UNIVERSITY OF PATRAS

School of Engineering
Department of Mechanical Engineering & Aeronautics
Laboratory for Manufacturing Systems & Automation

PhD Dissertation

A Method for the Generation of Assembly Information from Product Design:

Applications in Automotive Industry

GEORGIOS S. PINTZOS
Mechanical & Aeronautical Engineer

PATRAS 2017
 UNIVERSITY OF PATRAS
School of Engineering
Department of Mechanical Engineering & Aeronautics
Laboratory for Manufacturing Systems & Automation

PhD Dissertation

A METHOD FOR THE GENERATION OF ASSEMBLY INFORMATION FROM PRODUCT DESIGN:

APPLICATIONS IN AUTOMOTIVE INDUSTRY

GEORGIOS S. PINTZOS
Mechanical & Aeronautical Engineer

PATRAS 2017
iv
Η παρούσα Διδακτορική Διατριβή εγκρίθηκε ομόφωνα με βαθμό «ΑΡΙΣΤΑ» στις 25
Σεπτεμβρίου 2017, από την επταμελή εξεταστική επιτροπή η οποία ορίστηκε από την γενική
συνέλευση με Ειδική Σύνθεση του Τμήματος Μηχανολόγων και Αεροναυπηγών Μηχανικών της
Πολυτεχνικής Σχολής του Πανεπιστημίου Πατρών, στην υπ΄αριθμόν 15/27-6-2017
συνεδρίαση, σύμφωνα με τον άρθρο 12 Ν. 2083/92, μετά από σχετική εισήγηση της Τριμελούς
Συμβουλευτικής Επιτροπής.

Μέλη της Επταμελούς Εξεταστικής Επιτροπής αποτέλεσαν οι:

1. Μούρτζης Δημήτριος (Επιβλέπων)
Αναπληρωτής Καθηγητής του Τμήματος Μηχανολόγων και Αεροναυπηγών Μηχανικών
του Πανεπιστημίου Πατρών
2. Χρυσολούρης Γεώργιος (Μέλος της Τριμελούς Επιτροπής)
Καθηγητής του Τμήματος Μηχανολόγων και Αεροναυπηγών Μηχανικών του
Πανεπιστημίου Πατρών
3. Δέντσορας Αργύρης (Μέλος της Τριμελούς Επιτροπής)
Καθηγητής του Τμήματος Μηχανολόγων και Αεροναυπηγών Μηχανικών του
Πανεπιστημίου Πατρών
4. Ανυφαντής Νικόλαος
Καθηγητής του Τμήματος Μηχανολόγων και Αεροναυπηγών Μηχανικών του
Πανεπιστημίου Πατρών
5. Καρακαπιλίδης Νικόλαος
Καθηγητής του Τμήματος Μηχανολόγων και Αεροναυπηγών Μηχανικών του
Πανεπιστημίου Πατρών
6. Αδαμίδης Εμμανουήλ
Αναπληρωτής Καθηγητής του Τμήματος Μηχανολόγων και Αεροναυπηγών Μηχανικών
του Πανεπιστημίου Πατρών
7. Τατσιόπουλος Ηλίας
Καθηγητής της Σχολής Μηχανολόγων Μηχανικών του Εθνικού Μετσόβιου
Πολυτεχνείου

v
 ΠΡΑΚΤΙΚΟ ΣΥΝΕΔΡΙΑΣΗΣ
Επταμελούς Εξεταστικής Επιτροπής για την κρίση της διδακτορικής διατριβής του
κ. ΓΕΩΡΓΙΟΥ Σ. ΠΙΝΤΖΟΥ

Η Τριμελής Συμβουλευτική Επιτροπή, αποτελούμενη από τους Μούρτζη Δημήτριο,

Αναπληρωτή Καθηγητή του Τμήματος Μηχανολόγων και Αεροναυπηγών Μηχανικών του
Πανεπιστημίου Πατρών (επιβλέποντα), Χρυσολούρη Γεώργιο Καθηγητή του Τμ. Μηχανολόγων
και Αεροναυπηγών Μηχανικών του Πανεπιστημίου Πατρών και Δέντσορα Αργύρη Καθηγητή
του Τμ. Μηχανολόγων και Αεροναυπηγών Μηχανικών του Πανεπιστημίου Πατρών,
συμπληρώθηκε σε Επταμελή Εξεταστική Επιτροπή με τους Νικόλαο Ανυφαντή Καθηγητή του
Τμ. Μηχανολόγων και Αεροναυπηγών Μηχανικών του Πανεπιστημίου Πατρών, Καρακαπιλίδη
Νικόλαο Καθηγητή του Τμ. Μηχανολόγων και Αεροναυπηγών Μηχανικών του Πανεπιστημίου
Πατρών, Αδαμίδη Εμμανουήλ Αναπληρωτή Καθηγητή του Τμήματος Μηχανολόγων και
Αεροναυπηγών Μηχανικών του Πανεπιστημίου Πατρών και Τατσιόπουλο Ηλία Καθηγητή της
Σχολής Μηχανολόγων Μηχανικών του Εθνικού Μετσόβιου Πολυτεχνείου σύμφωνα με την υπ’
αριθμ. 15/27-6-2017 συνεδρίαση της Γενικής Συνέλευσης με την Ειδική Σύνθεση του Τμήματος
Μηχανολόγων & Αεροναυπηγών Μηχανικών του Πανεπιστημίου Πατρών.
Η Εξεταστική Επιτροπή συνήλθε σήμερα 25 Σεπτεμβρίου 2017 και ώρα 16:00 σε
συνεδρίαση κατά την οποία η διατριβή με τίτλο “Μέθοδος για την παραγωγή πληροφοριών
συναρμολόγησης κατά το σχεδιασμό προϊόντων: εφαρμογές στο χώρο της
αυτοκινητοβιομηχανίας” υποστηρίχθηκε από τον υποψήφιο διδάκτορα κ. Γεώργιο Σ. Πίντζο.
Στην συνέχεια υπεβλήθησαν ερωτήσεις στον υποψήφιο και επερατώθει η εξέταση.
Μετά την αποχώρηση του υποψηφίου και των ακροατών, η επιτροπή συζήτησε επί της
γενομένης εξετάσεως και αποφάσισε ομόφωνα να κάνει δεκτή τη διατριβή, επειδή αυτή είναι
πρωτότυπη και αποτελεί ουσιαστική συμβολή στην επιστήμη, ο δε υποψήφιος την
υπερασπίστηκε με άκρως ικανοποιητικό τρόπο και τον βαθμολόγησαν με βαθμό: ΑΡΙΣΤΑ

vi
Στη μνήμη της μητέρας μου

vii
 ΕΥΧΑΡΙΣΤΙΕΣ

Όντας μέλος του Εργαστηρίου Συστημάτων Παραγωγής και Αυτοματισμού για επτά χρόνια, αρχικά ως
προπτυχιακός και μετέπειτα ως μεταπτυχιακός φοιτητής, είχα την δυνατότητα να ασχοληθώ με ένα
εύρος ερευνητικών προβλημάτων που αφορούσαν κυρίως στο σχεδιασμό προϊόντων και οργάνωσης
συστημάτων παραγωγής. Σε αυτά τα χρόνια, μου δόθηκε η ευκαιρία να εργαστώ ως ερευνητής
μηχανικός σε ευρωπαϊκά προγράμματα, στο πλαίσιο των οποίων συνεργάστηκα με πάρα πολλούς
ερευνητές και μηχανικούς από διαφορετικές χώρες της Ευρώπης. Αυτή η εμπειρία θεωρώ πως υπήρξε
καθοριστική στη διαμόρφωσή μου ως μηχανικό. Για όλα αυτά οφείλω να εκφράσω την ευγνωμοσύνη
μου στον καθηγητή κ. Γεώργιο Χρυσολούρη, Καθηγητή του Τμήματος Μηχανολόγων & Αεροναυπηγών
Μηχανικών, που μου προσέφερε τις ευκαρίες αυτές καθώς και για την πολύτιμη καθοδήγησή του.

Επιπροσθέτως, θα ήθελα να ευχαριστήσω τον επιβλέποντα μου κ. Δημήτριο Μούρτζη, Αναπληρωτή

Καθηγητή του Τμήματος Μηχανολόγων & Αεροναυπηγών Μηχανικών καθώς και τον κ. Αργύριο
Δέντσορα, Καθηγητή του Τμ. Μηχανολόγων & Αεροναυπηγών Μηχανικών του Πανεπιστημίου Πατρών
για τη συμβολή τους στη διαμόρφωση αυτής της εργασίας. Ευχαριστώ επίσης τα μέλη της επταμελούς
επιτροπής κ. Νικόλαο Ανυφαντή, Καθηγητή του Τμ. Μηχανολόγων & Αεροναυπηγών Μηχανικών του
Πανεπιστημίου Πατρών, κ. Νικόλαο Καρακαπιλίδη, Καθηγητή του Τμ. Μηχανολόγων & Αεροναυπηγών
Μηχανικών του Πανεπιστημίου Πατρών, κ. Εμμανουήλ Αδαμίδη, Αναπληρωτή Καθηγητή του Τμήματος
Μηχανολόγων & Αεροναυπηγών Μηχανικών του Πανεπιστημίου Πατρών και κ. Ηλία Τατσιόπουλο
Καθηγητή του Τμ. Μηχανολόγων Μηχανικών του Εθνικού Μετσόβιου Πολυτεχνείου για τη συνεισφορά
τους στην τελική διαμόρφωση της εργασίας.

Επίσης θα ήθελα να εκφράσω τις ευχαριστίες μου στους Δρ. Νικόλαο Παπακώστα, Δρ. Κοσμά
Αλεξόπουλο, κ. Χρήστο Γιαννούλη και κ. Κ. Σμπαρούνη για τις χρήσιμες και πολύτιμες συμβουλές τους.
Επίσης ευχαριστώ τις γραμματείς του Εργαστηρίου κα. Άντζελα Σμπαρούνη και την κα. Βασιλική
Βελαώρα για την υποστηριξή τους.

Ένα μεγάλο ευχαριστώ στους Δρ. Κωνσταντίνο Ευθυμίου και Δρ. Λουκά Ρέντζο για την καθοδήγηση και
υποστήριξη τους στα πρώτα χρόνια της εκπόνησης αυτής της διατριβής. Επίσης θα ήθελα να
ευχαριστήσω τους φίλους και συναδέλφους μου Δρ. Μιχάλη Δούκα, κ. Μάρκο Μάτσα, κ. Νίκο Μήλα
και κα. Κατερίνα Βλάχου για τη συνεργασία μας αλλά και για την πολύτιμη συμπαράσταση τους.

Εν συνεχεία δεν θα μπορούσα να μην ευχαριστήσω τον πάτερα μου Σωτήριο Πίντζο για την πίστη του
στις ικανότητές μου και για το ότι μου μετέδωσε την αγάπη του για την επιστήμη, καθώς και την κοπέλα

viii
μου Δήμητρα Μπερνιδάκη για την υπομονή και υποστήριξη της όλα αυτά τα χρόνια. Επίσης τους φίλους
μου Σταύρο Λιόση, Σοφία Σταματάτου, Βασίλειο Κόλια, Ισίδωρο Κρίσιλια και Βασίλειο Τριγλιανό για την
υποστήριξη τους και την πίστη τους σε εμένα.

Τέλος, δυστυχώς δεν μπορώ να εκφράσω με λέξεις πόσο ευγνώμων είμαι στη μητέρα μου Μαρία
Γιάννη-Πίντζου στη μνήμη της οποίας αφιερώνω αυτή τη διατριβή, για την ανιδιοτελή της αγάπη και
για τη δύναμη που ακόμα αντλώ από αυτήν.

Γεώργιος Σ. Πίντζος
Πάτρα, Σεπτέμβριος 2017

ix
 A Method for the Generation of Assembly Information from
Product Design: Applications in Automotive Industry

SUMMARY
Assembly planning methods have been in the centre of industrial and academic research for many
decades, since manual assembly costs may often account for even half of the total manufacturing
expenses. Existing and emerging manufacturing trends, such as mass customization and personalization,
require fast responses when it comes to the conception and realization of the relevant manufacturing
systems. Although, work methodologies have been proposed and applied, such as concurrent
engineering, gaps still exist between product development and manufacturing. Current Product
Lifecycle Management (PLM) systems focus on the coordination of activities among engineers of
different disciplines without, however, being capable of providing actual decision support functionality
to decision makers. Moreover, solutions for the different phases of assembly planning have been
proposed, without however taking into account the holistic nature of assembly planning that spans the
different engineering phases.

In this work, a method is proposed to enable the generation of assembly information, relevant to
different phases of a product’s lifecycle, from product data available during its design. The approach is
accompanied by three applications focusing on assembly planning, assembly execution and End of Life
(EoL) product handling. The core method is an algorithm that generates assembly precedence relations
between all components of a product. The algorithm uses collision / interference detection as well as
constraints and information extraction as the main means to identify the assembly relations. The results
of the algorithm are then used in three different applications: End of Life value estimation, Assembly
Line Balancing and Simulation and Assembly Instructions Generation. The applications target both
engineers that work in overlapping phases of product and production development as well as the
manual assembly operators during execution of the actual processes.

The Assembly Precedence Diagram Generation (APDG) algorithm was developed as an add-on to a
widely-used CAD software. The developed application has a minimum input requirement which
concerns the selection of the base part(s) as well as configuring the algorithm’s performance through
different options. The performance of the algorithm has been tested by data provided by the
automotive industry. The rest of the applications were developed using three different approaches.
More specifically, the applications relevant to assembly line planning were developed as part of a
collaborative web-platform that can also extract and use information residing in engineering files. The

x
assembly instruction generation and visualisation was developed as a separate module and an EoL
indicators calculator was developed as a macro application. The development of the applications
relevant to assembly line planning involved the elaboration of a new method integrating Discrete Event
Simulation (DES) with Assembly Line Balancing (ALB). All proposed applications have been tested on the
basis of industrial data and the respective results are provided using time and cost related indicators.
The results highlight the benefits of applying the developed methods in industrial environments. More
specifically, the effectiveness of the APDG in generating assembly information for use in both production
planning and EoL evaluation of products is presented, as well as the added value of the proposed
integrated approach (assembly line balancing and simulation) during production design.

xi
 ΜΕΘΟΔΟΣ ΓΙΑ ΤΗΝ ΠΑΡΑΓΩΓΗ ΠΛΗΡΟΦΟΡΙΩΝ ΣΥΝΑΡΜΟΛΟΓΗΣΗΣ ΚΑΤΑ ΤΟ
ΣΧΕΔΙΑΣΜΟ ΠΡΟΪΟΝΤΩΝ: ΕΦΑΡΜΟΓΕΣ ΣΤΟ ΧΩΡΟ ΤΗΣ
ΑΥΤΟΚΙΝΗΤΟΒΙΟΜΗΧΑΝΙΑΣ

ΠΕΡΙΛΗΨΗ
Οι μέθοδοι σχεδιασμού και προγραμματισμού συστημάτων συναρμολόγησης βρίσκονται στο κέντρο
της βιομηχανικής και ακαδημαϊκής έρευνας εδώ και πολλές δεκαετίες, κυρίως λόγω του κόστους της
χειρωνακτικής συναρμολόγησης, που μπορεί να αφορά ακόμη και τις μισές από τις συνολικές δαπάνες
παραγωγής ενός προϊόντος. Οι υπάρχουσες και αναδυόμενες βιομηχανικές τάσεις, όπως η μαζική
προσαρμογή και εξατοµίκευση προϊόντων (mass product customization and personalization), απαιτούν
μικρούς χρόνους για τη λήψη αποφάσεων υλοποίησης των σχετικών συστημάτων παραγωγής. Παρόλο
που έχουν προταθεί και εφαρμοστεί μεθοδολογίες όπως ο παράλληλος τεχνικός σχεδιασµός
(concurrent engineering), υπάρχουν ακόμα κενά μεταξύ του σχεδιασμού ενός προϊόντος και της
παραγωγής του. Τα υπάρχοντα συστήματα διαχείρισης του κύκλου ζωής των προϊόντων (Product
Lifecycle Management systems) εστιάζουν στο συντονισμό των δραστηριοτήτων μεταξύ μηχανικών
διαφορετικών ειδικοτήτων, χωρίς όμως να υποστηρίζουν άμεσα τη λήψη αποφάσεων από τους
ιθύνοντες μηχανικούς. Επιπλέον, αρκετές λύσεις που έχουν προταθεί και αφορούν τις διαφορετικές
φάσεις σχεδιασμού των συστημάτων παραγωγής, δεν λαμβάνουν υπόψη την ολιστική φύση του
σχεδιασμού και άρα τις επιμέρους φάσεις που τον αποτελούν. Κατά συνέπεια, οι υπεύθυνοι για το
σχεδιασμό και την ανάπτυξη προϊόντων εξαρτώνται ακόμα από τις ειδικές γνώσεις των μηχανικών που
ασχολούνται με τις μετέπειτα φάσεις του κύκλου ζωής ενός προϊόντος.

Η εν λόγω διατριβή προτείνει μια μέθοδο παραγωγής πληροφοριών συναρμολόγησης, που αφορούν
διαφορετικές φάσεις του κύκλου της ζωής ενός προϊόντος, από τα αρχεία σχεδίου των προϊόντων (CAD
files). Η προσέγγιση αυτή συνοδεύεται από τρεις εφαρμογές που εστιάζουν στο σχεδιασμό
συστημάτων συναρμολόγησης, στην εκτέλεση της συναρμολόγησης και στην ανακύκλωση ή
επαναχρησιμοποίηση του προϊόντος, ή τμημάτων αυτού, στο τέλος της ζωής του. Η κύρια μέθοδος
αποτελείται από έναν αλγόριθμο που προσδιορίζει τις σχέσεις προτεραιότητας μεταξύ όλων των
συστατικών στοιχείων ενός προϊόντος (υποσυναρμογές, κομμάτια και σύνδεσμοι). Ο αλγόριθμος κάνει
χρήση της ανίχνευσης σύγκρουσης (collision detection) μεταξύ των στοιχείων, καθώς επίσης και των
περιορισμών και άλλων συναφών πληροφοριών που έχουν οριστεί από τους σχεδιαστές του προϊόντος.
Έχοντας αυτά σαν κύρια μέσα, ο αλγόριθμος προσδιορίζει τις πιθανές ακολουθίες συναρμολόγησης.
Τα αποτελέσματα του αλγορίθμου χρησιμοποιούνται στην συνέχεια από τρεις διαφορετικές
εφαρμογές: εφαρμογή εκτίμησης αξίας στο τέλος ζωής (End of Life value estimation), εφαρμογή

xii
εξισορρόπησης και προσομοίωσης γραμμών συναρμολόγησης και εφαρμογή παραγωγής οδηγιών
συναρμολόγησης (Assembly Instructions Generation). Οι εφαρμογές αφορούν μηχανικούς που
εργάζονται στις φάσεις του σχεδιασμού και της ανάπτυξης προϊόντος και γραμμών συναρμολόγησης
καθώς και τους χειριστές κατά τη διάρκεια της εκτέλεσης των πραγματικών διαδικασιών.

Ο αλγόριθμος παραγωγής διαγραμμάτων προτεραιότητας συναρμολόγησης (Assembly Precedence

Diagram Generation - APDG) αναπτύχθηκε ως πρόσθετη εφαρμογή (add-on) σε ένα ευρέως
χρησιμοποιούμενο σύστημα σχεδιασμού (CAD). Η απόδοση του αλγορίθμου έχει εξεταστεί με τη χρήση
δεδομένων που έχουν παραχωρηθεί από την αυτοκινητοβιομηχανία. Οι υπόλοιπες εφαρμογές
αναπτύχθηκαν χρησιμοποιώντας τρεις διαφορετικές προσεγγίσεις. Πιο συγκεκριμένα, οι εφαρμογές
σχετικές με τον σχεδιασμό γραμμών συναρμολόγησης αναπτύχθηκαν ως τμήματα μιας πλατφόρμας
συνεργασίας που μπορεί επίσης να εξαγάγει και να χρησιμοποιήσει δεδομένα που βρίσκονται σε
αρχεία τεχνικού σχεδιασμού (CAx files). Η εφαρμογή παραγωγής και απεικόνισης οδηγιών
συναρμολόγησης αναπτύχθηκε ως ανεξάρτητη εφαρμογή ενώ η εφαρμογή αξιολόγησης του προϊόντος
για ανακύκλωση ή επαναχρησιμοποίηση αναπτύχθηκε ως μακρο-εφαρμογή (macro). Η ανάπτυξη των
εφαρμογών σχετικών με το σχεδιασμό γραμμών συναρμολόγησης συμπεριλαμβάνει την ανάπτυξη μιας
νέας μεθόδου ενσωμάτωσης μεθόδων εξισορρόπησης γραμμών συναρμολόγησης με μεθόδους
προσομοίωσης αυτών. Όλες οι αναπτυγμένες εφαρμογές ελέγχθηκαν χρησιμοποιώντας δεδομένα από
την βιομηχανία και τα αποτελέσματά τους παρουσιάζονται με την χρήση δεικτών που εκφράζουν χρόνο
και κόστος.

xiii
TABLE OF CONTENTS

ABBREVIATIONS ................................................................................................................................................... XVII

LIST OF FIGURES .................................................................................................................................................... XIX

LIST OF TABLES ...................................................................................................................................................... XXI

1 INTRODUCTION ................................................................................................................................... 23

1.1 CHALLENGES ..................................................................................................................................................23

1.2 PROBLEM DEFINITION ......................................................................................................................................26
1.3 DESCRIPTION OF THE APPROACH .......................................................................................................................27
1.4 DISSERTATION ROADMAP.................................................................................................................................28

2 STATE OF THE ART ............................................................................................................................... 31

2.1 ASSEMBLY INFORMATION IN PRODUCT FILES.......................................................................................................31

2.2 METHODS FOR ASSEMBLY GRAPH GENERATION ..................................................................................................33
2.3 END OF LIFE PRODUCTS VALUE ESTIMATION ........................................................................................................38
2.4 ASSEMBLY LINE BALANCING AND SIMULATION ....................................................................................................40
2.5 ASSEMBLY INSTRUCTION INTERFACES .................................................................................................................44
2.6 RESEARCH CONTRIBUTION OF THE THESIS............................................................................................................48

3 METHODS DESIGN AND DEVELOPMENT .............................................................................................. 51

3.1 OVERVIEW OF THE PROPOSED APPROACH ..........................................................................................................51

3.1.1 Modeling of the problem and identifies solutions...................................................................................51
3.1.2 Developed tools infrastructure .................................................................................................................54
3.2 PRECEDENCE DIAGRAM GENERATION ALGORITHM ..............................................................................................56
3.2.1 User input and CAD requirements ............................................................................................................57
3.2.2 Separation of components ........................................................................................................................59
3.2.3 Fasteners removal and first collision detection .......................................................................................59
3.2.4 User input for singular extraction paths directions .................................................................................63
3.2.5 Assembly tiers and associated fasteners ..................................................................................................63
3.2.6 Association of components belonging to sequential tiers ......................................................................66
3.3 ASSESSMENT OF PRODUCTS FOR END OF LIFE SEPARATION AND MATERIAL HANDLING .............................................67
3.3.1 Collision detection algorithm and components’ information .................................................................68

xv
3.3.2 Disassembly cost calculation .................................................................................................................... 69
3.4 COLLABORATION SOFTWARE ............................................................................................................................ 71
3.4.1 Extended database .................................................................................................................................... 72
3.4.2 Information extraction application ........................................................................................................... 77
3.4.3 Assembly line balancing and simulation .................................................................................................. 79
3.5 DESIGN OF THE ASSEMBLY INSTRUCTIONS INTERFACE............................................................................................ 84

4 IMPLEMENTATION............................................................................................................................... 87

4.1 APPLICATIONS AND TOOLS OVERVIEW ................................................................................................................ 87

4.2 ENGINEERING ASSISTANCE APPLICATIONS............................................................................................................ 87
4.2.1 Assembly / Disassembly Precedence Diagram Generation add-on ........................................................ 87
4.2.2 EFRR calculation application ..................................................................................................................... 90
4.2.3 Assembly Instructions Generation and Visualization application ........................................................... 91
4.3 COLLABORATION ASSISTANCE APPLICATIONS ....................................................................................................... 94

5 CASE STUDIES AND RESULTS ................................................................................................................ 99

5.1 GENERATION OF PRECEDENCE DIAGRAMS FOR DIFFERENT PRODUCT DESIGN FILES ..................................................... 99
5.2 END OF LIFE VALUE ESTIMATION AND COMPARISON BETWEEN DIFFERENT PRODUCT DESIGNS....................................102
5.3 ASSEMBLY LINE PLANNING..............................................................................................................................104
5.4 ASSEMBLY INSTRUCTIONS GENERATION AND VISUALISATION USING AUGMENTED REALITY GOGGLES ..........................110

6 CONCLUSIONS AND DISCUSSION ....................................................................................................... 115

6.1 DISCUSSION ON THE RESULTS..........................................................................................................................115

6.2 LIMITATIONS OF THE PROPOSED APPROACH ......................................................................................................117
6.3 OUTLOOK AND INDICATIONS FOR FUTURE CONTINUATION OF THE PRESENTED WORK...............................................119

7 REFERENCES ...................................................................................................................................... 123

BRIEF BIOGRAPHICAL NOTE – PUBLICATIONS ....................................................................................................... 131

xvi
 ABBREVIATIONS

3D : Three Dimensional
ALB : Assembly Line Balancing
ALBA : Assembly Line Balancing Agent
ALBP : Assembly Line Balancing Problem
APDG : Assembly Precedence Diagram Generation
AR : Augmented reality
BOM : Bill of Materials
BOP : Bill of Processes
CAD : Computer-Aided Design
CAM : Computer-Aided Manufacturing
CAx : Computer-Aided engineering applications
CPD : Collaborative Product Development
CT : Cycle Time
DES : Discrete Event Simulation
DESA : Discrete Event Simulation Agent
DPF : Diesel Particulate Filter
DSM : Design Structure Matrix
DTO : Data Transfer Object
EFRR : Economic Feasibility for Recycling or Reuse
EoL : End of Life
GALBP : General Assembly Line Balancing Problem
HMD : Head Mounted Displays
ICT : Information and Communication Technologies
IPT : Integrated Product Team
IT : Information Technology

xvii
ABBREVIATIONS

MAS : Multi-Agent System

MES : Manufacturing Execution Systems
MMALBP : Mixed-Model Assembly Line Balancing Problem
MTBF : Mean Time between Failures
MTM : Methods Times Measurements
MTTR : Mean Time to Repair
MuMALBP : Multi-Model Assembly Line Balancing Problem
PC : Personal Computer
PD : Precedence Diagram
PLM : Product Lifecycle Management
PPR : Product-Process-Resource
PSO : Particle Swarm Optimization
SALBP : Simple Assembly Line Balancing Problem
SDK : Software Development Kit
UI : User Interface
UML : Unified Modelling Language
VR : Virtual Reality
WC : Work Centre (or Work Station)
XML : Extensible Markup Language

xviii
 LIST OF FIGURES

FIGURE 1: PRODUCT’S LIFECYCLE PHASES INFLUENCED BY THE PROPOSED APPROACH....................................................................27

FIGURE 2: ASSEMBLY AND DISASSEMBLY OPERATIONS IN THE LIFE CYCLE OF TECHNICAL PRODUCTS [3]............................................31
FIGURE 3: COMMON CONSTRAINT DEFINITIONS IN CAD SOFTWARE. ..........................................................................................33
FIGURE 4: (A) ASSEMBLY PRECEDENCE GRAPH OF A SIMPLE DESK ASSEMBLY AND (B) THREE POSSIBLE SEQUENCES THAT CAN BE
EXTRACTED FROM THE PRECEDENCE GRAPH...................................................................................................................34

FIGURE 5: METHODS FOR INFORMATION GENERATION BASED ON THEIR OUTPUT. ........................................................................37

FIGURE 6: ASSEMBLY LINE BALANCING PROBLEM AND SOLUTION METHODS CLASSIFICATION. ........................................................41
FIGURE 7: SIMPLE ALBP EXAMPLE SOLUTION. ........................................................................................................................42
FIGURE 8: NECESSARY INFORMATION THAT HAVE TO BE GATHERED TO BUILD A MANUAL ASSEMBLY LINE MODEL. ..............................44
FIGURE 9: MANUAL ASSEMBLY INSTRUCTIONS USING (A, B) MONITORS AND (C) PAPER. ................................................................45
FIGURE 10: DEVICE ASSESSMENT FOR SHOPFLOOR APPLICATIONS...............................................................................................47
FIGURE 11: STEPS OF THE PRODUCT’S LIFE AND RELATED DEVELOPMENTS. ..................................................................................52
FIGURE 12: SOFTWARE NEED DISTRIBUTION IN A COMMON PROJECT WITH AN OEM. ..................................................................53
FIGURE 13: OVERVIEW OF THE PROPOSED INFRASTRUCTURE. ....................................................................................................54
FIGURE 14: APPLICATIONS WITH REFERENCE TO THE PROPOSED INFRASTRUCTURE. ......................................................................55
FIGURE 15: ASSEMBLY TIERS FROM THE DESK EXAMPLE. ...........................................................................................................56
FIGURE 16: OVERVIEW OF ALGORITHM’S STEPS.......................................................................................................................57
FIGURE 17: EXPLODED VIEW OF THE DIFFERENTIAL ASSEMBLY. ..................................................................................................58
FIGURE 18: COLLISION DETECTION ALGORITHM. ......................................................................................................................60
FIGURE 19: OUTPUT OF THE FIRST COLLISION DETECTION STEP. .................................................................................................62
FIGURE 20: OUTPUT OF THE FIRST COLLISION DETECTION STEP FOR THE DIFFERENTIAL. .................................................................63
FIGURE 21: ALGORITHM FOR FASTENER ASSOCIATION TO COMPONENTS. ....................................................................................64
FIGURE 22: TRANSLATION OF DISASSEMBLY TIERS TABLE TO ASSEMBLY TIERS. .............................................................................65
FIGURE 23: THE TRANSLATED DISASSEMBLY TIERS TABLE TO ASSEMBLY TIERS FOR THE DIFFERENTIAL. ..............................................65
FIGURE 24: FINAL PRECEDENCE DIAGRAM TABLE CONTAINING ASSOCIATED FASTENERS AND INSERTION AXES. ...................................66
FIGURE 25: FINAL PRECEDENCE DIAGRAM TABLE OF THE DIFFERENTIAL. ......................................................................................67
FIGURE 26: REPRESENTATION OF THE PRECEDENCE DIAGRAM GENERATED FOR THE DIFFERENTIAL. .................................................67
FIGURE 27: INFORMATION AVAILABLE AT THE END OF COLLISION DETECTION. ..............................................................................69
FIGURE 28: APPLICATIONS REGARDING COLLABORATION. .........................................................................................................72
FIGURE 29: BASIC CLASSES OF THE PROPOSED DATA MODEL AND THEIR INTERRELATIONSHIPS.........................................................76
FIGURE 30: CLASS TYPE TAXONOMY. .....................................................................................................................................77

xix
LIST OF FIGURES

FIGURE 31: ENGINEERING DATA OBJECTS’ STRUCTURE. ........................................................................................................... 77

FIGURE 32: PROCESS SIMULATION FILE TOPOLOGY. ................................................................................................................. 78
FIGURE 33: EXAMPLE OF A PRECEDENCE DIAGRAM CONTAINING ALTERNATIVE CONFIGURATIONS. .................................................. 80
FIGURE 34: GENERATION OF LINE ALTERNATIVES BY THE ALB AGENT. ........................................................................................ 82
FIGURE 35: GENERATION OF A SIMULATION MODEL BY THE DESA. ............................................................................................ 83
FIGURE 36: ASSEMBLY OPERATIONS BREAKDOWN. .................................................................................................................. 84
FIGURE 37: INFORMATION USED FOR GENERATION OF ASSEMBLY INSTRUCTIONS FROM APDG OUTPUT. ........................................ 85
FIGURE 38: DEFINITION OF START POSITION FOR PATH CREATION. ............................................................................................. 86
FIGURE 39: VBSCRIPT EDITOR IN CATIA. .............................................................................................................................. 88
FIGURE 40: INPUT AND MESSAGE BOXES IN CATIA. ................................................................................................................. 88
FIGURE 41: EFRR CALCULATOR. ........................................................................................................................................... 90
FIGURE 42: IDENTIFICATION OF COORDINATE SYSTEM THROUGH IMAGE TARGET.......................................................................... 92
FIGURE 43: ASSEMBLY INSTRUCTION APPLICATION STRUCTURE. ................................................................................................ 92
FIGURE 44: EXAMPLE OF A PART SUPERIMPOSED ON A BASE PART. ............................................................................................. 93
FIGURE 45: MANUFACTURING PROCESS INFORMATION AGENT AND PPR DATABASE DEPLOYMENT. ................................................ 94
FIGURE 46: THREE-LAYER IMPLEMENTATION OF THE AGENTS. ................................................................................................... 96
FIGURE 47: GUI OF THE ALBA AND DESA AGENTS................................................................................................................. 96
FIGURE 48: PRODUCT DESIGN FILES USED IN EXPERIMENTS. ...................................................................................................... 99
FIGURE 49: DISASSEMBLY DIRECTIONS FOR “CAP” COMPONENT IN DIFFERENTIAL ASSEMBLY. ......................................................101
FIGURE 50: END OF LIFE HANDLING SCENARIO. .....................................................................................................................102
FIGURE 51: OLD (A) AND NEW (B) REAR AXLE ASSEMBLIES’ CHARACTERISTICS. ...........................................................................103
FIGURE 52: EXCERPT FROM PRECEDENCE DIAGRAM OF OLD AXLE CAD. ....................................................................................104
FIGURE 53: FOCUS OF THE CASE STUDY IN A LINE DESIGN PROJECT. ..........................................................................................105
FIGURE 54: EXTRACT FROM THE PRECEDENCE DIAGRAM OF THE PROCESSES OF THE CASE STUDY...................................................106
FIGURE 55: USER INTERFACE OF THE WEB-APPLICATION FOR THE CONTROL OF THE DEVELOPED AGENTS. .......................................108
FIGURE 56: RUNNING COSTS, PRODUCTION RATE AND CYCLE TIME OF THE FIVE BEST ALTERNATIVES GENERATED BY THE ALBA AND
DESA. ..................................................................................................................................................................110
FIGURE 57: DPF CAD ASSEMBLY. ......................................................................................................................................111
FIGURE 58: HARDWARE USED IN THE CASE STUDY. ................................................................................................................112
FIGURE 59: (A) DPF FULLY ASSEMBLED ON THE SUPPORT STRUCTURE AND (B) DPF WITHOUT COVER PARTS. .................................112
FIGURE 60: ANIMATION OF THE CASE BEING ASSEMBLED ON THE DPF USING AR.......................................................................113
FIGURE 61: DYNAMIC RELATION BETWEEN RESULTS OF SUB-STEPS AND INPUT OF INITIAL SUB-STEPS. ............................................117
FIGURE 62: EXAMPLE OF CYLINDER GEOMETRY INFORMATION AS STORED IN PROPRIETARY CAD FORMAT AND AS STEP. ................118

xx
 LIST OF TABLES

TABLE 1: COMPONENTS AND FASTENERS EXTRACTION TIMES. ....................................................................................................70

TABLE 2: SENSITIVITY FACTOR FOR DIFFERENT MATERIALS. ........................................................................................................89
TABLE 3: CHARACTERISTIC OF PRODUCT DESIGN FILES............................................................................................................ 100
TABLE 4: STEP USED AND RESULTS FOR EACH PRODUCT DESIGN. ............................................................................................. 101
TABLE 5: EFRR RESULTS OF THE ALGORITHM. ...................................................................................................................... 103
TABLE 6: CASE STUDY SPECIfiCATIONS. ................................................................................................................................ 107
TABLE 7: FIVE BEST RESULTS GENERATED BY THE DESA. ........................................................................................................ 109

xxi
INTRODUCTION

1 INTRODUCTION

In this chapter an introduction to the scope and content of this dissertation is given. The
chapter begins with the description of the current methods and practices of the industry, as
well as the academic work concerning assembly information generation, planning and lifecycle
management methods.

1.1 Challenges

In Europe, approximately 12.9 million people work in the automotive sector. This means that
5.3% of the EU employed population is directly or indirectly professionally occupied in the
automotive industry [1]. The total turnover of the industry in 2012 was 843.4 €bn which
accounts for 6.6% of EU27 GDP [2]. It has become obvious that the profit of the automotive
manufacturers is becoming increasingly dependent on the design and final assembly business
operations [3]. Assembly is a manufacturing process in which components are added to a
product in order to achieve the optimally planned logistics to create a finished product in the
fastest possible way [4]. In the industry, more than 50% of direct labour costs are linked to the
assembly of the products [5]. This number illustrates the importance of optimizing manual
assembly plans and their design / configuration in an industry that gradually moves towards
mass personalization. Beyond this trend, modern manufacturing organisations must increase
the effectiveness of manual processes in view of shrinking natural resources [6].

Assembly is usually carried out manually and as a result it can cause even half of the total
manufacturing expenses. Manual assembly processes are affected by a number of factors, the
combination of which dramatically affects the efficiency of the processes by altering results
such as the frequency that a finished product or sub-assembly is completed [7]. This is also the

23
INTRODUCTION

case in automated or semi-automated assembly systems where alternative process sequences

are evaluated in order to optimize the lead time, cycle time and costs [8].

The first step in designing an assembly system is the formal documentation / representation of
the assembly processes and their respective constraints. The design of an assembly system
requires methods to represent the assembly components and hierarchy, and to generate the
alternatives of the sequences of the assembly [9]. This can be a time consuming process,
especially for multi-variant products that allow different configurations by the customers. The
result of such a process, is essentially a Bill of Materials (BOM) that can have different forms
depending on the level of information that need to be communicated [10]. However, beyond
the list of components to be assembled, every other form of a BOM requires a certain effort in
order to be generated. Beyond the generation of this information for assembly systems’ design,
industrial practices also focus on its use for maintenance operations which include assembly
and disassembly processes. The current industrial practices regarding the handling of this
information is primarily performed through the use of Information Technology (IT) solutions
[11].

Information Technology (IT) is one of the most essential technological assets to any
contemporary manufacturing organization [4]. Technologies that enable computer-aided
engineering have formed the current situation, known as digital product and production
engineering [11]. This is characterized by a large number of different IT tools, based on web-
based and workstation applications. Digital product engineering refers to the processes in
which engineers use digital tools to design and develop a product. Product engineering usually
entails an activity dealing with issues of cost, manufacturability, quality, performance,
reliability, serviceability and user-related features. According to [12], digital product
engineering should provide the means leading to product designs that satisfy all functional
requirements as well as helping to reduce the product’s complexity. A relatively new paradigm
of digital product engineering is the Collaborative Product Development (CPD) [13]. CPD is an
internet-based computational architecture that supports managing, sharing and transferring
product knowledge and information among geographically distributed organizations in order
for engineering decisions to be made in collaborative environments [14]. The CPD applications
proposed in academic literature, span from product design and reviewing, down to product

24
INTRODUCTION

customization [15]. Digital production engineering refers to the processes that involve
engineering teams, using heterogeneous IT tools by which they collaboratively design and
implement a manufacturing system [11]. These teams can be part of a single organization or
different organizations collaborating towards a common goal. Approaches such as in [16],
propose specific workflow systems that enable the collaborative computer-aided production
engineering.

However, most commercial solutions that are available and provide the necessary
functionalities, support only proprietary engineering file formats except for standardized
graphic exchange ones, such as STEP. This creates a gap between different organisations that
are engaged in the same projects since the direct sharing of files and information is not possible
[17]. This can cause significant delays in product and production delivery times in an extended
enterprise context.

In parallel to the current practices regarding the assembly of products, recycling and reuse of
products and their components (which are more related to the disassembly of a product) are
not fully and systematically considered by contemporary product engineers and designers
during the design phases [18]. This is partly due to the fact that the designers do not have
standard tools to automatically assess their designs, since current CAD tools do not typically
provide any metrics regarding the recyclability of the designer’s choices. On the other hand,
CAD systems have grown from graphics editors with few built-in design functionalities, to
sophisticated tools with the ability to generate and provide additional information on the
characteristics of a designed product, such as parts’ collision and automatic fastener selection
[4].

Finally, beyond the cooperation of engineers, the communication between engineers and shop
floor personnel is still an issue that is frequently examined in academic research [19].
Technologies such as Augmented Reality (AR) have evolved and can provide efficient solutions
to assist mutable, human-related manufacturing tasks [20]. The combination of these
technologies with traditional methods can provide integrated solutions to be used in mass
customization oriented systems. Few methods for the integration of technologies such as AR

25
INTRODUCTION

to industrial practices have been proposed however, usually discounting the effort that has to
be given by a company’s engineers to realise them [21].

1.2 Problem definition

The challenges that contemporary companies face regarding assembly systems’ design and
realisation have been summarised above. It is clear that the problem affects most, if not all, the
business phases of assembly design and execution. The goal of the approach presented in this
work is to attempt to address the challenges belonging to the phase of assembly systems’
design and investigate different applications following this phase. Therefore, the definition of
the problem is done separately for each of the specific aforementioned challenges.

The key challenge, as mentioned in the previous section, is the lack of integration mechanisms
between product design and production design. More specifically, the design of an assembly
system requires methods to enable the generation of sequences of the assembly [9]. In order
to define the problem in a way that fully encompasses its dimensions in real life, we must first
consider the affected people / engineers that are involved in the respective processes. In
contemporary product design environments, Integrated Product Teams (IPTs) are formed in
order to design products that satisfy requirements coming from different phases of their
lifecycle [22]. During product design, assembly requirements can be viewed as links or gaps.
More explicitly, links are defined in the forms of mates and contacts [23]. From there, the
possible assembly sequences are established in order to create the first assembly Bill of
Processes (BOP). As stated previously, this can be a procedure which, when dealing with diverse
product designs, can be time consuming. Until recent times, this was sufficient during assembly
design, since the focus was more on the creation of value through manufacturing;
contemporary paradigms however, take into account the life cycle of technical products and
the optimisation of value and benefits, including disassembly for reuse or recycling [3].

Considering the above facts, the problems addressed in this work are summarized below:

 No method acceptable by industrial criteria for the generation of assembly information

from product designs.

26
INTRODUCTION

 Lack of a methodology to sustain the assembly information in extended engineering

environments through a product’s production design phases.
 Lack of methods to automatically calculate EoL indicators during the design of a
product.
 Lack of tools to enable the exchange and evolution of product and production
information regarding manual assembly.

1.3 Description of the approach

A method is proposed to enable the generation of assembly related information, relevant to

different phases of a product’s lifecycle, from the design files of products or sub-assemblies.
The method is demonstrated through three different applications relevant to the product’s
lifecycle. The core of the approach is an algorithm that generates assembly precedence
relations between components of a product and also identifies the fasteners relevant to each
component. The algorithm uses collision detection as well as information extraction (mass,
material, assembly constraints) as the main means to identify the assembly relations. Figure 1
shows the core method of assembly information generation and which phases of a product’s
lifecycle it can feed this information to, namely; product’s assembly line planning, assembly
execution and End of Life handling, in the order that appearing in a product’s lifecycle.

Figure 1: Product’s lifecycle phases influenced by

the proposed approach.

27
INTRODUCTION

As mentioned earlier, for each of these phases an application has been developed in which uses
of the method are demonstrated. The applications concern both engineers that work in
overlapping phases of product and production development as well as the manual assembly
operators during execution of the actual processes: Integrated Assembly Line Balancing and
Simulation, Assembly Instructions Generation and Visualisation and quantification of the
product’s End of Life value.

1.4 Dissertation Roadmap

The dissertation consists of 6 chapters -beyond the introduction- and is structured as follows:

Chapter 2 presents a review of the academic literature related to the topics investigated in the
dissertation. Based on the structure of the proposed approach the literature review is classified
into different sections covering the different approaches proposed in the literature. More
specifically, the individual sections cover the following topics: assembly information in product
files, methods for generating assembly information from product files, End of Life products
value estimation, assembly line balancing approaches and assembly instruction interfaces.
Following the above, an overview of the research contribution of this work is presented.

Chapter 3 initially provides an overview of the proposed methods. Each individual application’s
requirements are presented together with the realised software architecture. Following this,
the core method is presented, as well as the accompanying applications regarding the
generation and use of assembly information from product design files. Each section describes
the approaches both from the methodological as well as from the development point of view.
The Assembly Precedence Diagram Generation algorithm (APDG) is described in detail and the
complete dataflow is presented and explained. The following sections of the chapter provide a
detailed overview of the developments regarding the rest of the applications.

In Chapter 4, the implementation of the methods described in the previous section is

presented, highlighting the relevant tools used as well as some development details where it is
necessary.

28
INTRODUCTION

In Chapter 5 the case studies that the developed methods were tested through are presented.
In addition to the data used in each case, a description of the results for each application is
given in order to highlight the applicability of the introduced approaches to real life
manufacturing problems, namely to Assembly Precedence Diagram Generation, estimation of
EoL metrics, Assembly Line Balancing and Simulation as well as the generation of assembly
instructions and communication to shopfloor personnel.

Chapter 6 includes an assessment of the proposed methods based on the respective results.
For each case study’s results some conclusions are drawn, especially regarding the actual use
of the proposed method in industrial context. Finally, some aspects of the applications
developed are pointed out, indicating aspects of improvement in relation to existing practices.
Furthermore, the outlook of the dissertation is presented. Directions for future research
activities are proposed.

29
 State of the Art

2 STATE OF THE ART

2.1 Assembly Information in Product Files

In the lifecycle of contemporary technical products, assembly and disassembly can be

considered as critical processes [3]. Beyond the observable assembly operations during a
product’s life (components assembly, final assembly etc.), disassembly operations are included
in all phases of a product’s lifecycle; initially appearing during the product’s components’
assembly, to repairing defects down to the end of the product’s life, for recycling and reuse
purposes.

Figure 2: Assembly and disassembly operations in the

life cycle of technical products [3].

31
State of the Art

In order to define the assembly / disassembly processes of a product, engineers start by

defining assembly constraints in the Bill of Materials (BOM). The Bill of Materials (BOM) is
considered as key information regarding the assembly of a product; it integrates product data
through the product lifecycle, and it has been used as a hub of product data for product design,
production planning, procurement, maintenance and repair [24]. Contemporary Computer-
Aided Design (CAD) software presents a hierarchical view of a product, where all parts and
components are placed on different levels [25]. This is used during assembly planning in order
to distinguish the pre-assembled components from the components and parts that need to be
assembled in the current design. As an example, in the final assembly of a laptop computer,
the motherboard comes pre-assembled with all of the necessary components integrated and
therefore the whole motherboard assembly is considered as one component to be assembled
[9]. This distinction of components into “pre-assembled” and “requiring assembly” has been
investigated also during product design. The main reason behind this is to warrant the
modularity of the products in all their possible variations, in order to decrease the complexity
of planning their assembly [26].

Beyond the relations between distinct parts, information on most types of fasteners are
available within the CAD tools. As fasteners we consider any part that mechanically joins two
or more components together. Usually fastener libraries are provided as extensions of the CAD
tools, with recent approaches offering web-based libraries that enable easier application for
standardisation purposes [27]. The libraries use standardized sizing and naming conventions
dictated by different standardization organisations [28].

Outside the hierarchical structure of product assemblies and the existence of libraries of
fasteners and other joining mechanisms (geometric assembly constraints, welding seams etc.),
any other information directly related to the planning of assembly has to be generated later on
through additional CAx tools. Nevertheless, CAD tools currently provide sophisticated solutions
for feature definition, including constraints and materials of parts. The materials that can be
defined for unique parts are also provided through appropriate libraries and are stored as
additional data on product files. The constraints’ terms differ between commercial tools but
they can be summarised as follows [29] [30](Figure 3):

32
 State of the Art

 Axes coincidence constraint: Two or more components with cylindrical features (holes
or cylindrical geometries) have axes that coincide.
 Contact constraint: Surfaces from two or more components have contact.
 Offset constraint: Two or more elements from different components have a standard
offset.
 Fix component: A component is fixed and not movable within the assembly.
 Binding constraint: Two or more components are bound together (fixed by reference).

Figure 3: Common constraint definitions in CAD software.

Each defined constraint can be used to determine the assembly relations between the
components [5] [31] [32]. Nevertheless, the extraction of the assembly relations using the
already defined constraints has to be done manually by the engineers that are working on
assembly processes definition. Beyond this, the biggest challenge that has been identified
regarding assembly information is the increase of the efficiency regarding its conservation,
development and adaptation for the different applications throughout the product’s lifecycle
[33].

2.2 Methods for Assembly Graph Generation

Assembly planning includes the determination of a feasible method and layout in order to
assemble a product from its components. A product’s assembly plan affects the efficiency of

33
State of the Art

both the assembly processes and the assembly line. Planning and using efficient and effective
assembly processes can actively contribute to the reduction of the manufacturing cost of a
product [4]. In assembly systems design, assembly representation methods that can support
collaborative design and development are considered as key enablers [9]. Provided that
contemporary systems are designed and developed through the collaboration of globally
distributed manufacturers and suppliers, this is becoming increasingly important.

The assembly planning starts by defining the relations between the parts that comprise an
assembly. Different representation methods have been proposed which allow different views
on the components of a product or a sub-assembly. One of the earlier methods was the Liaison
Graph which provided the groundwork for future approaches [34]. The Liaison Graphs are
based on the depiction of physical mates between the components of an assembly but do not
include any assembly relations. Assembly relations are mainly represented through precedence
and sequence diagrams. The precedence diagrams represent the direct relations between
assembly tasks that correspond to different components. In the same manner AND / OR graphs
were developed in order to model assembly relationships so that valid sequences can be
generated [35].

The representation mostly used in assembly planning is the precedence diagram or precedence
graph, which contains all the valid sequences of an assembly. Even for the precedence
diagrams, different representations have been proposed in the literature that include process
times, levels of assembly etc. [36] [37]. The diagram of a precedence graph, theoretically, offers
all the possible assembly sequences of tasks for assembling a product (Figure 4).

Figure 4: (a) Assembly precedence graph of a simple desk assembly and (b) three possible
sequences that can be extracted from the precedence graph.

34
 State of the Art

The advantage of generating precedence relations rather than generating an assembly

sequence directly is that a precedence diagram contains more than one possible assembly
sequences. This allows the selection of the most appropriate one based on additional assembly
planning criteria which can be applied in the different phases of assembly planning. For
example, during process planning, alternative process times and costs can be calculated
therefore enabling the further evaluation of the different alternative sequences. However, the
generation of precedence relations can be a complicated and time consuming task and since it
is currently based on the individual engineers’ perception of the assembly, the results may
exclude possible relations.

Most available methods for sequence generation assume sequential tasks. Consideration of
assembly hierarchy and parallel assembly allows other none sequential sequence choices which
may lead to simplified sequence generation and new system configurations [9]. In current
industrial practices, after the definition of the assembly tasks and their relations, the tasks are
further detailed and enhanced with additional sub-tasks. A common example concerns the
condition at which a component arrives at the line. A component might be outfitted in
protective materials which require a “pre-task” of removing them [38]. Additionally, the use of
Methods Times Measurements (MTM) or similar means, assist engineers in determining the
individual process times and define the relevant resources to be used (e.g. choosing between
electric or pneumatic screwdrivers). Based on all this information, assembly line balancing is
performed which leads to the final assembly sequence which may include parallel tasks.

A popular view on the generation of assembly relations was based on the following practice;
“assembly by disassembling”. It was introduced in 1990 and is based on the fact that the
knowledge of how a product is disassembled can provide the necessary information to
assemble it [39]. This practice employed the linking features between components to represent
the assembly or disassembly precedence relationships of these components based on the
direction they are examined in. For example, going from the completed assembly of the
product towards its base parts would characterize the disassembly, while the other way around
would depict its assembly.

35
State of the Art

Later on, a model that generated assembly sequences by considering precedence relationships
and different combinations of subassemblies was introduced [40]. The innovation of this
approach consists in the fact that a bigger assembly could be broken down in different sub-
assemblies based on the proposed representation of the product. An approach regarding the
modularization of a product’s components utilizing neural network algorithms and design
structure matrices (DSMs) was also introduced in [41]. However, beyond the definition of a
products’ structure and parts’ clustering, what is also necessary for the generation of assembly
plans is the preliminary generation of precedence relations. All the precedence relations have
to be converted to a graph frame, i.e. precedence diagram, from which assembly plans can be
generated. This method was adopted by some authors in order to generate sets of assembly
sequences [42].

The two stages involved in this method were:

 The generation of the precedence relations between liaisons or logical combinations of

liaisons in a product.
 The verification of the liaison sequence.

This method however, required as input the precedence relations, the proper generation of
which can be considered as a time consuming task.
Regarding the direct assembly sequence generation methods, a number of approaches have
been proposed in the recent years. A fully automated assembly sequence planner has also been
proposed, by directly extracting geometric information from STEP CAD files [43]. In this
approach, the geometric information to determine geometric assembly constraints are
analyzed and interference-free matrices are generated to represent spatial constraints
between components during assembly operations. Another notable approach, was an adaptive
genetic assembly planner that could find an optimal or near-optimal assembly sequence of a
product [44]. However, this approach required the pre-existence of geometric constraints
information between the components of an assembly. Other approaches focused on the
generation of disassembly sequences using software modules to generate disassembly
constraints [45]. In conclusion, all assembly information generation methods from product

36
 State of the Art

design files, can be grouped in three categories -based on the results- as seen in the figure
below:

Figure 5: Methods for information generation based

on their output.

The main motivation for the work related to the generation of precedence diagrams is
summarized in the following points:

 Current approaches for assembly sequence generation assume sequential tasks which
disregard parallel assembly choices or the re-generation of sequences from precedence
diagrams [9] [46].
 Extraction of assembly information as early as the product design phases is currently
complex and time consuming. The task of generating a precedence diagram in the
industry is currently being performed using pen and paper methods and the correctness
of the task sequences based on geometric constraints are initially dependent on the
individual engineer’s perception [47].
 As modern customer trends move towards the mass-personalization paradigm, new
assembly design methods in support of open product architecture and personalization

37
State of the Art

design need to be developed [9]. These should be supported by tools that can extract
production information fast.

2.3 End of Life products value estimation

Manufacturers are responsible for the economic and environmental performance of their
products throughout their life cycle. Life cycle engineering aims to integrate issues and
parameters into product development throughout the life cycle of a product, taking into
account the requirements of long-time usage and recycling [3].

Recycling and reuse of products and their components are still not fully and systematically
taken into consideration by contemporary product engineers and designers [4] [48]. This is
partly due to the fact that the designers do not have standard tools to automatically assess
their designs, since the CAD tools do not typically provide any metrics regarding the recyclability
of the designer’s choices. On the other hand, CAD systems have grown from graphics editors,
with initially few built-in design functionalities, to sophisticated tools, providing information on
the characteristics of a designed product, including simulation of its functions [4]. On the other
side, recyclability indicators focus more on the materials and usually disregard any metrics that
concentrate on the requirements of reusing components, concerning the dismantling or
separation phases [49].

In order to accurately estimate early enough the disassembly cost, detailed representations of
the disassembly process should be available at the design phase of a product. Currently,
assembly/disassembly representations are limited in terms of the comprehensiveness of
process information [9]. This creates the need for establishing additional mechanisms that can
help engineers generate this information manually, thus increasing the overall complexity of an
engineering project [50].

A number of recyclability indicators have been proposed in the literature for assessing
recyclability and reusability of product designs. These indicators may refer to the materials used
in the product or to the processes required for its dismantling and recycling. These approaches

38
 State of the Art

focus, in principle, on providing feedback on the potential value that can be extracted from a
product at the end of its useful life.

The indicators solely concentrating on the materials used, typically provide a percentage of the
material’s mass that can be recovered in relation to the total mass of the product [51]. Since
these indicators only provide the potential value to be extracted from EoL products in terms of
mass, they do not provide any feedback on the actual cost of recycling the product. Also, a few
indicators have been proposed considering as recycling output only the materials that have
value over a certain threshold, i.e. the percentage content with high recyclability index [49].
Similarly, the recyclability rate is one of the metrics, which again uses the mass of a certain
recyclable component multiplied by an index that expresses the recycling rate of components
that have the same material. This indicator has been used together with other factors, in order
to provide alternative dimensions or materials of certain components for the products’
recyclability optimization [52].

Although most indicators may express the potential recycling value, they do not take into
account the economic feasibility of recycling or reusing a product’s components. Another
indicator that expresses the profit to loss margin is the PLMrecycle, which provides the gains
minus the costs of recycling a product. However, the great challenge for this indicator is the
estimation of the values used for the disassembly costs [49].

Summarizing, the main challenges for estimating the value that could be yield from special
handling of EoL products are:

 Lack of indicators that sufficiently provide information on the actual value that could be
extracted from a product by also considering the costs for handling it.
 No information available to product designers regarding disassembly processes.
 Absence of a method for the estimation of disassembly costs.

39
State of the Art

2.4 Assembly Line Balancing and Simulation

One of the main objectives in the process of designing assembly lines is the maximization of
the ratio between throughput and the required costs. This is often deemed as a problem of
considerable industrial importance [36]. One of the most relevant problems that has been
widely researched is the Assembly Line Balancing Problem (ALBP). In the ALBP, an assembly line
consists of stations arranged along a conveyor or transportation system, where the tasks are
consecutively launched down the line and are transferred from one station to the next. In
today’s highly competitive market, it is often necessary that engineers use flexible, user-friendly
decision support systems that may be applied to real-world assembly line design problems [53].
At the same time, the production planning of manufacturing systems, including assembly lines,
has become one of the key areas of production systems’ research. The production plans have
to cope with market, production and logistics uncertainty [54].

Today’s manufacturing organizations usually offer diverse product personalization and

customization options, in order for the customers’ needs to be better fulfilled. This offering of
customization options may often lead to significant competitive advantages [4]. However,
broad product customization makes the production planning process increasingly complex,
since any changes taking place in the demand profile may considerably affect the production
settings and configurations. This is one of the main reasons why the design of an assembly line
is usually carried out in a series of discrete, consecutive phases. The demand profile is often
considered fixed for the subsequent line design phases, such as the one related to the ALBP.
The ALBP, in particular, addresses only a subset of the entire line design problem and deals with
the way that the required processes for the assembly of a production item, can be efficiently
assigned to the various resources (e.g. work centres or stations). In the literature, numerous
methods have been developed and proposed in order to provide solutions to the different
forms of ALBPs. The assembly lines can be classified into three main categories with respect to
the number of product models they produce; simple (SALBP), mixed-model (MMALBP), where
various models of a generic product are produced on the line, in an intermixed situation, and
multi (MuMALBP), where more than one product is produced on the line in batches [55].
According to the adopted classification, different approaches for optimization can be assumed
(Figure 6). Optimum seeking ones have been used for ALBPs that have a small number of

40
 State of the Art

constraints while heuristics have mostly been used for problems that have greater number of
constraints. Moreover, in the past decades additional research has been performed on the
integration of knowledge systems and expert systems that can either provide similar solutions
from past problem identification or guide multi-objective searches based on preferences to
specific criteria [56].

Figure 6: Assembly Line Balancing Problem and solution methods classification.

A simple ALBP is presented in Figure 7. This problem falls into the SALBP category and includes
seven tasks / processes. It is presented in the form of a precedence graph. The numbers inside
the nodes represent each task’s identification number, while the ones outside the nodes
correspond to the processing times for these tasks. The fourth processing task has, for instance,
a 5-time unit. The precedence constraints are represented by arcs, in which case, the second
and third tasks have to be completed before the fourth task begins. One of the ways of
minimizing the cycle time problem of this assembly line, is by grouping the tasks in three (3)
work centres, depicted with dashed boxes, as shown in Figure 7. In this example, the cycle time
of WC1, WC2 and WC3 are 7, 7 and 8 time units respectively, and therefore, the cycle time for
the entire assembly line is 8 time units.

41
State of the Art

3 4
WC1
2 5
4 5 1
1 4 7
2 3
3 6

WC2 WC3

Figure 7: Simple ALBP example solution.

The most studied instance of the ALBP family of problems is the Simple Assembly Line Balancing
(SALBP) as described in [57]. The Simple Assembly Line Balancing is a label on the field of
research with numerous simplifying assumptions in the definition of an ALBP. A recent study
related to demand uncertainty on SALBP proposed an optimization approach [58]. In this
definition of SALBP, the exact quantity of products to be manufactured is unknown. The
objective is that the workload variance over all the stations in the line be minimized. To this
effect, there are two ways of solving the problem: by using a min–max indicator, which
minimizes the maximum workload variance (distributing the workload as evenly as possible)
and by considering the so called α-worst approach. Another way of considering uncertainty in
ALBP is by assuming that the processing times are uncertain and not deterministic. A related
stability study has been performed for the case of the General Assembly Line Balancing Problem
(GALBP) [59]; the GALBP covers a series of extensions of the SALBP by integrating elements met
in industrial practice, such as U-shaped lines, parallel stations or processing alternatives. The
level of demand uncertainty was expressed in [59] via its mean absolute percentage deviation
and the performance of the assembly system was measured by a fill rate. This arrangement
was proposed on the basis of the direct influence of demand uncertainty on the so-called fill
rate for customer demand. The larger the demand variability was, the more sensitive the fill
rate was.

The assembly lines are commonly balanced for the production of mixed products, even slightly
different, whilst the demand is usually not fixed and furthermore, each individual customer
generally desires not only a specific quantity at a certain time, but also with specific
characteristics depending on his / her needs. A change in demand, in the assembly lines that

42
 State of the Art

are set to produce mixed or multi-variant products, in a given sequence, would cause a change
in the actual operation times of the required processes [60]. This would lead to a re-balancing
of the line which would mean production losses for the enterprise, as all resources would have
to comply with new processes within the station.

A Mixed-Model Assembly Line Problem (MMALBP) is regarded as being much closer to the
industrial assembly line design problems, and may consider a series of uncertain factors, with
greater accuracy, as described in [61]. MMALBP may also consider demand variation, since it
allows the production of different models on a common base product in an assembly line.

The consideration of equipment performance and demand uncertainty in the process of

designing assembly lines has been an important issue of many manufacturing firms. However,
there is no simple way of addressing this challenge. In essence, this problem is being
interpreted as the modelling and optimization of an assembly line’s configuration, by
addressing, at the same time, the problem of scheduling the line.

In order to overcome the limitations of line balancing approaches and to obtain more detailed
information on the designed systems, simulation methods are used both prior and after the
performance of assembly line balancing. As an example, process simulation or calculation of
process times with MTM methods is performed prior to the performance of balancing in order
to obtain the necessary process times. After line balancing, discrete event simulation is
performed for the estimation of additional metrics such as production rate, reliability of the
systems etc.

The discrete event simulation is performed manually, i.e. by creating models and testing them
in a Discrete Event Simulation (DES) application. In order to create the model, the relevant
information must be gathered by the different engineers working in the design of the processes
(Figure 8). From all the gathered information the simulation model is built using specific
elements [62]. This process is done manually by engineers and in cases where a different model
has to be tested (e.g. as a result of an updated line balancing) the existing model has to be
reconfigured manually as well. Although, time consuming, in different industrial cases, the

43
State of the Art

combination of line balancing and discrete event simulation can result in the increase of
productivity dramatically [63].

Figure 8: Necessary information that have to be gathered to build a manual assembly line model.

It is evident, that the seamless integration of the discrete tasks of line balancing and simulation
would result in more detailed outcomes concerning line planning in shorter durations. In order
to achieve this however, a larger system, that would incorporate a common information
structure of the designed system as well as methods that would automatically integrate line
balancing results and discrete event simulation input, would be necessary.

2.5 Assembly instruction interfaces

Although in human - based systems (non-automatic, semiautomatic) flexibility is considered

being higher than that of fully automated systems and the systems themselves are considered
more adaptable to changes, there is still potential for the flexibility and response of the
shopfloor operators to be increased [4].

To this end, the use of smartphones or similar devices as support tools for operators was
examined in [64]. In a similar approach, the authors discuss a software system utilizing wireless
information and communication for supporting assembly line operators in performing their

44
 State of the Art

assembly tasks [65]; the proposed system sends information to operators to handheld devices
(iPod Touch) of each operator. Furthermore, the use of tablets has been explored in a case
where the concept of a mobile dashboard for shop floor operators, providing information on
process context, performance, knowledge and communication was presented [19]. In these
approaches relevant data models were also created to support such activities.

At the same time, the use of Augmented Reality (AR) technologies in support of human
operators is becoming increasingly investigated. Through AR, digital information and
knowledge can be reused, while supporting assembly operators [19]. With the new advances
in the AR hardware field, digital content can be visualized in high quality and fidelity, while the
tracking systems utilize the advancements in ubiquitous computing, providing accurate tracking
of objects in a real environment. This could help to significantly reduce the time required for
shopfloor personnel to read instructions which currently is being done through (Figure 9):

 paper-based instructions attached to the body of the assembly that include the parts to
be assembled, and
 monitors (screens) mounted inside the assembly stations that present the instructions
to the operators in a combination of text and other visual information.

Figure 9: Manual assembly instructions using (a, b) monitors and (c) paper.

45
State of the Art

However, there is a great need for integrating AR systems with information structures,
associated to manufacturing systems [20]. In another approach, an iPod Touch device was used
as a gesture-control mechanism that enabled image based manipulation [66]. Although it has
been argued that Interfaces of AR applications should be intuitive for users and easily controlled
with natural human movements, such as gestures and kinesthetic control, this is hardly the case
for industrial applications, where users need to provide more elaborate input, including filling
in alphanumeric fields [67]. Moreover, performing gestures in industrial environments, where
heavy machinery is around should be avoided.

A typical AR system consists of a hardware device that can be either head mounted or
handheld. A typical AR software system has two main functionalities; user tracking and
information augmentation. The hardware system components are usually bundled in one
device, where the display and computational functionalities are supported on the same device.
The most common devices in the shopfloor applications are Head Mounted Displays (HMDs)
[20]. However, wearing HMD devices can cause headaches, dizziness and nausea. This is due to
the graphics lag in the direct view and can quickly leads to “simulator sickness1” [68].

Both AR equipment and handheld devices, such as the iPod Touch or similar devices, have been
explored as communication tools on the shopfloor. However, the following limitations can be
observed in each case.

1. AR goggles limit the input that can be provided by the user since the only techniques
for feedback are vocally or through gestures; although, the AR goggles are very easy to
carry, and in some cases, don’t even interfere with the users’ field of view, they cannot

1
A type of motion sickness that is experienced when users perform simulations in Virtual or Augmented Reality
systems for extended periods of time.

46
 State of the Art

be used as input devices as they require the use of additional tools (such as a
smartphone bundled through Bluetooth technology) [20].
2. Handheld devices, in the form of tablets, are not flexible in terms of transportation and
use; tablets can be difficult to carry, especially in cases that users have to perform
manual tasks. Although they have adequate surfaces for information and allow more
elaborate input, they have to be carried in a case or in one hand, thus limiting the user’s
movements.
3. Handheld devices in the form of “smartphones” can provide limited information due to
their size. They can be quite difficult to read, especially when there is a lot and complex
information to be provided.
4. AR HMD devices can lead to headaches, dizziness and nausea, especially when the
augmented content does not follow the real world context presented to the user (static
or dynamic with lag).
5. Although the simultaneous use of AR goggles and handheld devices has been discussed
in literature, the combination of additional features of Augmented Reality hasn’t been
explored [67]. The following figure summarizes the disadvantages and advantages of
each hardware application:

Figure 10: Device assessment for shopfloor applications.

47
State of the Art

2.6 Research contribution of the thesis

The key research contribution of the dissertation is summarised in the following points:

1. Method for the generation of assembly information from assembly designs

This work presents a method for extracting assembly information from product design data.
The method is based on the concept of assembly tiers for the separation of parts / components
to be assembled into groups, based on their geometric characteristics.

2. Consideration of End of Life handling metrics during design phases

Recycling and reuse of products and their components are still not fully engrossed by product
engineers and designers. This is partly due to the fact that the designers do not have standard
tools to automatically evaluate their designs, since the CAD tools do not typically provide any
metrics to the recyclability of the designer’s choices. A method is presented for the estimation
of recyclability indicators of product designs. It involves the automated CAD-integrated
disassembly precedence diagram generation algorithm and a recyclability assessment module.
The algorithm employs collision detection and CAD feature recognition for the extraction of
disassembly tiers as well as the sequences of specific components. This information is then
processed in addition to the components’ characteristics, including materials, by the second
module, which is capable of providing an initial evaluation about the product’s recyclability.

3. Investigation of an integrated approach to assembly line balancing through the simultaneous

use of discrete event simulation
An internet-based approach is proposed, for the performance of assembly line balancing and
simulation in an integrated manner. The resulting software is capable of generating, simulating
and evaluating a series of alternative assembly line configurations and is implemented as a
cloud service, taking into account diverse market and manufacturing scenarios.

4. Integration of augmented reality tools into final assembly operations

The assembly information generation algorithm is also used for the definition of assembly
steps, carried out by engineers with the use of CAD data and additional input concerning the
product. This information is then delivered to an AR (Augmented Reality) application, which is

48
 State of the Art

responsible for the authoring of visual instructions of assembly tasks. The AR application is
based on an algorithm that creates the virtual instructions, having as basis the assembly
precedence relations of the specific product’s parts. The instructions are stored in the form of
a template.

5. Proposal of a platform architecture for the support of collaborative activities during assembly
systems’ design
A collaborative platform is proposed for the more efficient management of engineering
information within an organisation as well as a group of organisations (extended engineering).
The platform supports different phases from product design down to assembly line balancing
and simulation and is based on a semantic core that expresses information about Products,
Processes and Resources. The platform is internet-based, connecting engineers in the extended
enterprise with shop-floor personnel. The concept is based both on a series of distinct
components and elements comprising the knowledge base, and on a set of web-based services.
The latter are interfaced with a set of Computer Aided (CAx) systems though information
extraction applications and provide decision support capabilities for the design of a production
line.

6. Validation of the above in real industrial case studies with data provided by the European
automotive industry
All proposed methods were demonstrated through industrial case studies formed by the
requirements and data of automotive industrial companies.

49
 Methods Design and Development

3 METHODS DESIGN AND DEVELOPMENT

3.1 Overview of the Proposed Approach

The following sections present the approach that this work elaborates on the aforementioned
problems (section 1.2) and how this approach was realised through a common infrastructure
design.

3.1.1 Modeling of the problem and identifies solutions

The approach for assisting engineers in the process of assembly planning starts by the
development of a method for generating assembly sequence plans. When first proposed,
assembly information generation was a breakthrough that could reduce the time and cost of
the planning and as a result the competitiveness of the manufactured products (see section
2.2). Additionally, it could disengage the procedure from the necessity for an expert assembly
sequence planner. Therefore, the core research has been carried out for generating assembly
sequence plans automatically or semi-automatically. In addition to the generation of assembly
information from product files, their handling and use in different areas is improved through
the development of specific applications targeting different phases of the lifecycle of the
product.

In the following figure the tasks that engineers and shopfloor personnel carry on during specific
phases of the product’s lifecycle can be seen with reference to the different applications
proposed in this thesis.

51
Methods Design and Development

Figure 11: Steps of the product’s life and related developments.

Each of the aforementioned solutions is implemented as one or more software applications.

Since this thesis tries to view the holistic nature of the assembly design and development
processes, the methods proposed here, are designed on the basis of a common infrastructure
that views the engineering processes as both individual and collaborative. The first category
refers to the processes performed by engineers locally, namely those belonging to the same
department or who are in close distance and in almost direct communication with each other.
The second category includes the processes performed by different engineers, belonging to
different departments of the same or different organization. This distinction is made to
separate the applications that refer to the two main categories on the basis of their
characteristics.

The applications fitting in the first category are named engineering assistance applications and
comprise primarily engineering applications that are used by all engineers. These can be
common CAD and CAM tools as well as spreadsheet and text editing applications. Furthermore,
add-on applications are considered. Add-on applications are those working on top of the
aforementioned software, and are used for the generation of information that is relevant to a
certain type of project.

The second category of applications refers to the collaboration assistance applications that
enable engineers to work by sharing information in a common space through the same

52
 Methods Design and Development

“vocabulary”. This is a strong requirement, since in today’s engineering world there is a

plethora of software tools that use different formats of engineering files. As a result, the
engineering companies that collaborate with Original Equipment Manufacturers (OEMs) should
use the same software tools used by OEMs for the efficient exchange of information. This places
a financial restriction to the organisations that participate in the same projects, such as system
integrators or service / part / equipment suppliers (Figure 12).

Figure 12: Software need distribution in a common

project with an OEM.

At the same time, not all results of a certain process have to be or can be shared. As an example,
a Small to Medium Enterprise (SME) that manufactures fixtures for assembly lines, will have to
share the external dimensions during the production design phase, but will not wish to share
the specific design methods that are also stored into the resulting CAD file. The main
requirement of the collaboration assistance applications is the interpretation of data having
resulted from engineering processes into:

1. the minimum amount of information that can be shared

2. a format that is usable and understandable by the different participants in a common
project.

This was partially performed through the use of open standards [69]; however, this was not
adopted by industrial organizations for two reasons. The first reason is that engineering file
formats constantly change, due to upgrades used by the software tools, and therefore the open
standards are constantly subject to changes. Secondly, more complex, engineering files cannot
be converted into a common open format, owing to the difference in information structures
that various software vendors have chosen. In order to support this, the proposed approach

53
Methods Design and Development

for collaborative assistance applications is separated into two types of applications; conversion
or information extraction applications and collaborative web-based applications. The first type
of applications focuses on the extraction of all relevant information from the engineering files
that are supported by different software vendors. The second type of collaborative applications
deals with those using information that is received automatically from different project
participants. As an example, in this thesis, an assembly line balancing and simulation application
is presented. This application makes use of information generated by various engineers that
might belong to different departments of different organizations.

3.1.2 Developed tools infrastructure

In order for the two categories of applications to be effectively integrated, a universal approach
to data management is proposed. The data management is divided on the basis of the storage
requirements of different files and information that are used in engineering projects. Initially,
there is a local storage used only within an organization and comprises local servers or drives
that can store files relevant only to the work performed by a particular person or organization.
Then, an extended database is proposed for the sharing of engineering information and a
repository for storing relevant files that have to be used during a project’s execution.

Figure 13: Overview of the proposed infrastructure.

In the following sections, the core methods and algorithms of the applications that have been
developed are described. As engineering assistance applications, there is a description of two
tools, one for the generation of the product components’ precedence diagram from

54
 Methods Design and Development

engineering files and another for the estimation of EoL metrics and a developed indicator.
Additionally, as collaborative assistance applications, there is an information extraction
application and an assembly balancing and simulation application presented. With reference
to Figure 1, the applications that are developed, including the implementation of the
precedence diagram algorithm, are:

1. Assembly precedence diagram generation application

2. Disassembly precedence diagram generation and estimation of a product’s EoL
indicator
3. Collaboration assistance applications:
a. Information extraction
b. Assembly Line Balancing
c. Discrete Event Simulation
4. 3D instructions generation and visualisation

The following figure shows how each of the above applications fits to the proposed
infrastructure:

Figure 14: Applications with reference to the

proposed infrastructure.

55
Methods Design and Development

3.2 Precedence Diagram Generation Algorithm

The proposed method for generating assembly and disassembly information from CAD files,
introduces the concept of assembly tiers which correspond to different groups of parts or
subassemblies of a product that can be assembled at the same time, based on their geometric
characteristics. In a previous work, the parallel disassembly groups have been described as
assembly levels or layers for different applications (e.g. onion peeling) [70]. Assembly tiers can
be considered as groups of parts that have to be assembled before or after other groups of
parts. Using again the desk example from section 2.2, its assembly precedence diagram with
the relevant tiers defined can be seen below (Figure 15):

Figure 15: Assembly tiers from the desk example.

The proposed approach focuses on generating an assembly precedence diagram using

assembly tiers. The added value of using assembly tiers is that it significantly reduces the
solution space when trying to identify the relations between components in a diagram. If no
assembly tiers are defined prior to the identification of the assembly relations, then each part
is eligible to be followed in sequence by the all of the other components. As an example from
Figure 15, component 1 could be followed by all of the components of the assembly, i.e.
components 2, 3, 4, 5, 6 and 7. When identifying that the next tier has as member components
3 and 4, the algorithm has to only search for relations between component 1 and components
3 and 4. The proposed method includes as a core method the collision or clash identification
between parts or sub-assemblies of the examined assembly. Furthermore, in order to complete
the assembly task definition, the relevant fasteners are associated to the relevant parts or
components. Finally, through an additional collision detection routine, the assembly tiers form
the precedence diagram by identifying liaisons between parts of sequential tiers. In the

56
 Methods Design and Development

generated diagram, each node is represented by a component and the links between these
nodes represent assembly precedence relations.

Beyond some initial requirements that have to be satisfied for the CAD data to be processed,
there are two steps where user input is required which are only necessary at the start of the
algorithm’s data processing and after the first results have been submitted to the user. The
precedence diagram is provided to the user in the form of a spreadsheet containing all
components and their precedence relations as well as their insertion directions to the
assembly. The overview of the method can be seen in the figure below:

User Input – Step 1 User Input – Step 2

Assembly CAD Algorithm specifications Singular extraction directions

Separation of components
Sub- Assembly tiers generation
Fasteners Main components
assemblies
Assembly tiers Associated Insertion
fasteners directions

Remove fasteners

Collision detection – Step 1 Collision detection – Step 2

Dissasembly tiers Extraction directions Precedence diagram

User
UserInput
Inputstep
step Algorithm
Algorithmstep
step

Figure 16: Overview of algorithm’s steps.

The different phases are described in the next sections. The two phases are separated by the
two steps where input from the user is required.

3.2.1 User input and CAD requirements

The first step of the method requires the user to provide the CAD data as well as specific
parameters regarding the algorithms’ application. In the CAD software used for the method’s

57
Methods Design and Development

implementation, each product is represented by a CAD tree containing different levels of parts
or assemblies. The algorithm identifies the components to be assembled as the components
on the first level of the tree. Therefore, each component that will be pre-assembled should be
on the first level and its parts or sub-components on the second level. The algorithm can then
be initiated as a Microsoft Visual Basic application.

After the algorithm has been initiated the user is requested to select the assembly’s base
component(s). These components will be considered being fixed, during the assembly
processes, meaning that this may include additional assemblies representing fixtures.
Furthermore, the user is prompted to provide a step for each component to be moved by,
during the algorithm’s collision tests as well as a value for the collision detection sensitivity. The
collision / interference detection is a standard feature of contemporary CAD software and
usually makes use of spatial partitioning for faster identification of interferences between
different components.

In order for the sequence of the algorithm’s steps to be demonstrated, a car differential (Figure
4), is presented. The assembly contains 16 components on the first level and the base
component selected by the user is the “Case” component.

Figure 17: Exploded view of the differential assembly.

58
 Methods Design and Development

3.2.2 Separation of components

The separation of components into fasteners and assembly components is based on a previous
method that identified the fasteners through their names [71]. The algorithm goes through the
first level of the assembly, identifies the fasteners and deactivates / hides them. This is done to
exclude them from the collision detection results. The fasteners are further separated into two
groups; primary and secondary fasteners. The keywords used for identifying the primary
fasteners in the current implementation are “bolt”, “screw”, “clip”, “wedge” and “pin” while
for the secondary fasteners the keywords are “nut” and “washer”. The name identification is
independent of the case that the letters of the name are provided (upper or lower). Finally, the
rest of the components are separated into parts and sub-assemblies based on the tree
hierarchy.

For the differential, the 16 components of the first level are separated into nine (9) assembly
components, six (6) primary fasteners and one (1) secondary. The primary fasteners are
identified by the word “screw” contained in their names (“Screw1.1”, “Screw1.2”, “Screw1.3”,
“Screw1.4”, “Screw2.1” and “Screw2.2”). The secondary fastener is “Nut.1” due to the keyword
“nut”.

3.2.3 Fasteners removal and first collision detection

After the fasteners have been recognized, and categorized into primary and secondary, they
are removed from the scene (hidden) and their geometries are excluded from collision testing.
When the collision detection begins, the algorithm’s specifications for each part are calculated
and stored. The specifications include the distance that each component has to be moved at
different directions. The distances are defined based on the dimensions of the overall assembly.
From these distances a final position is calculated that the object must reach in a global or local
axis without collision. If this occurs, then the algorithm recognizes that the component can be
removed from the assembly. As proposed in [71], additional directions are examined after a
component has reached a position after which collision occurs. Based on constraints defined
by the user during design which are present in the CAD file, the algorithm prioritizes the
possible directions that a component can be moved to. As an example, if there is an axis
coincidence between two parts, both parts will be firstly checked for removal on this axis. The
following diagram shows the logical processes within the algorithm (Figure 18). The algorithm

59
Methods Design and Development

starts after the user has provided his input and stops when no component has been left in the
scene except the base component(s).

Start

Move xi component, S
distance in dj

Yes
Collision

No

Yes
No xi component reached
First coll. detection
final position PFij

No
Yes

Move xi component,
-S distance in dj
Move component xi
to initial position Pi

Store temporary initial

position Pij
j = j+1

Yes No Yes
j≤ J j > J/2

No Yes No
J=6 J=6
i = i+1
j=1
No Yes

D2 = D ∖ {dj,dj+3} D2 = D ∖ {dj,dj-3}
Yes Global coordinates No
coincide with local

Move xi component to P2i

k=1
J=6 J = 12
Yes
No Move xi component, S
k ≤ J-2 distance in d2k
Yes
i≤I
Yes
No k = k+1 Collision
Set all components
with valid No
disassembly
directions to tier Tl xi component reached No
final position PFijk

l = l +1 Yes

Store dj as a valid
Yes Remove all disassembly
components with direction for xi
I>0
valid disassembly
directions
No

End

Figure 18: Collision detection algorithm.

60
 Methods Design and Development

Where:

I: Total number of components that are in the scene minus the components selected
as base parts. The total number of these components is reduced during the algorithm’s
execution; when a tier has been identified, all components that belong to it are
removed from the scene.

X: Set of all components that are not excluded as fasteners or user defined base parts.
Where xi ∈ X, i = 1, 2, 3….I.

S: The step (distance) that each component is moved to a certain direction when
performing collision detection. The initial step is provided by the user (default value is
set to one millimetre).

l: Number of current assembly tier. l = 1, 2, 3, …

D: Primary set of all axes (local and global) that a component can be moved on.
D = { xgl, ygl, zgl, -xgl, -ygl, -zgl, xloc, yloc, zloc, -xloc, -yloc, -zloc }
Each primary direction is denoted as dj ∈ D, j = 1, 2, 3, ...,12 where d1 = xgl, d2 = ygl etc.
The sequence of the axes can change based on possible constraints defined by the user
(e.g. axes coincidence). As described previously the axes corresponding to constraints
are checked first.

D2: Secondary set of all directions (local and global) that a component can be moved at.
Each direction is denoted as d2k ∈ D2, k = 1, 2, 3, …, 10
Where D2 ⊂ D, since the axis that the component has already been moved at and the
opposite axis are excluded from the secondary set.

The algorithm has the following steps:

1. The algorithm starts by moving a component at the primary directions defined by the
local and global axes dj (starting with xgl) at a certain step (S), and then checks if the

61
Methods Design and Development

component has collided with any of the other components including the base
components. If the local and global axes coincide, only the global axes are used. In the
differential’s case, every component’s local axes system is parallel to the global axes
system.
2. If a collision occurs after the first iteration, the algorithm stores the last position of the
component that no collision occurred (Pij).
3. Then the component is moved towards the secondary directions (d2k) which are the
same as the primary with the identified dj and its opposite omitted.
4. At each component moving step of the algorithm, a final position is defined as explained
earlier, which is different for the primary and secondary directions; PFij and PFijk
respectively.
5. If the object can reach any of the final positions without colliding with another object,
then the primary axis dj is stored as a valid disassembly direction. The same process is
repeated until all primary axes have been checked.
6. Finally, after all components have been checked, the ones that can be disassembled at
this stage (tier) are stored to the current tier (Tl) and removed from the scene.
7. If components remain in the scene (I > 0) the process is repeated until no components
are left, except the base components.

The results of this process are the disassembly tiers as well as the possible disassembly
directions/ axes that a component can be extracted from the assembly. These are provided in
a table containing one column per axis for the directions and one column containing the
respective tier (Figure 19).

Figure 19: Output of the first collision detection step.

62
 Methods Design and Development

The format of the output of the first collision detection step for the differential is shown in
Figure 20. It should be noted that the collision detection routine is not applied to the
component “Casing.1”, since it is the base component and is indicated with a zero tier number.

Figure 20: Output of the first collision detection step for the differential.

3.2.4 User input for singular extraction paths directions

After the possible primary extraction directions have been generated, the user is prompted to
select only one that will be used for generating the precedence diagram. If one or more
extraction directions exist on the global axes then the local axes are dismissed. The user is
prompted for each component of the assembly which in the implementation is highlighted, as
well as the global or local axes he is expected to select one from. After the process has been
completed, each component has only one extraction direction which will later be inversed for
the assembly precedence diagram generation. The only assembly component of the
differential, with more than one extraction direction, is “Cap.1”. As a result, the user is
prompted to select the appropriate extraction direction only for this component. During the
implementation, the selected direction was Xgl.

3.2.5 Assembly tiers and associated fasteners

After the disassembly tiers and disassembly directions have been defined, the algorithm
identifies the fasteners that are linked to each component. At the beginning of this step, all
components are removed from the scene and the base components and fasteners are the only
components left in the scene. Then, the components of each tier are reintroduced, one tier at
a time and moved one step (S) to their disassembly direction. When a fastener collides with a
part, it is introduced into a new row under the moved component and removed from the scene.
The component is then returned to its initial position. The whole process is repeated until all

63
Methods Design and Development

components of all tiers have been checked and therefore every fastener has been associated
to a specific component. The following figure shows the flow of the algorithm.

Start

Introduce all
components of Tl

Move xi component, S
distance in di

Yes Associate fastener f

Collision
to component xi
No
i = i+1
Remove fastener f
from scene
Yes
i≤I

No
Move component xi
Remove all to initial position Pi
components of Tl
from scene

l=l+1
i=1

Yes
l≤L

No

End

Figure 21: Algorithm for fastener association to components.

Where:
L: maximum number of tiers.
di: disassembly direction of component xi.
f: fastener that collided with component xi.

After the algorithm ends, each axis’ value that is one in the table presented in Figure 22 is set
to zero and the opposite axis’ value is set to one. Then the tier values are reversed except the
base tier (T = 0) where the value of each tier is calculated as shown:

64
 Methods Design and Development

T = (L+1) – T (1)

Considering the example of Figure 19 for L = 4, part of the resulting table would be:

Figure 22: Translation of Disassembly tiers table to Assembly tiers.

Considering the example of the differential, the translation of the Disassembly tiers table in
Figure 20, results in the Assembly tiers table depicted in Figure 22.

Figure 23: The translated disassembly tiers table to assembly tiers for the differential.

As seen in Figure 22, the component “Cap.1” has collided with the 6 primary fasteners, showing
that all 6 screws were attached to the cap of the differential. In addition, the component
“Axle_mount.1” has collided with the secondary fastener “Nut.1” and, thus, is associated with

65
Methods Design and Development

it. It should be noted that in this case the “Nut” word was considered as indicating a primary
fastener by the algorithm.

3.2.6 Association of components belonging to sequential tiers

Since each component that belongs to an assembly tier, is associated through a precedence
relation to one or more components belonging to the previous tier, these associations can be
identified through collision detection. In this step, a component of a certain tier is tested for
contact to a component of the previous tier, therefore generating a direct link between these
two components. In the case where there is no direct connection to any component of the
previous tier, a direct relation to all components of the previous tier is assumed. An example of
this are components which are positioned inside a casing; the cap of the casing might have no
collision to any of the enclosed components, however there is a precedence relation to all
components that are inserted last into the casing.

Finally, the corresponding preceding components are added to the column “Preceding
Components”. In Figure 8, the column “Tiers” is replaced with the column “Preceding
components” (Figure 24).

Figure 24: Final precedence diagram table containing associated fasteners and insertion axes.

For the differential, each one of the two drives has as reference the corresponding drive mount,
as seen in the final precedence diagram table (Figure 25).

66
 Methods Design and Development

Figure 25: Final precedence diagram table of the differential.

In Figure 26, it can be seen that the “Nut.1” fastener was associated to both the
“Axle_mount.1” and the “Pinion.1” components (since both are joined by the nut). However,
the first instance was disregarded as appearing in an earlier tier than the first one did. This
means that the “Nut.1” component will only be inserted after the “Axle_mount.1”.

Figure 26: Representation of the precedence diagram

generated for the differential.

3.3 Assessment of Products for End of Life Separation and Material Handling

Considering the existing processes for the handling of products, after the end of their useful
life, there are discrete steps to be carried out, the first of which is dismantling the product and
separating it into reusable and recyclable components. From the perspective of a dismantling
center, the savings from the disposal can be considered unrelated, since a possible disposal

67
Methods Design and Development

would yield no profit. A product’s post-recycled value, when considering standard recycling
costs per mass, is always proportional to the pre-recycled value of the material. Therefore, a
new indicator, called Economic Feasibility for Recycling or Reuse (EFRR) is proposed in this
dissertation. This indicator provides a monetary evaluation that is required both for the
materials and the processes for the manual separation of the EOL products.

 m Vr   Cd
i i i
EFRR  n n
( €/kg ) (2)
m n
i

Where:
n: Total number of components
mi: Mass of component i (kg)
Vri: Value of pre-recycled material of component i (€/kg)
Cdi: Cost of disassembly for component i (€)

3.3.1 Collision detection algorithm and components’ information

In order for the cost of disassembly to be estimated during the design phase, the disassembly
processes have to be defined first. Since this is a time consuming process and would normally
require the direct input of all pertinent information by a process engineer, the algorithm
presented in section 3.2 is partially used in order to identify the necessary processes and their
relations. Beyond the form of the algorithm for the generation of assembly processes, this
variation manages to extract specific information about the product’s components. The input
required by the product engineer or designer is again limited to the selection of a base
component to be regarded as fixed during disassembly. The first step of the algorithm is the
identification of the main components and parts of the assembly. This includes all the sub-
assemblies at each level of the assembly tree. The second step is the extraction of all the
relevant information that will directly or indirectly be used in the following steps. This
information includes:

68
 Methods Design and Development

 Part constraints: The constraints defined during the product’s creation. The main
constraints used are the axes’ coincidence and the parts’ faces contact. All constraints
are stored with reference to the parts they are applied to.
 Part material: The material defined by the designer for each part is stored as its
reference.
 Process information: During design, some pieces of process information, such as the
edges or points to be welded, are available. These are stored together with the relevant
specifications, such as reference parts and the welding seams’ length.

The specific steps of the core algorithm aiming at the definition of the disassembly precedence
diagram are the same as described among sections 3.2.1 - 3.2.6, without the reverse of the
opposite extraction axes’ values and the reverse of the tier values after the extraction of the
disassembly tiers and assembly relations (section 3.2.5). This is done in order to have as a result
the disassembly precedence diagram and not the assembly precedence diagram. In the end,
the information that refers to the product is gathered for each component separately. All the
components consisting of more than one part are checked again by the algorithm based on the
assembly’s tree structure.

Figure 27: Information available at the end of collision detection.

3.3.2 Disassembly cost calculation

In order to quantify the cost related to the disassembly processes, the process times
corresponding to each component are calculated separately. In automotive dismantling
centers, the entire product is initially jigged and dismantled into its main components, i.e.

69
Methods Design and Development

engine, differential, axes, wheels and so forth. After this step, each component is separated
into reusable components or components for recycling. In the current study, the disassembly /
dismantling costs, related to manual labor of the product’s main components, have been
considered. The dismantling process times are classified into three categories; fastener
removal, part extraction and seams removal.

All components have a part extraction process. The time of this process is calculated on the
basis of the relevant assembly times of the part for carrying a distance of three meters [72]. All
the times depend on the component’s weight (Table 1). Regarding the fasteners, the same
approach is followed based on the length of the fastener during disassembly. In the current
implementation, the incorporated fasteners are bolts, screws and nuts and the estimated times
also consider lubrication for fast removal. Finally, the welded parts are considered only when
the weld seams join different materials, namely different types of steel. The time regarding the
seams’ removal depends on their length and type.

Table 1: Components and fasteners extraction times.

Parts extraction times

Weight of part Extraction time
m < 13 Kg 6.5 sec
20 kg > m ≥ 13 kg 20 sec
m ≥ 20 Kg 31 sec
Fasteners extraction times
Length of fastener Extraction time
L < 5 cm 10 sec
10 cm > L ≥ 5 cm 18 sec
L ≥ 10 cm 26 sec

In the end, the cost for the disassembly (Cd) of a component is calculated as a sum of all the
costs considered:

70
 Methods Design and Development

Cd  c(text   t f  Lwtw )€  (3)

Q

Where:
c: Labour cost per time (€/sec)
text: Time for extracting and moving the component (sec)
Q: Total number of fasteners referenced to a component
tf: Time for removing fastener f (sec)
Lw: Length of welding seam w referenced to component (mm)
tw: Reciprocal speed of removing seams of welding w (sec/mm)

The cost of each process is calculated as labour cost. Each process time is multiplied by the
estimated wage cost per time unit and the additional costs are excluded. After all costs have
been calculated and summed up per component, they are added together for the entire
product. For the implementation presented, an average of 28.20 €/hr wage has been
considered for the European Union area EU-18, i.e. 0.0078 €/sec [73].

3.4 Collaboration Software

In order for all engineers involved in a project to be able to make use of the relevant information
regarding assembly planning, a shared database is proposed that will store all relevant
information regarding the system and its processes. As described earlier this is common
practice in extended engineering environments. However, in addition to the storage of
engineering files, software agents are proposed for the execution of collaboration related tasks,
which will be able to provide, as well as to collect, relevant information from the database and
execute specific engineering tasks. The overall view of the applications can be seen below.

71
Methods Design and Development

Figure 28: Applications regarding collaboration.

The applications described are:

 Information Extraction application (Inf. Extr. app): This application is responsible for
extracting the relevant information that engineers have stored in engineering files.
 Assembly Line Balancing application (ALB app): This application is responsible for
performing assembly line balancing using the available information that reside in the
repository.
 Discrete Event Simulation application (DES app): This application performs discrete
event simulation on a selected number of alternative lines generated by the ALB app.

3.4.1 Extended database

The structure of the information, as regards to production design and deployment, is based on
two main requirements:

 The model that will be used should have a definite set of concepts including the
relations between the concepts.
 The concepts should be as generic as possible in order to satisfy expected differences
within the organisations in extended engineering (i.e. collaborative engineering
between different enterprises).

These requirements are somewhat contradictory, since the greater the depiction of concepts,
the lower the generic character of the model. Therefore, the modelling was performed on the

72
 Methods Design and Development

basis of cases created by industry partners that served both as Original Equipment
Manufacturers (OEM) and production system Integrators and parts/ equipment suppliers.

The Product, Process and Resource views are common terms in PLM systems and represent
the different views that engineers have on CAx files during the production design execution. In
the proposed data model, the first sets of concepts of the Product-Process-Resource (PPR)
include the modelling of the actual digital, production planning related data. The main concepts
here are Product, Process and Resource. The modelling of these concepts derives from the
higher level structures of relevant files, used by commercial CAx products, at the extent of their
being considered consistent with each other, i.e. use of the same terminology. The selection of
the CAx tools was based on questionnaires, delivered both to OEMs and SMEs, as well as a
production system integrator. The lower levels of the structures were not used since there were
inconsistencies between the different CAx products’ file structures. In cases where information
from the lower levels of a file’s structure needs to be extracted, e.g. for the identification of
process times from a CAM file, this can be done through the parsing of files or the direct input
by the user of the relevant attribute values through a Graphical User Interface.

The data model proposed, acts as the core of information used or produced by the applications
described at the beginning of this section. Through the combination of the database and the
applications, the engineers involved in the collaborative design of the assembly line can
exchange information with each other in an extended engineering environment. The model
serves as a communication layer between the collaborating organizations. The database, is also
linked to an extended engineering repository which acts as an individual component that aims
at managing and storing files exchanged among Original Equipment Manufacturers (OEM),
production system integrators and equipment suppliers.

The core concept of the proposed data model is the manufacturing system class that consists
of one or more assembly production lines. The manufacturing system is capable of representing
vital information such as the production capacity and availability while its association with the
demand class, provides the demand profiles for the different product types, manufactured in
the lines included in the manufacturing system. The line class is the core of the model, as it is
linked to key production concepts, which are products, components, processes and resources.

73
Methods Design and Development

The line includes key parameters for all assembly planning tasks such as the product cycle time,
number of stations, number of operators per station, etc. The basic concepts of the data model
are briefly described below:

 Product: The Product class can hold all necessary information about a product produced
by the line developed. The key characteristic of this class is its connection with the
Component class; a product is composed of several sub-assemblies or parts
(components) to form a final product.

 Component: The Component class may contain all information related to the individual
sub-assemblies or parts that are handled by processes and are connected in a sequence
to form a complete product. A component may also consist of sub-components, which
are shown in the model through a relevant relation.

 Process: This class can include information related to processes required for the
execution of the entire production. Beyond their association with other classes, the
processes are connected to each other with relations concerning precedence, linkage
and incompatibility in order for the process sequence plan of a production line to be
defined. The sequence plan is later used as the basis for the line balancing procedure
and other planning tasks.

 Resource: In this class, all attributes regarding the production resources upon which the
production process is carried out, may reside. The resource class comprises three
subclasses; namely, Equipment, Human and Conveyor System. The breaking down of
the resources into these concepts was made in order to support the identification of
specific parameters, related to the resources used in various processes. The Equipment
class used for the processes is further analysed into specific equipment such as robots,
handheld tools or milling machines etc. This resource sub-class is essential for the
design of an assembly line, since knowledge identification for the selection of suitable
equipment in a process, based on older similar cases is currently implemented to a great
extent by line designers.

74
 Methods Design and Development

 Configuration: The Configuration class may include all possible combinations of

processes and resources for the generation of alternative process parameters. For
instance, a certain process may be performed with two different tools, where the one
requires less time but the other has smaller running costs. Both process configurations
can be stored into the class, including the relevant cost and time parameters, and will
be used as criteria for the generation of alternative process sequences, prior to the
balancing of the assembly line. The assignment of process configurations to stations,
being necessary for the assembly line balancing task, is achieved through the relevant
association between the two classes.

 Engineering Data Object: This is the most important class in terms of connecting
production information and parameters to engineering files. It may include information
about the type of a file and the model’s concepts they are linked to, such as CAD files
linked to a sub-assembly or text instruction files linked to specific equipment, etc.

 Demand: The demand class refers to the distribution type used in the demand profile
of a certain product. The different kinds of demand distributions such as Volatile,
Uniform and Seasonal are expressed in the data model, through the relevant sub-class
hierarchy.

The following figure shows the basic classes of the proposed data model with the help of a UML
diagram. The diagram includes basic attributes of every class and their relations with other
classes. The structure of the data model and the classes’ attributes were selected by
considering a series of criteria in support of engineers in tasks, related to the design and
reconfiguration of a line. However, a lot of attributes were not included in the figures, due to
lack of space.

75
Methods Design and Development

Figure 29: Basic classes of the proposed data model and their interrelationships.

The type and subtypes attributes correspond to the different types of resources, processes,
systems etc. and the subtypes of some concepts that the data model is capable of representing.
For the concepts that these attributes were assigned to a taxonomy was implemented based
on current industrial practices. For instance, all process types should be defined for the data
model in order for any possible manufacturing process, stored into the repository to be
referenced to a certain type of process. For this purpose, an integer was assigned to every
process type, as it can be seen in the figure below. The subtype taxonomy was also similarly
implemented for the model’s concepts with a higher level of detail. For example, a certain type
of a manufacturing process (e.g. Welding: 2) has given the respective numbers for its subtypes
(e.g. Laser beam welding: 205).

76
 Methods Design and Development

Figure 30: Class type taxonomy.

The concept of Engineering Data Object actually represents several types of documents,
spreadsheets and files that are relevant to the main components of a production line. These
types are expressed through the subclasses, depicted in the following figure and assist
engineers in storing, handling and sharing any necessary file that contains information about
process sequences, resources, etc.

Figure 31: Engineering Data Objects’ structure.

3.4.2 Information extraction application

The Information extraction application is realised through a user interface and a software
agent. The software agent is initiated when a process simulation file is provided by the user
through the user interface. The agent then opens the file via the relevant software’s API and
extracts the pertinent information. Most engineering files related to 3D digital simulation have
a certain topology, comprising three concepts; process, product and resource (Figure 6).

77
Methods Design and Development

Figure 32: Process simulation file topology.

Each process step description in the file includes its simulated duration as well as the
precedence relations to the rest of the steps. Furthermore, the resources linked to each step
are provided with a reference to the specific step. These types of files are constantly updated
throughout a project’s duration, at specific milestones. Based on the certain project’s phase,
this file includes more or the same information, in greater detail. As an example, at the end of
the “Product analysis” phase, this file will include the process steps, their precedence, as well
as the initial estimations of their durations. After the “Physical layout and simulation” phase,
this file will include the specific resources and the process steps’ durations will be based on
detailed tests / simulations.

Through the proposed approach, the information on these files can be used progressively in
the sequence of the phases by providing the relevant input to other engineering tasks, such as
line balancing. Whenever a file is being uploaded through the aforementioned service, the
agent looks for all this information, stores it temporarily and then uploads the relevant part of
it on the extended database and the files on the local storage.

As mentioned, the database comprises different classes that describe the production system’s
elements’ parameters that are relevant to the design of the line. The main concepts used for
uploading information are the ones expressing the Product-Process-Resource (PPR) concepts.
The proposed classes are in line with the product-process-resource paradigm used by
engineering software and can be easily used for storing and updating all information related to
the creation of the line.

78
 Methods Design and Development

3.4.3 Assembly line balancing and simulation

The applications regarding assembly line balancing and simulation are designed as one
application that can create and evaluate assembly line models in two levels:

 On the first level alternatives are generated using an assembly line balancing method.
 On the second level, a number of the alternatives generated previously (evaluated
through user defined criteria) are simulated.

The core of the applications are two agents, the Assembly Line Balancing Agent (ALBA) and the
Discrete Event Simulation Agent (DESA).

3.4.3.1 Assembly Line Balancing Agent (ALBA)

The ALBA considers the following information residing in the database:

 The assembly processes including their precedence relations.

 Different process configurations with diverse process parameters and resources.

The processes taken into account are only at the highest possible level of description, such as
“Tighten screws on part A”. Based on the precedence relations of the processes, the ALBA
generates a series of alternative process sequences that differ in their sequence, as well as in
the process configurations used each time for different processes. For each product, the agent
calculates the maximum number of possible alternative sequences, calculated as follows:

k k

TNA= ∏ ni ∏ pi ! (4)
1 1

Where:
k: Total number of processes for a product.
𝑛𝑖: Number of possible configurations for process i.
𝑝i: Number of outgoing relations to other processes (for process i).

79
Methods Design and Development

As an example, in the following figure the process precedence diagram has 8 possible
alternative sequences since process 1 has 2 outgoing relations and processes 5 and 6 have 2
configurations each. Therefore TNA = (2*1*1*1*1*1) *(1! *1! *1! *1! *2! *2!) = 8.

Figure 33: Example of a precedence diagram containing

alternative configurations.

For ALB an existing method found in the literature was used to generate alternative process
sequences and group them into stations, on the basis of a maximum cycle time and minimum
number of stations [74]. However, the selection of this method was done based on criteria of
flexibility in order to accommodate a larger selection of problem definitions and should not be
considered as a part of this study. Any solution method covering the relevant criteria could be
used. The cycle time provides the maximum threshold for the station times, i.e. a group of
sequential processes. The algorithm generates alternatives by changing the sequences of the
processes or their configurations.

The precedence and compatibility relations of the processes derive from the processes that
reside in the database and are linked to the examined products or variants. Each process
configuration and the resource(s) it is linked to have been previously entered by engineers
working on equipment specification. Each resource provides specific investment costs as well
as running costs to a process configuration. For example, if a screwing process is performed
with a pneumatic tool, it will take less time but will have higher running and investment costs,
in comparison with the same process performed with a conventional tool. The number of

80
 Methods Design and Development

alternative sequences to be included for the discrete event simulation of each product is
calculated as follows;

𝑄 𝐴𝐿𝐷𝐸𝑆𝑡𝑜𝑡𝑎𝑙
𝐴𝐿𝐷𝐸𝑆𝑝 = 𝐴𝐿𝐴𝐿𝐵𝑝 | √ 𝑄 | (5)
∏𝑞=1 𝐴𝐿𝐴𝐿𝐵𝑞

Where:
𝐴𝐿𝐷𝐸𝑆𝑝 : Alternatives to be considered for the discrete event simulation of product p.
𝐴𝐿𝐴𝐿𝐵𝑝 : The maximum number of alternatives for all precedence relations and process
alternatives for the examined product p.
Q: Number of examined products or variants.
𝐴𝐿𝐴𝐿𝐵𝑞 : The maximum number of alternatives for all precedence relations and process
alternatives for the examined product q.
𝐴𝐿𝐷𝐸𝑆𝑡𝑜𝑡𝑎𝑙 : Total number of combined alternatives to be simulated.

The alternatives to be considered for each product result in a number of combinations around
the input number of 𝐴𝐿𝐷𝐸𝑆𝑡𝑜𝑡𝑎𝑙 . The goal is to achieve a number of combinations close to the
𝐴𝐿𝐷𝐸𝑆𝑡𝑜𝑡𝑎𝑙 . This restricts the DESA from creating more combinations and therefore, limits the
time required for the simulation. The rounded up number of the best alternatives for each
product (𝐴𝐿𝐷𝐸𝑆 ) is used for the generation of the combined alternatives. As a final step, the
ALBA uses the alternatives selected for each product or variant and combines them to create
new lines. The combined lines will have the maximum number of alternatives as stations and
as cycle time, the maximum cycle time of their combination.

81
Methods Design and Development

Figure 34: Generation of line alternatives by the ALB agent.

3.4.3.2 Discrete Event Simulation Agent (DESA)

The DESA, is built via the DesmoJ library [75] and takes as input the alternatives generated and
stored in the data repository by the ALBA (Figure 34). The overview of the implementation is
given in section 4.2.3. As described earlier, an alternative defines the process sequence, the
stations where each process is dispatched to and the process configuration of each one, linked
to the necessary resources.

At the same time, the way the processes are distributed over the stations, for each alternative,
dictates the way that a corresponding simulation model is built by the DESA. Based on the
resources used by the process configurations selected, the DESA considers reliability and
maintenance parameters for all stations, specifically, the Mean Time Between Failures (MTBF)
and the Mean Time To Repair (MTTR). These parameters may be obtained by Manufacturing
Execution Systems (MES) and may be edited by engineers, who deal with the equipment
planning, predictive and unscheduled maintenance and process design; both MTTR and MTBF
can be updated on the basis of measurements from the existing production systems that use

82
 Methods Design and Development

the relevant equipment. After the simulation model has been generated, the DESA executes
the corresponding simulation experiment.

The user defines the simulation time and the products’ demand profiles. The simulation time
is the “virtual” time that the line will be running on, e.g. a year. The demand profile determines
the quantity of products per type required over time, e.g. per day or month.

Figure 35: Generation of a simulation model by the DESA.

Once all assembly line alternatives have been simulated, the DESA returns the best alternative
on the basis of a series of criteria with user defined weights, regarding Cycle Time, Number of
stations, Running costs and Production Rate; if the user chooses to use the ALB agent, only the
first two are taken into account. In any other case, all criteria along with their weights are
considered.

83
Methods Design and Development

3.5 Design of the assembly instructions interface

A manual assembly operation, such as the assembly of an engine, is divided into tasks and then
into specific steps. As an example, an assembly task might be the assembly of a part on to the
main assembly. This task might be require specific steps such as placing the part on the
assembly, placing the fasteners and then tightening them.

Figure 36: Assembly Operations breakdown.

This information is currently provided to assembly personnel through papers or monitors as

explained in section 2.5. The authoring of the instructions is done manually by production
engineers which is a time consuming process, considering the many variants of different
products. However, based on the availability of the assembly processes and namely the

 parts they refer to,

 their assembly relations (precedence) as well as the
 fasteners that are referenced to each part,

an automatic authoring mechanism is proposed, that can automatically generate instructions

for shopfloor personnel. In addition to the information provided, the software mechanism can
also use the assembly directions for creating animations in order to animate objects. The
method is realised in the following steps:

84
 Methods Design and Development

 Determination of assembly tiers; all parts are grouped in different tiers starting from
“1”, which excludes the base part.
 For each part, a path is created that can be used to create an animation of the part
being brought to its final position.
 The fasteners of each part are also treated in the same way, generating a similar path
for them.

All the information for the steps above are collected through the output of the APDG algorithm
as shown below:

Figure 37: Information used for generation of assembly

instructions from APDG output.

The grouping of the parts based on their assembly tiers, is done in order to determine the
sequence that the instructions will be presented to the user, i.e. tier 1 parts will be presented
first, then tier 2 and so on. The paths generated will be used for the generation of animations
that will show the parts moving, while being placed, using Augmented Reality technologies. For
each part two positions are defined, a starting position and the end position. The end position
is given based on the CAD data (i.e. relative position of parts in assembly). The start position is
calculated based on the end position, the insertion axis defined by the APDG and the
dimensions of the bounding box of the part.

85
Methods Design and Development

Figure 38: Definition of start position for path creation.

After all the definition of the paths for each part and fastener the instructions are created, per
tier. This means that the user will be able to see the instructions for each tier sequentially. The
details on the implementation of the system are given in the implementation section.

86
 Implementation

4 IMPLEMENTATION

4.1 Applications and tools overview

As described in section 3.1, all the tools that the proposed methods are realised through were
developed following an approach of a structured platform. More specifically, with reference to
Figure 14, the developed applications were developed either as engineering assistance
applications or as collaborative assistance applications (as part of a web-based platform).
Concretely, the two categories include the following:

Engineering assistance applications:

1. Assembly / Disassembly Precedence Diagram Generation add-on
2. EFRR calculation application (EoL value estimation)
3. Assembly instructions generation and visualisation application

Collaboration assistance applications

4. Information extraction application
5. ALBA and DESA

Based on this grouping, the following sections describe the technical implementation of each
application.

4.2 Engineering assistance applications

4.2.1 Assembly / Disassembly Precedence Diagram Generation add-on

The algorithms for the APDG and DPDG was developed as a macro application using the VBA
embedded editor of CATIA V5-6R2013 (Figure 39). A macro is a series of functions written in a

87
Implementation

scripting language that is grouped in a single command to perform the requested task
automatically [76]. The application can be loaded on top of CATIA as a VBA library. The user
input at any phase of the algorithm is requested using input and message boxes (Figure 40). In
case the user inserts a zero value in the step or sensitivity definition boxes, the algorithm
automatically uses the default values.

Figure 39: VBScript editor in CATIA.

Figure 40: Input and message boxes in CATIA.

The collision detection was checked through oClash.Compute. At each time two groups of
components are checked, the one group contains the component that is being moved and the
other one the components that are in the scene or belong to the previous tier depending on

88
 Implementation

the step of the algorithm where collision detection is performed (see Section 3.2). The
sensitivity of the collision has a default value of 0.1 millimetres (Abs(oConflict.Value) < 0.1) in
order to ignore contacts, however, it can be further configured by the user. This default value
is modified during the execution of every collision detection based on a pertinent sensitivity
factor. This factor is dependent on the material of each part and has empirical values which are
stored in the script. The materials considered are those found in the default material catalogue
of the CAD SW used, and apply to components of industrial or mechanical products. The reason
behind using a collision detection sensitivity factor is to compensate for elastic deformations
during assembly or disassembly. The values used, in relation to the elastic modulus of each
material can be seen in Table 2.

Table 2: Sensitivity factor for different materials.

Material Elastic Modulus, E (GPa) Suggested

factor
Metal
Lead 16 1.5
Magnesium 44.81 1.2
Aluminium 70 1.2
Chroma 70 1.2
Silver 76 1.2
Gold 78 1.2
Zinc 97 1.2
Yellow Brass 105 1.2
Bronze 110 1.2
Copper 110 1.2
Titanium 114 1.2
Brass 131 1.2
Steel 200 1
Iron 211 1
Nickel 214 1
Tungsten 400 1
Other
Rubber 2·10-3 10
Plastic 2.2 5
Plexiglass 2.9 5
Epoxy 3 5
Raw polystyrene 3 5
Construction
PVC 3 5

89
Implementation

It should be noted that the values of the factor are subject to change and can be modified based
on additional experiments with the APDG or DPDG algorithms. The resulting arrays that are
produced from the algorithm are exported as spreadsheets.

4.2.2 EFRR calculation application

When the DPDG produces the disassembly information together with additional information
for the disassembled parts, these are all stored in a spreadsheet. This way, any information that
need to be altered can be right on the spreadsheet by the end user through the use of the
relevant spreadsheet application (e.g. Microsoft excel). In order to avoid the creation and usage
of an additional / external application, the relevant calculation mechanisms for the metrics and
indicator proposed in section 3.3 were implemented as a macro code in the spreadsheet
application. The user has only to enable the macro once the DPDG results have been generated
and press a button that appears, which run a script for the calculation of EFRR. Beyond the
calculation of the EFRR, the spreadsheet also presents the metrics for each part (materials
value, disassembly costs etc.) in the relevant part’s row.

Figure 41: EFRR calculator.

90
 Implementation

4.2.3 Assembly Instructions Generation and Visualization application

The assembly instructions generation and visualisation application has two components, the
assembly instructions generation using the results of the APDG and the visualisation
component. Both components are built in Unity2, using C#, on a project that has access to the
CAD geometries of the processed product and the output of the APDG. The visualisation
application requires input from an augmented reality device, in order to recreate the input with
augmented information (3D geometries) and provide it back. AR devices include tables and AR
goggles. However, due to the nature of the application (manual assembly), the application is
intended for AR goggles as described in section 2.5 (see Figure 10).

The first component performs the following steps:

1. As a first step, Unity has to scan the spreadsheet and translate it to a new array. The
new array includes the names of the parts, their assembly axes, as well as the assembly
tiers.
2. Using the stored information, the application creates a script for each part using the
assembly axis and its position. This script describes the animation of each part, with
reference to a central coordinate system.

The visualisation component uses a standard image (marker) in order to have a reference in
the environment within which the application is used. This image is named Image Target and is
used by Unity’s libraries in order to identify a common reference in the pertinent environment.
Through the Image Target the application can identify the origin of the global coordinate
system in that environment. The visualisation can therefore render the 3D geometries of the
provided parts onto the environment. The visualisation component, takes two types of input
from the user:

2
https://unity3d.com/

91
Implementation

 Images through the VR goggles and

 feedback from a handheld device.

Figure 42: Identification of coordinate system through Image Target.

In the current implementation both components have been developed as a desktop application
and are run through a computer to which the AR goggles are connected (Figure 43). This was
necessary in order to provide the necessary input in a high framerate for faster processing of
images.

Figure 43: Assembly Instruction application structure.

The rendering of the parts happens in batches based on the tier number. Initially, the
application only renders the base part(s) and the user can move forwards or backwards, using

92
 Implementation

a handheld input device. The application also renders the referenced fasteners. In addition, in
case the fasteners use their standard names, the application can also render the relevant tools
for specific fasteners (e.g. bolts). The following figure shows the superimposed cap of a
differential together with the accompanied fasteners and tools.

Figure 44: Example of a part superimposed on a base part.

93
Implementation

4.3 Collaboration assistance applications

As described previously there are three main collaborative components proposed, namely:

 Information extraction application

 Assembly Line Balancing application
 Discrete Event Simulation application

as well as an extended database. All the components were considered as components of a

common platform. The extended database was implemented using available commercial tools
through the class structure presented in section 3.4.1. Since the database follows the Product-
Process-Resource (PPR) paradigm, it will be referenced as PPR database in this section. In
addition to the agents described above, a Manufacturing Information agent was developed in
order to manage the data sharing between the different agents. In the current implementation,
two main modules are used; a client library providing access to data (deployed in the same
server as the ALBA and DESA) and the database, where all the data are stored. In addition, the
term Data Transfer Object (DTO) represents each PPR object implemented within the platform.
The implementation of the Data access involves remote operations in the data repository
through the Web Service of the database used. This makes the deployment of the technology
components irrelevant to the actual physical place, where the data repository is deployed (i.e.
it can be used remotely by people working for one organisation).

Figure 45: Manufacturing process information agent and

PPR database deployment.

94
 Implementation

The three main agents (Inf. Extraction, ALBA and DESA), have been implemented as a part of
the same platform as a cloud-based system, integrating information exchange as well as
assembly line balancing and discrete event simulation. This system can be accessed and used
via a standard web browser. The system’s architecture is centralised MAS, having as a core the
data repository and the Manufacturing process information agent. This is accomplished
through a three-layer architecture (Figure 46). The relevant data required by the agents are
automatically retrieved from the PPR database.

All software agents are implemented with the use of the Java programming language and the
JADE agent framework [77], which provides the means for a formal collaboration among
independent components of software. The ALB agent consists of three main components:
 Task Sequencer: Generates alternative sequences according to their precedence, linked
and incompatibility relations.
 Station Former: Forms stations based on the task sequence provided and the relevant
cycle time.
 Multi-criteria Evaluator: Evaluates the alternatives with reference to selected criteria
and user –provided weights.

The DESA components utilise the DesmoJ library [75], and include the following:

 Product Generator: Generates orders, based on the demand distribution provided for
each product type.
 Part Processing: Handles the lifecycle of a part or subassembly and dispatches it to the
next station according to its process sequence.
 Resource Process: Handles the lifecycle of the resources by starting and stopping them.
 System Model: Handles the system’s lifecycle and holds all of the above components.

95
Implementation

Figure 46: Three-layer implementation of the agents.

The graphical user interfaces of the resulting prototypes, which were also designed using
standard Java libraries, is presented below.

Figure 47: GUI of the ALBA and DESA agents.

In Figure 47, the GUI of the ALBA and DESA is presented. Both agents are controlled using the
same GUI (screen) through elements of it. More specifically:

96
 Implementation

 Figure 47.a: Input elements for selection of products to be assembled as well as input
of assembly constraints (No. of stations, Cycle time, Invesment Cost and the definition
of the labour cost in €/h).
 Figure 47.b: Parameters definition for multi-criteria assessment of the produced
alternatives, in the first phase by the ALBA and then by the DESA.
 Figure 47.c: Configuration of the simulation parameters as to its virtual duration and
product demand.
 Figure 47.d: Best alternative produced in the form of a Nummi wall. The Nummi wall
was introduced by New United Motor Manufacturing, Inc. and shows the processes
stacked on different columns that represent the stations. The vertical axis is the
duration (i.e. station times) in seconds.

97
 Case studies and Results

5 CASE STUDIES AND RESULTS

In order to evaluate the Precedence Diagram Generation algorithms (APDG and DPDG) as well
as the accompanying applications, the work developed here was ultimately tested using data
from or related to the automotive industry. For each application, the results were measured
in different ways, i.e. using different metrics referring to time or cost.

5.1 Generation of precedence diagrams for different product design files

In order to obtain a better assessment on the APDG algorithm’s efficiency, eight product design
files with diverse characteristics were used (Figure 48).

Figure 48: Product design files used in experiments.

99
Case studies and Results

The main characteristics that were taken into account for each product design are the total
number of parts placed on all of the assembly tree levels, the number of components (distinct
parts or sub-assemblies of parts) on the first level, the number of fasteners, the total volume
of the assembly as well as its dimensions on the X, Y, Z axes. Finally, the total geometric
elements of the assembly were recorded prior to the actual experiments in order to obtain a
metric on the capacity required for performing collision checks (Table 3).

Table 3: Characteristic of product design files.

Total Comp. Total

Volume X dim. Y dim. Z dim.
No. Design number of on first Fasten. Geom.
(m3) (mm) (mm) (mm)
parts level Elem.
1 Gear box 19 19 0 5.27 * 10-4 122 217 92 2417
Wankel
2 17 17 0 5.64 * 10-4 69 63 82 2398
engine
Mechanical
3 15 15 0 2.67 * 10-4 101 450 224 836
jack
4 Disc Clutch 64 31 33 0.011 328 328 750 2847
5 Stirling engine 41 17 20 4.76 * 10-4 102 247 157 1272
6 Radial Engine 115 27 76 0.044 1160 1136 1056 10164
7 Differential 19 12 7 0.005 400 230 354 5233
8 Rear Axle 322 13 38 0.059 600 2059 696 6138

The system used for the implementation of the precedence diagram generation case studies
had the following specifications: Windows 7 Professional SP1 64-bit operating system, Intel(R)
i7-2600 CPU 3.4GHz, 8.00 GB installed memory (RAM) and NVIDIA GeForce GTS 450 1GB GPU.
The results regarding time until the assembly tiers generation can be seen in Table 4. The
recording of the time started when the first component was moved, after the user selected the
base component(s).

The experiments were performed using different steps during collision detection phases, which
covered the minimum width of the examined components. Some design files required no user
input for the determination of extraction directions, since there was only one possible for each
component, and as a result the algorithm moved directly to the precedence determination and
fasteners’ association. Some components required additional input such as the differential

100
 Case studies and Results

which required input regarding the direction of the “Cap” component which had five possible
disassembly directions (Figure 49). The time used by the user to select the directions was not
included in the final measurements.

Figure 49: Disassembly directions for “Cap” component

in differential assembly.

The results for all product design files can be seen below:

Table 4: Step used and results for each product design.

Step used Number of

No. Design Duration
(mm) tiers
1 Gear box 2 5 1’21’’
2 Wankel engine 2 5 1’29’’
3 Mechanical jack 5 3 0’53’’
4 Disc Clutch 5 8 5’05’’
5 Radial Engine 10 6 6’04’’
6 Steam Engine 5 8 20’10’’
7 Differential 10 4 1’09’’
8 Rear Axle 10 3 7’56’’

Although a direct link between one characteristic to the duration of the algorithm cannot be
obtained, it can be nonetheless seen that the characteristics that have a bigger impact on the

101
Case studies and Results

algorithm’s performance, are the components on the first level as well as the geometric
elements. On the other side, it can be observed that for most design files, the algorithm,
required less than two minutes. In the case of the steam engine (66 components on the first
level, 95 fasteners and 7798 geometric elements) required around 20 minutes for the algorithm
to determine the precedence relations and fasteners association. It should be noted that in the
current literature there is no similar method providing the generation of components’
precedence diagrams and fasteners’ associations directly from CAD files. Therefore, a direct
benchmarking of the algorithm’s performance is not feasible. However, it can be pointed out
that the algorithm satisfies contemporary industrial needs since the actual time for the manual
conception of the relevant information exceeds that of the algorithm.

5.2 End of life value estimation and comparison between different product

designs

The end of life value estimation application uses the DPDG algorithm as well as another
application for the calculation of the EFRR. Two rear axle CAD files were used for testing the
application, following a scenario of total separation, i.e. full dismantling of different
components for reuse, as it is performed in dismantling centres. The parts reused are generally
gears or bearings, which reside typically at the last disassembly tiers of an assembly. The
purpose of calculating the EFRR in this case is to evaluate the design in the scenario where none
of the parts would be reused and all parts would be recycled (Figure 50).

Figure 50: End of Life handling scenario.

102
 Case studies and Results

The two files used for the testing of the algorithm, contained the components in a tree
structure (multi-level), while the fasteners were placed at the same assembly level of the
components they secured. The two assemblies have the various sub-assemblies at different
levels of the assembly tree. Namely, the new axle design had one sub-assembly for the cover,
which contained the sub-assemblies of the pinion gear and the differential gears. The old axle
design is composed of two sub-assemblies, one for the cover, containing the pinion gear sub-
assembly and that of the differential gears sub-assembly. For the better recyclability evaluation
of the product, cast iron and steel scrap materials were considered separately with different
pre-recycled material prices [78].

Figure 51: Old (a) and new (b) rear axle assemblies’ characteristics.

The EFRR results were provided in spreadsheets together with all the other information
generated. An excerpt representation belonging to a generated precedence diagram of the old
axle CAD can be seen in Figure 52, whilst the EFRR results for both products can be seen in
Table 5.

Table 5: EFRR results of the algorithm.

Assembly Materials’ value (€) Disassembly costs (€) EFRR (€/kg)

Old Rear Axle 162.00 15.71 0.30

New Rear Axle 137.45 13.82 0.39

103
Case studies and Results

Figure 52: Excerpt from precedence diagram of old axle CAD.

The increased disassembly cost of the old rear axle can be linked to the higher number of
fasteners as well as to the removal of welding seams that connect the brake support system to
the main axle body (comprising iron and steel respectively). Finally, due to the higher density
in steel components, the overall value of the new axle is higher, when considering the total
mass.

5.3 Assembly line planning

The industrial case study that the ALBA and DESA were tested on, involves the design of an
assembly line for two variants of a car’s engine; a gas and a diesel engine variant. Both products
share similar parts and some identical processes. However, certain processes are different in
terms of duration and required resources, i.e. they are variant-specific. Furthermore, the diesel
variant will be provided to the line in a more advanced assembly state which requires less
assembly processes than the gas variant does. In the collaboration environment that the
current case study is a part of, the engineers from the OEM producing the engines, provide
information about the products and line requirements. The case study was structured based
on a possible scenario that involve commissioned engineers, from a system integrator, who
perform simulations and provide process descriptions for assembling the product as well as the
relevant resources. The case study addresses the performance of the line balancing when all
information has been provided or uploaded and starts with the engineer that performs the line
balancing and simulation (Figure 53).

104
 Case studies and Results

Figure 53: Focus of the case study in a line design project.

The equipment used in the line, beyond the conveyor systems, consists of pneumatic and DC
power tools for screwing as well as cranes for moving large engine components and
transmission systems. Concerning the human resources of the current experiments, one
operator per station has been considered. Each process is associated with different process
configurations, which result in different process durations and running costs, depending on the
resource used. The processes examined in this case include operator tasks, namely reading the
product information and loading the product to the moving palette, followed by actual
assembly tasks, such as installing and tightening components. A portion of the precedence
diagram of the processes for the gas engine variant is shown in Figure 54. In this diagram, two
alternative configurations of process 3 are depicted with different durations and average
running costs. The process refers to the loading of the engine to the moving pallet, where two
alternatives have been proposed by the process and equipment specification engineers.
Beyond defining the process alternatives, the engineers also link the configurations to the
relevant resources.

Following the previous example, the first process alternative configuration of process 3 will use
the type of crane that has an investment cost of around 10k €, while the second one 7,5k €. At
the same time, for the first crane, the mean time between failures is considered being around
one month, while for the second one 12 days. For all process alternatives, all this information

105
Case studies and Results

will be used by the DESA during the simulation of the line, when the final performance criteria,
including the production rate, will be formed.

The relevant process data, including alternative configurations, have been already uploaded
and stored into the data repository and are retrieved by the agents. This action is the resulting
task of the “planning of process” and “planning of equipment” phases, carried out by the
system integrator’s engineers, as illustrated in Figure 53. In this case study, the engineer who
is in charge of the assembly line design task initiates the entire process through a web-browser,
which communicates with all software agents, hosted in one server.

Figure 54: Extract from the precedence diagram of the processes of the case study.

The parameters used are depicted in Table 6. Fifty two (52) processes have to be balanced and
simulated in a line, with a maximum cycle time (CT) of 80 seconds.

106
 Case studies and Results

Table 6: Case study specifications.

Parameters Input

Product variants Gas engine variant

Diesel engine variant (Engine B)
(Engine A)
Processes 33 19

Alternative process 2 4
configurations
Max. CT 80

Demand Uniform; Mean value: 750; Standard deviation: 30

Once this information has been entered, the ALBA and DESA carry out all the steps described
in section 3.4.3. The final result, i.e. the best alternative based on the criteria and constraints
provided by the user, is presented to the user in the form of a Nummi Wall. The Nummi Wall is
essentially a chart, showing the distribution of the processes among the different stations. The
processes are per station and based on the product they refer to, are visually differentiated
through their colours. It should be noted, that the financial figures stated in this case study are
indicative, since they have been modified to maintain the confidentiality of the companies that
provided them.

A series of experiments was carried out, and in all cases, the average computation time of each
run was under 3 minutes for the assembly line balancing and simulation combined. All software
agents are hosted in standard servers with Intel® Xeon® 5600 series processors. In the specific
series after the initial generation of alternatives by the ALBA, the best 105, based on equation
2 of section 3, were then simulated by the DESA. In actual industrial practice, this process
normally requires that the assembly engineer identifies and defines all required data, then
perform the ALB process, and finally use the results derived for building and executing the
corresponding simulation model. In most cases, it is required that the ALB and DES process be
re-launched many times, until the expected performance is obtained. In the proposed
approach, the entire process is carried out automatically. More importantly, the user can
choose whether the DESA can be run after the ALBA, or if the results should be supplied only
by the ALBA. Thus the engineer has the flexibility to perform line balancing in different phases

107
Case studies and Results

of the line design project, when different data are available. When the best alternative is
visualised by the user, a Nummi wall appears showing the distribution of the processes in the
stations for the best alternative (Figure 12). The Nummi wall contains all the processes for each
product, as they are distributed per station. In the screen below, the dark grey columns
represent the gas engine processes, while the blue ones represent the diesel engine processes.
An additional functionality, allows the user to make a mouse hover over each process, to bring
up an annotation box that shows the respective process information, such as process times,
total station time as well as the relevant product. Each station has a column for the processes
of each product. The right section of the GUI is used for inputting the relevant parameters, such
as maximum cycle time, maximum number of stations, labour cost per hour for the given
simulation, weights for line balancing and simulation etc. prior to running the ALBA and DESA.
Furthermore, the user can also select if the DESA can be run through a check box, by providing
the flexibility only to enable line balancing when the relevant simulation data are not available.

Figure 55: User interface of the web-application for the control of the developed agents.

Regarding the results of the line balancing and simulation, the five best alternatives of the case
presented can be seen in Table 7. If only the results of the ALBA were to be considered, namely
Cycle Time, Number of stations and Investment costs, the Alternative #47 would be regarded

108
 Case studies and Results

as the best solution. This alternative has the lowest combination of cycle time and number of
stations as well as the lowest investment cost. However, considering the actual production rate
recorded by the DESA as well as the Running costs, it was decided that the alternative #117 was
the best solution (Figure 13). More specifically, although this alternative based on the balancing
results has a higher cycle time it has also 13% higher production rate, while requiring lower
costs to run. This example clearly shows how the results, coming directly from the balancing
line can be overcome by results following simulation. The simultaneous execution of line
balancing and simulation may support the selection of the proper alternative, in the early
stages of line design, avoiding unnecessary iterations. Finally, through the central data
repository, the existing process configurations from previously formed assembly systems can
be reused, providing additional alternatives.

Table 7: Five best results generated by the DESA.

Line alternative Cycle Time Number of Production rate Investment Running costs
stations (engines/day) costs (€) (€/ day)
Alt#117 79 9 737,6451613 65100 19105,43148

Alt#11 78 10 737,9032258 61300 20295,64242

Alt#19 80 9 738,7741935 61000 20726,28161

Alt#47 78 9 653,2903226 43500 20027,04439

Alt#113 79 9 645,5806452 61800 19002,33865

109
Case studies and Results

Figure 56: Running costs, Production Rate and Cycle Time of the five best alternatives
generated by the ALBA and DESA.

5.4 Assembly Instructions generation and visualisation using Augmented reality

goggles

The case study of the assembly instructions generation concerned the generation and
visualisation of AR instructions of a Diesel Particulate Filter (Figure 57). More specifically, the
case study revolved around the assembly of the heat cover parts on the DPF which include the
actual cover as well as two metal tubes and their support.

110
 Case studies and Results

Figure 57: DPF CAD assembly.

The CAD file of the DPF was processed by the APDG where the components to be assembled
were placed on the first level of the assembly tree. The pre-assembled components were
placed as sub-assemblies. The end results provided the assembly tiers (four – including the
base part) as well as the assembly directions of the components and the referenced fasteners.
The APDG output was processed by the relevant application and the instructions were
generated. The instructions were tested on an actual shopfloor were assembly personnel used
the application and provided feedback regarding the implemented technologies (including the
AR goggles). The hardware used were the Vuzix 1200XLD3 and the LG Nexus 54 which was used
as a handheld device for controlling the visualisation application (navigating through the
sequential assembly tiers).

3
https://www.vuzix.com/augmented-reality/products_star1200xld/
4
http://www.lg.com/gr/mobile-phones/lg-D821-nexus-5

111
Case studies and Results

Figure 58: Hardware used in the case study.

Regarding the product to be assembled, the DPF was placed on a support structure that is
normally used for its assembly. The AR marker (i.e. physical Image Target) was placed on this
structure (Figure 59).

Figure 59: (a) DPF fully assembled on the support structure and (b) DPF without cover parts.

In total ten people concerned with manual assembly operations participated in the
demonstrations, ranging from quality inspection managers to shopfloor operators. The
demonstration included the generation of the instructions as well as the use of the instructions
using the AR visualisations.

112
 Case studies and Results

Figure 60: Animation of the case being assembled on the DPF using AR.

The overall feedback that was provided is presented in three groups based on the focus of the
feedback, namely:

Software related
 The visualization is user friendly and easy to use and operate. This is important for
future developments since, if the tool is not user friendly, the shopfloor operators will
not use it (they cannot be forced to use tools or devices).
 Visual feedback should be provided in order to notify operators that the sequence is
complete.
 Clearer instructions should be provided (e.g. through text) on how to assemble each
part.
 Additional functionalities that enable the recognition of the parts that are assembled
(i.e. being picked up and used) would be relevant for future developments.

Hardware related
 The field of view of the AR goggles used is narrow compared to the “180°” operators
are used to. Goggles with greater field of view should be used / investigated in the
future.
 The screen should be even more transparent to allow easy looking-through.
 The handheld device is sufficient for communication and could be used as is.
 The light entering from the bottom and sides of the glasses is sometimes a problem as
it interacts with the inside part of the goggles’ screen.

113
Case studies and Results

 The cables connecting the goggles should be reduced; this is very important when
working in the operation station to avoid tangling of the cables around objects in the
station.
 The marker has to be always visible to the camera in order to calculate the relevant
position of the 3D geometries and this can be a problem; an inertial sensor embedded
into the AR goggles can be used to try and keep the image, even if not 100% precise,
when moving away the goggle’s field of view from the marker/assembly.

General recommendations
 This technology in its current form is more suitable for long sequence operations or
complex activities, not for short operations; the system could be especially ideal for
training new operators, rather than for directly supporting the assembly people for
simple short operations.

114
Conclusions and Discussion

6 CONCLUSIONS AND DISCUSSION

6.1 Discussion on the results

Through the case studies presented in the previous sections, some conclusions can be drawn
for the methods proposed as well as their implementation. The main method suggested,
concerned the generation of assembly components’ precedence diagrams which has been
recognised by the academic and industrial communities as an important point of improvement
to the current practices [9].

The method was implemented through the APDG as an add-on to a commercial CAD software.
The results showed that precedence diagrams using assembly tiers can be generated for
complex assemblies within minutes which, compared to paper based methods is an evident
improvement. More specifically, regarding the duration of the execution of the method as it
was implemented, it can be observed that even with modest hardware most assemblies were
processed under ten minutes. Although a direct link between one or more characteristics to
the duration of the algorithm cannot be obtained, it can nonetheless be seen that the
characteristics that have a bigger impact on the algorithm’s performance are the number of
the components to be assembled (first level of assembly tree) as well as the number of
geometric features. Additionally, it can be observed that for most assembly files, the algorithm,
required less than two minutes. In the current literature there is no parallel method providing
the generation of components’ precedence diagrams and fasteners’ associations directly from
CAD files. The nearest approach in terms of input and output is the generation of assembly
sequences, which, reduces the number of alternative process sequences to one. Therefore, a

115
 Conclusions and Discussion

direct benchmarking of the algorithm’s performance is not feasible. Finally, the fact that the
outcome can be stored in a digital form (through the physical storage of the resulting
spreadsheet or through a database) is an improvement with respect to collaborative and reuse
aspects of engineering business processes.

The case study regarding the EoL value estimation, presented a method for using existing CAD
technology features in order to provide feedback to designers and engineers, concerning the
recyclability of their designs in the phases of EoL treatment. The present approach focuses on
the economic feasibility estimation of handling EoL products via a collision detection algorithm.
The corresponding application is developed as an add-on to a CAD system. Beyond the
comparison of designs in an automatic way, the results of this case study highlight the
importance of estimating the EoL process costs. In comparison to the results obtained when
the dismantling costs are not considered, there was an average drop of 12.5% of value per mass
unit for each design. This is due to the fact that the proposed method also considers manual
assembly costs. The proposed method may support the increasing effort of manufacturing
companies towards employing reuse and remanufacturing paradigms.

Regarding the results related to assembly planning (line balancing and simulation), it may be
underlined how an improved model for linking tasks (in terms of input – output of information)
along the different process planning phases emerges out of the use of cloud technologies. In
addition, the use of agents, for tasks that require input, i.e. for the final design of a line from a
number of engineers, may lead to better results. In the current industrial practice, constant
communication is necessary so that all information required for the design of a manufacturing
process be properly exchanged. Most parts of the data / information have to be collected and
exchanged at each iteration; an activity that is often counter-productive and time consuming.
By utilising this approach, the complexity of exchanging data and information in each phase can
be significantly reduced, while performing certain consecutive design phases concurrently
(balancing and simulation).

116
Conclusions and Discussion

Furthermore, since a series of alternatives may be provided in minutes, different demand

profiles or capacity configurations may also be investigated. For example, the initial
requirement for a line in terms of production rate can be re-evaluated with respect to cost,
since the design of the line is not treated as a sequence of static processes, but rather as a
dynamic aggregation of input, provided by engineers (Figure 61).

Figure 61: Dynamic relation between results of sub-steps and input of initial sub-steps.

The Augmented Reality case demonstrated one of the possibilities of how the APDG results can
be used during production execution for the generation and presentation of assembly
instructions. The feedback received from people in the industrial sector has been very positive
also highlighting the possible improvements for future implementations.

6.2 Limitations of the proposed approach

The introduction of the assembly tiers concept through the APDG algorithm aims at forming a
basis for further generation of assembly sequences. Through this approach, the solution space
for identifying the assembly relations of each component is reduced to the minimum possible
for each component (i.e. only the components that belong to the next tier), based primarily on
the geometric characteristics of the product. It was demonstrated that with existing
technologies, even the brute-force search for the identification of all assembly tiers can be
performed in minutes providing quick support to the engineers involved. The second part of
the APDG algorithm presented performs selective search to identify the connections between
components and fasteners and components belonging to sequential tiers. Limitations
regarding the performance of the algorithm have to do mainly with its application to open file
formats such as STEP files. This happens for two reasons; the first reason has to do with the

117
 Conclusions and Discussion

fact that the methods for collision detection in commercial CAD files are optimised for their
own proprietary formats. When a proprietary file is translated into an open source format, e.g.
STEP, all the information regarding its creation as well as the basic geometrical features are
converted to surfaces (vertices and faces). Due to this, interpretations take place, that result in
geometries that have rough features which need more time to be processed in terms of
interference checking (crucial function of collision detection) and visualisation (Figure 62).
Secondly, the resulting geometries are based on the simplified features of the original file (e.g.
a circle will be adapted as a polygon), causing the collision detection algorithm to perform
slower and to detect collisions that would normally not occur.

Figure 62: Example of cylinder geometry information as stored in proprietary

CAD format and as STEP.

Finally, the APDG algorithm only partially addresses flexible geometries. As an example, cables
are only addressed when they have been specifically defined as such through their position on
the assembly tree as well as their structure. A cable must be a sub-assembly with three or more
components; one representing the components that can be bent, and the rest being the plugs,
i.e. the parts that connect the cable to the rest of the assembly. The first component must

118
Conclusions and Discussion

contain the word “cable” in its naming. The algorithm completely removes the part that can be
bent and tries to identify how the rest of the components can be removed. This, is a controlled
way of dealing with cables and the matter should be further researched.

Regarding the calculation of EoL product values, the proposed application uses monetary rates
that are present-day. However, automotive products’ lifecycles typically exceed five years,
therefore the most appropriate values would be the forecasted ones for this timeframe.
However, the reliable forecasting of these values is not feasible, therefore, this is a constraint
that cannot be easily overcome. As a matter of fact the prices of relevant metals cannot be
forecast or estimated for far periods into the future. Consequently, the EoL indicator proposed,
or any other indicator reviewed, must be regarded as a metric based on the current state of
business.

The applications related to Assembly Line Balancing and Simulation make use of a specific data
model; since the model was constructed based on existing work that can be found in academic
literature, it is limited in terms of the cases it can be used for. The same is true for the balancing
and simulation algorithms which make use of data from this model.

Finally, the Augmented Reality application is limited mainly by the current state of hardware
that are available for this type of applications. In section 5.4, the limitations of the current
application were documented as feedback from industrial personnel. In order for applications
such as the one proposed to be applicable in real industrial scenarios, more flexible types of
hardware must become available.

6.3 Outlook and indications for future continuation of the presented work

Future work regarding the methods presented should be mainly focused on the APDG
algorithm since this is the core of the research presented. More specifically, the further
exploration of rotation of the components beyond their constrained axes as well as their
motion in further axes should be investigated. This will require the identification of similar
constraints defined by different CAD software regarding rotating components. Moreover, this

119
 Conclusions and Discussion

should be accompanied by the simplification of the collision detection routines through

heuristics for the faster execution of the algorithm. Regarding components that are flexible,
such as cables or chains, future research should be concentrated on the inclusion of these
components without any requirements on their position in the assembly tree. As far as the EoL
value estimation is concerned, the cost elements related to transportation, as well as the
removal of wastes and energy costs should be investigated. A possible extension of the
algorithm, could also address the inclusion of different values for possibly reusable
components. However, this would require the labelling of these components as reusable by the
designers.

As described in the previous section, the applications related to assembly planning are mainly
restricted by the information that can be stored -and therefore used- by the data model.
Consequently, the main consideration should be to further extend the data model. However,
as a general remark, it must be pointed out that the use of cloud technologies (Software as a
Service etc.) can enhance the efficiency of engineers involved in production design, partially
through the reduction of the time and effort spent, formerly required for the communication
of previous or on-going task results. Nonetheless, these solutions have only been recently
investigated in the context of product design, but not extensively for process design. Additional
effort should be dedicated to the utilisation of such technologies in the manufacturing systems’
design domain. Moreover, the information to be shared, residing in the engineering files of an
organisation, in an extended engineering environment, should become more easily obtainable
by those partners who do not possess the same commercial platforms as the OEM. As described
in section 2, work has been done in developing open standards in order for product files to be
shared between different commercial CAD platforms (such as STEP), while no standards have
been created and accepted by many industrial companies for process related engineering files,
except for machining (e.g. STEP-NC). Therefore, additional work should be carried out towards
the development of methods for sharing production related information and knowledge. In the
future, it is expected that more effort will be devoted towards the development of advanced
methods for controlling the access to shared technical data and knowledge in the extended
engineering environment, comprising departments and engineering teams of the same

120
Conclusions and Discussion

organisation. One of the most common approaches to be investigated is the Role Base Access
Control (RBAC) which has two distinctive advantages; the first is that all access rights can be
calculated on the basis of the user’s initially specified role. The second refers to the access to
certain data objects in repositories, which can also be controlled by linking data objects to
people through their relationships to different roles. This approach does require an additional
software component that will include workflow and task management. Finally, towards the
direction of seamless collaboration, additional software services integrated into the proposed
multi-agent system will be developed and integrated. The services will focus on the automatic
extraction of engineering information from legacy CAx files, such as process simulation files
created by engineers during the process and equipment specification phases. This is part of the
ongoing effort towards the elimination of non-adding value tasks that engineers are bound to
perform in modern extended engineering environments.

As a final point, in order to increase the applicability of the methods proposed towards the
automated generation and visualisation of Augmented Reality instructions, future research
should mainly be concerned with the development of new hardware that will be acceptable by
the people that perform assembly operations. Although new hardware has been developed
that might be acceptable, such as the Microsoft HoloLens5, they should be thoroughly tested
in industrial environments and if needed refined. In the context of the methods proposed, the
use of additional tracking mechanisms that do not require markers but are rather only
dependent on the inclusion of any part of the product in the user’s field of view for achieving a
reference coordinate system, should be considered.

5
https://www.microsoft.com/microsoft-hololens/en-us

121
References

7 REFERENCES

[1] ACEA, “The Automotive Industry: Pocket Guide,” 2013.

[2] ACEA, “The Automobile Industry: Pocket Guide,” 2014.

[3] E. Westkämper, “Assembly and Disassembly Processes in Product Life Cycle

Perspectives,” CIRP Annals - Manufacturing Technology, vol. 52, no. 2, p. 579–588, 2003.

[4] G. Chryssolouris, Manufacturing Systems: Theory and Practice. 2nd ed., New York:
Springer-Verlag, 2006.

[5] T. Eng, Z.-K. Ling, W. Olson and C. McLean, “Feature-based assembly modeling and
sequence generation,” Computers & Industrial Engineering , vol. 36 , no. 1, pp. 17-33,
1999.

[6] T.-D. Nguyen, R. McFarland, M. Kleinsorge, J. Krüger and G. Seliger, “Adaptive

Qualification and Assistance Modules for Manual Assembly Workplaces,” Procedia CIRP,
vol. 26, pp. 115-120, 2015.

[7] G. Chryssolouris, D. Mavrikios, D. Fragos and V. Kara, “A virtual reality-based

experimentation environment for the verification of human-related factors in assembly
processes,” Robotics and Computer-Integrated Manufacturing, vol. 16, no. 4, pp. 267-
276, 2000.

123
 References

[8] N. Papakostas, G. Michalos, S. Makris and D. Zouzias, “Industrial applications with

cooperating robots for the flexible assembly,” International Journal of Computer
Integrated Manufacturing , vol. 24 , no. 7, pp. 650-660., 2011.

[9] S. Hu, J. Ko, L. Weyand, H. ElMaraghy, T. Lien, Y. Koren, H. Bley, G. Chryssolouris, N. Nasr
and M. Shpitalni, “Assembly system design and operations for product variety,” CIRP
Annals - Manufacturing Technology, vol. 60, no. 2, p. 715–733, 2011.

[10] W. Hopp, Factory Physics. 3rd ed., Irwin/McGraw-Hill, 2007.

[11] D. Mourtzis, N. Papakostas, D. Mavrikios and S. Makris, “The role of simulation in digital
manufacturing: applications and outlook,” International Journal of Computer Integrated
Manufacturing , vol. 28, no. 1, 2015.

[12] N. Suh, “Theory of complexity, periodicity and the design axioms,” Research in
Engineering Design – Theory, Applications, and Concurrent Engineering, vol. 11, no. 2,
pp. 116-131, 1999.

[13] D. Mavrikios, K. Alexopoulos, V. Xanthakis, M. Pappas and G. Chryssolouris, “A Web-

Based Platform for Collaborative Product Design, Review and Evaluation,” Digital Factory
for Human-oriented Production Systems, pp. 35-56., 2011.

[14] K. Rodriguez and A. Al-Ashaab, “Knowledge WebBased System Architecture for

Collaborative Product Development,” Computers in Industry, vol. 56, no. 1, p. 125–140,
2005.

[15] G. Chryssolouris , D. Mavrikios , M. Pappas , V. Xanthakis and K. Smparounis , “A Web

and Virtual Reality-based Platform for Collaborative Product Review and Customisation,”
Collaborative Design and Planning for Digital Manufacturing, pp. 137-152, 2009.

[16] K. Alexopoulos, S. Makris, V. Xanthakis and G. Chryssolouris, “A web-services oriented

workflow management system for integrated digital production engineering,” CIRP
Journal of Manufacturing Science and Technology, vol. 4, no. 3, pp. 290-295, 2011.

[17] G. Pintzos, L. Rentzos, K. Efthymiou, N. Papakostas and G. Chryssolouris, “A Knowledge

based collaborative platform for the design and deployment of manufacturing systems,”
in International Conference On Product Lifecycle Management, Nantes, 2013.

124
References

[18] N. Perry , A. Bernard , F. Laroche and S. Pompidou , “Improving Design for Recycling –
Application to Composites,” CIRP Annals-Manufacturing Technology 61(1), pp. 151-154,
2012.

[19] C. Groger, M. Hillmann, H. Friedemann, B. Mitchang and E. Westkamper, “The

Operational Process Dashboard for Manufacturing,” Proceedings of the 46th CIRP
Conference on Manufacturing Systems, vol. 7, no. 1, pp. 205-210, 2013.

[20] A. Nee, S. Ong, G. Chryssolouris and D. Mourtzis, “Augmented reality applications in

design and manufacturing,” CIRP Annals - Manufacturing Technology, vol. 61, no. 2, p.
657–679, 2012.

[21] S. Makris, G. Pintzos, L. Rentzos and G. Chryssolouris, “Assembly support using AR

technology based on automatic sequence generation,” CIRP Annals - Manufacturing
Technology 62 , p. 9–12, 2013.

[22] L. Mathieu and B. Marguet, “Integrated Design Method to Improve Producibility based
on Product Key Characteristics and Assembly Sequences,” CIRP Annals - Manufacturing
Technology, vol. 50, no. 1, p. 85–88, 2001.

[23] R. Mantripagada and D. Whitney, “The Datum Flow Chain: a systematic approach to
assembly design and modeling,” Research in Engineering Design, vol. 10, pp. 150-165,
1998.

[24] Y. Peng, J.-X. Gong and K.-D. Huang, “Research on Reuse-oriented similarity evaluation
of BOM conceptual model,” Journal of National University of Defense Technology, vol.
33, no. 2, 2011.

[25] A. Hoque, P. Halder, M. Parvez and T. Szecs, “Integrated manufacturing features and
Design-for-manufacture guidelines for reducing product cost under CAD/CAM
environment,” Computers & Industrial Engineering, vol. 66, no. 4, pp. 988-1003, 2013.

[26] J. Paralikas, A. Fysikopoulos, J. Pandremenos and G. Chryssolouris, “Product modularity

and assembly systems: An automotive case study,” CIRP Annals - Manufacturing
Technology, vol. 60, no. 1, p. 165–168, 2011.

125
 References

[27] L. Qiong, X. Yang, P. Rushu, X. Jing and W. Yulin, “Research on the key technologies of
web parts library in product configuration system,” International Conference on
Advances in Engineering, vol. 24, p. 677–681, 2011.

[28] ISO, “Fasteners Standards catalogue,” International Organization for Standardization,

2014. [Online]. Available:
http://www.iso.org/iso/iso_catalogue/catalogue_ics/catalogue_ics_browse.htm?ICS1=
21&ICS2=060&ICS3=10.

[29] CATIA, “CATIA Version 5 - Sketcher,” 1999. [Online]. Available:

http://www.catia.com.pl/tutorial/z2/sketcher.pdf. [Accessed 2014].

[30] S. NX, “http://www.catia.com.pl/tutorial/z2/sketcher.pdf,” Siemens, 2014. [Online].

Available: http://www.plm.automation.siemens.com/en_us/products/nx/. [Accessed
2015].

[31] E. Wang and Y.-S. Kim, “Feature-based assembly mating reasoning,” Journal of
Manufacturing Systems, vol. 18, no. 3, p. 187–202, 1999.

[32] C. Mascle, “Feature-based assembly model for integration in computer-aided assembly,”

Robotics and Computer-Integrated Manufacturing, vol. 18, no. 5–6, p. 373–378, 2002.

[33] M. David and F. Rowe, “What does PLMS (product lifecycle management systems)
manage: Data or documents? Complementarity and contingency for SMEs,” Computers
in Industry, 2015.

[34] A. Bourjault, “Contribution a une approche méthodologique de l’assemblage

automatisé: elaboration automatiquede séquences opératoires,” Besançon (France):,
Université de Franche-Comté, 1984.

[35] L. S. Homem de Mello and A. Sanderson, “AND/OR Graph Representation of Assembly

Plans,” IEEE Transactions on Robotics and Automation , vol. 6, no. 2, p. 188–199, 1990.

[36] B. Rekiek, A. Doigui, A. Delchambre and A. Bratcu, “State of art of optimization methods
for assembly line design,” Annual Reviews in Control , vol. 26, no. 2, pp. 163-174, 2002.

[37] R. J. Riggs and S. Hu., “Disassembly liaison graphs inspired by word clouds,” Procedia CIRP
, vol. 7, p. 521 – 526, 2013.

126
References

[38] D. Kima, J. Portera and J. Buddhakulsomsiri, “Task time estimation in a multi-product

manually operated workstation,” International Journal of Production Economics, vol.
114, no. 1, p. 239–251, 2008.

[39] S. Lee and Y. G. Shin, “Assembly planning based on geometric reasoning,” Computers &
Graphics , vol. 14, no. 2, p. 237–250, 1990.

[40] D. Ben-Arieh and K. Kramer, “Computer-aided process planning for assembly: generation
of assembly operations sequence,” International Journal of Production Research, vol. 32,
no. 3, pp. 643-656, 1994.

[41] J. Pandremenos and G. Chryssolouris, “A neural network approach for the development
of modular product architectures,” International Journal of Computer Integrated
Manufacturing, vol. 24, no. 10, pp. 879-887, 2011.

[42] T. L. De Fazio, T. E. Abell, G. Amblard and D. Whitney, “Computer-aided assembly

sequence editing and choice: editing criteria, bases, rules and technique,” IEEE
International Conference on Systems Engineering, pp. 416-422, 1990.

[43] C. Pan, S. Smith and G. C. Smith, “Automatic assembly sequence planning from STEP CAD
files,” International Journal of Computer Integrated Manufacturing, vol. 19, no. 8, pp.
775-783, 2006.

[44] S.-F. Chen and Y.-J. Liu, “An adaptive genetic assembly-sequence planner,” International
Journal of Computer Integrated Manufacturing, vol. 14, no. 5, pp. 489-500, 2001.

[45] J. Li, Q. Wang and P. Huang, “An integrated disassembly constraint generation approach
for product design evaluation,” International Journal of Computer Integrated
Manufacturing, vol. 25, no. 7, pp. 565-577, 2012.

[46] P. Gu and X. Yan, “CAD-directed automatic assembly sequence planning,” International

Journal of Production Research, vol. 33, no. 11, pp. 3069-3100, 1995 .

[47] G. Dini, F. Failli, B. Lazzerini and F. Marcelloni, “Generation of Optimized Assembly

Sequences Using Genetic Algorithms,” CIRP Annals – Manufacturing Technology, vol. 48,
no. 1, p. 17–20, 1999.

127
 References

[48] N. Perry, A. Bernard, F. Laroche and S. Pompidou, “Improving Design for Recycling –
Application to Composites,” CIRP Annals-Manufacturing Technology, vol. 61, no. 1, pp.
151-154, 2012.

[49] G. Villalba, M. Segarra, J. Chimenos and F. Espiell, “Using the Recyclability Index of
Materials as a Tool for Design for Disassembly,” Ecological Economics, vol. 50, no. 3-4,
pp. 195-200, 2004.

[50] E. Lutters , F. van Houten , A. Bernard and E. Mermoz , “Tools and techniques for product
design,” CIRP Annals-Manufacturing Technology 63, pp. 607-630, 2014.

[51] D. Nelen, S. Manshoven, J. Peeters, P. Vanegas, N. D’Haese and K. Vrancken, “A

Multidimensional Indiacator Set to Assess the Benefits of WEEE Material Recycling,”
Journal of Cleaner Production, vol. 83, pp. 305-316, 2014.

[52] Y. Umeda, S. Fukushige, T. Mizuno and Y. Matsuyama, “Generating Design Alternatives

for Increasing Recyclability of Products,” CIRP Annals-Manufacturing Technology, vol. 62,
pp. 135-138, 2013.

[53] C. Becker and A. Scholl, “A survey on problems and methods in generalized assembly line
balancing,” European Journal of Operational Research, vol. 168, no. 3, pp. 694-715, 2006.

[54] A. Gupta and C. Maranas, “Managing demand uncertainty in supply chain planning,”
Computers and Chemical Engineering, vol. 27, no. 8-9, pp. 1219-1227, 2003.

[55] A. Scholl, Balancing and sequencing of assembly lines, Reading, MA, Heidelberg: Physica-
Verlag, 1999.

[56] K. N. and M. L., “Assembly Line Balancing: A Review of Developments and Trends in
Approach to Industrial Application,” Global Journal of Researches in Engineering, vol. 13,
no. 2, 2013.

[57] A. Scholl and C. Becker, “State-of-the-art exact and heuristic solution procedures for
simple assembly line balancing,” European Journal of Operational Research, vol. 168, no.
3, pp. 666-693, 2006.

[58] W. Xu and T. Xiao, “Strategic Robust Mixed Model Assembly Line Balancing Based on
Scenario Planning,” Tsinghua Science & Technology, vol. 16, no. 3, p. 308–314, 2010.

128
References

[59] E. Gurevsky, O. Battaïa and A. Dolgui, “Stability measure for a generalized assembly line
balancing problem,” Discrete Applied Mathematics, vol. 161, no. 3, p. 377–394, 2013.

[60] M. Chica, O. Cordon, S. Damas and J. Bautista, “A robustness and visualization model for
time and space assembly line balancing under uncertain demand,” International Journal
Production Economics, vol. 145, pp. 761-772, 2013.

[61] Z.-X. Guo, W.-K. Wong and Z. P. R. Li, “Modelling and Pareto optimization of multi-
objective order scheduling problems in production planning,” Computers and Industrial
Engineering, vol. 64, no. 4, pp. 972-986, 2013.

[62] B. Schroer and F. Tseng, “Modelling complex manufacturing systems using discrete event
simulation,” Computers & Industrial Engineering, vol. 14, no. 4, p. 455–464, 1988.

[63] H. Zupan and N. Herakovic, “Production line balancing with discrete event simulation: A
case study,” IFAC- Papers On Line , vol. 48, no. 3, p. 2305–2311 , 2015.

[64] A. Fasth, J. Stahre, T. Fassberg and G. Nordin, “iPod touch – an ICT tool for operators in
factories of the future,” Proceedings of 43rd CIRP International Conference on
Manufacturing Systems, 2010.

[65] S. Makris, G. Michalos and G. Chryssolouris, “A Pushlet-Based Wireless Information

Environment for Mobile Operators in Human Based Assembly Lines,” Proceedings of the
7th International Conference on Digital Enterprise Technology, pp. 622-631, 2011.

[66] Y. M. Chen, S. H. Hsieh and S. C. Kang, “Multi-touch Gestural 4D Viewing on Handheld

Devices,” Proceedings of the Conference on Computer Applications in Civil and Hydraulic
Engineering (CCACHE), pp. 116-122, 2011.

[67] H.-L. E. Chi, S.-C. Kang and X. Wang, “Research trends and opportunities of augmented
reality applications in architecture, engineering and construction,” Automation in
Construction, vol. 33, no. 1, pp. 116-122, 2013.

[68] A. Utsumi, P. Milgram, H. Takemura and F. Kishini, “Investigation of Errors in Perception

of Stereoscopically Presented Virtual Object Locations in Real Display Space,”
Proceedings of the 38th Human Factors and Ergonomics Society Annual Meeting, vol. 38,
no. 4, pp. 250-254, 1994.

129
 References

[69] F. Petit, “Interactive design of a product and its assembly system,” PhD thesis, Universite
Catholique de Louvain., 1999.

[70] S.-F. Chen, J. H. Oliver, S. Chou and L.-L. Chen, “Parallel Disassembly by Onion Peeling,”
Journal of Mechanical Design, vol. 119, no. 2, pp. 267-274, 1997.

[71] L.-M. Ou and X. Xu, “Relationship matrix based automatic assembly sequence generation
from a CAD model,” Computer-Aided Design, vol. 45, no. 7, pp. 1053-1067, 2013.

[72] G. Boothroyd, P. Dewhurst and W. Knight, Product Design for Manufacture and
Assembly, 3rd ed., Vols. Boothroyd G, Dewhurst P, Knight WA (2010) Product Design for
Manufacture and Assembly, 3rd ed. CRC Press, Taylor & Francis Group., CRC Press, Taylor
& Francis Group., 2010, pp. Boothroyd G, Dewhurst P, Knight WA (2010) Product Design
for Manufacture and Assembly, 3rd ed. CRC Press, Taylor & Francis Group..

[73] Eurostat, “Hourly Labour costs,” European Commission, 2015. [Online]. Available:
http://ec.europa.eu/eurostat/statistics-explained/index.php/Hourly_labour_costs.

[74] R. Eberhart and K. J. , “A new optimizer using particle swarm theory,” Proceedings of the
6th International Symposium on Micro Machine and Human Science, p. 39–43, 1995.

[75] Desmo-J, “Demo-J,” University of Hamburg, Department of Computer Science, 2012.

[Online]. Available: http://desmoj.sourceforge.net/home.html. [Accessed 2015].

[76] E. Ross, VB Scripting for CATIA V5: How to program CATIA macros, 2012.

[77] J. A. D. Framework, “JAVA Agent DEvelopment Framework,” 2014. [Online]. Available:

http://jade.tilab.com/.

[78] MetalPrices.com, “Metal prices,” 2015. [Online]. Available:

http://www.metalprices.com/.

130
BRIEF BIOGRAPHICAL NOTE – PUBLICATIONS

BRIEF BIOGRAPHICAL NOTE – PUBLICATIONS

Georgios S. Pintzos was born in Athens on the 18th of October 1986. He

received his Diploma in Mechanical and Aeronautics Engineering from
the Department of Mechanical Engineering and Aeronautics of the
University of Patras in September 2010. Being a PhD candidate at the
University of Patras from April 2011, he has been involved in a number
of EU funded research projects as well as in the educational activities of
the Laboratory for Manufacturing Systems and Automation. He is currently part of the
Industrialization team of BIC Shavers’ R&D department. He is a member of the Technical
Chamber of Greece (TEE).

During his work at the Laboratory for Manufacturing Systems and Automation he published a
number of peer reviewed papers in international scientific journals and conferences:

International journal publications

1. G. Pintzos, C. Triantafyllou, N. Papakostas, D. Mourtzis, G. Chryssolouris, "Assembly
Precedence Diagram Generation through Assembly Tiers Determination", International
Journal of Computer Integrated Manufacturing, Volume 29, Issue 10, pp. 1045-1057,
DOI 10.1080/0951192X.2015.1130260, (2016)
2. N. Papakostas, G. Pintzos, C. Giannoulis, G. Chryssolouris, "An agent-based collaborative
platform for the design of assembly lines", International Journal of Computer Integrated
Manufacturing, Volume 29, Issue 4, pp.374-385, DOI
10.1080/0951192X.2015.1066862, (2016)
3. N. Papakostas, G. Pintzos, C. Triantafyllou, "Computer-aided design assessment of
products for end of life separation and material handling", CIRP Annals - Manufacturing
Technology, Volume 64, Issue 1, pp.185-188, DOI 10.1016/j.cirp.2015.04.023, (2015)

131
 BRIEF BIOGRAPHICAL NOTE – PUBLICATIONS

4. S. Makris, G. Pintzos, L. Rentzos, G. Chryssolouris, "Assembly support using AR

technology based on automatic sequence generation", CIRP Annals – Manufacturing
Technology, Volume 62, No. 1, pp. 9–12, 10.1016/j.cirp.2013.03.095, (2013)
5. S. Makris, L. Rentzos, G. Pintzos, D. Mavrikios, G. Chryssolouris, "Semantic-based
taxonomy for immersive product design using VR techniques", CIRP Annals -
Manufacturing Technology, Volume 61, No.1, pp.147–150, DOI
10.1016/j.cirp.2012.03.008, (2012)

International conferences

1. G. Pintzos, M. Matsas, N. Papakostas, D. Mourtzis, "Disassembly Line Planning Through

the Generation of End-of-Life Handling Information from Design Files", 49th CIRP
Conference on Manufacturing Systems (CIRP CMS 2016), 25-27 May, Stuttgart,
Germany, (2016)
2. G. Pintzos, N. Nikolakis, K. Alexopoulos, G. Chryssolouris, "Motion parameters
identification for the authoring of manual tasks in digital human simulations: an
approach using semantic modelling ", 48th CIRP Conference on Manufacturing Systems
(CIRP CMS 2015), Procedia CIRP, 24-26 June, Naples, Italy, Volumue 41, pp. 752-758
(2016)
3. G. Pintzos, L. Rentzos, N. Papakostas, G. Chryssolouris, "A Novel Approach for the
Combined Use of AR Goggles and Mobile Devices as Communication Tools on the
Shopfloor", (DET 2014), 8th International Conference on Digital Enterprise Technology,
25-28 March, Stuttgart, Germany (2014)
4. G. Pintzos, L. Rentzos, K. Efthymiou, N. Papakostas, G. Chryssolouris, "A Knowledge
based collaborative platform for the design and deployment of manufacturing
systems", (PLM13) International Conference On Product Lifecycle Management, 6-10
July, Nantes (2013)
5. G. Pintzos, M. Matsas, N. Papakostas,G. Chryssolouris, "Production Data Handling Using
a Manufacturing Indicators’ Knowledge Model", (CIRP CMS 46) Procedia CIRP, 46th CIRP
Conference on Manufacturing Systems, 29-31 May, Sesimbra, Portugal (2013)

132
BRIEF BIOGRAPHICAL NOTE – PUBLICATIONS

6. G. Pintzos, M. Matsas, G. Chryssolouris, "Defining Manufacturing Performance

Indicators using Semantic Ontology Representation", (CMS2012), 45th CIRP Conference
on Manufacturing Systems, Athens, Greece, pp.7-13 (2012) PROCEDIA
7. M. Matsas, G. Pintzos, A. Kapnia, D. Mourtzis, "An Integrated Collaborative Platform for
managing Product-service Across their Life Cycle", Open Access Procedia CIRP, Vol 59,
pp. 220-226 , (2017)
8. C. Bikas, A. Argyrou, G. Pintzos, C. Giannoulis, K. Sipsas, N. Papakostas, G. Chryssolouris,
"An automated assembly process planning system", Procedia CIRP, 6th CIRP Conference
on Assembly Technologies and Systems (CATS), Volume 44, pp. 222-227, (2016)
9. S. Antoniou, L. Rentzos, G. Pintzos, K. Smparounis, D. Mavrikios, G. Chryssolouris,
"Ergonomic Validation of Manual Processes through Posture Detection Using Optical
Sensors", EuroVR Conference , 15-16 October, Lecco, Italy (2015)
10. N. Papakostas, G. Pintzos, M. Matsas, G. Chryssolouris, "Knowledge-enabled design of
cooperating robots assembly cells", 5th CIPR Conference on Assembly Technologies and
Systems, (CATS 2014) ,Procedia CIPR, Volume 23, pp. 165-170 (2014)
11. N. Papakostas, G. Pintzos, C. Giannoulis, N. Nikolakis, G. Chryssolouris, "Multi-Criteria
Assembly Line Design under Demand Uncertainty", (DET 2014), 8th International
Conference on Digital Enterprise Technology, 25-28 March, Stuttgart, Germany (2014)
12. S. Hesse, B. Wolf, M. Rosjat, D. Nadoveza, G. Pintzos, “Reference Model Concept for
Structuring and Representing Performance Indicators in Manufacturing”, IFIP
International Conference on Advances in Production Management Systems, APMS
2012, pp. 289-296 (2012)
13. L. Rentzos, G. Pintzos, K. Alexopoulos, D. Mavrikios, G. Chryssolouris, "Prototype
Designing with the help of VR Techniques: The Case of Aircraft Cabin", (DET 2011), ISBN
978-960-88104-2-6, 7th International Conference on Digital Enterprise Technology,
Athens, Greece, pp. 568-576 (2011)
14. L. Rentzos, G. Pintzos, K. Alexopoulos, D. Mavrikios, G. Chryssolouris, "An Analysis of
Human-Based Assembly Process for Immersive and Interactive Simulation", (DET 2011),
ISBN 978-960-88104-2-6, 7th International Conference on Digital Enterprise
Technology, Athens, Greece, pp. 558-567 (2011)

133
 BRIEF BIOGRAPHICAL NOTE – PUBLICATIONS

15. L. Rentzos, G. Pintzos, K. Alexopoulos, D. Mavrikios, G. Chryssolouris, "Interface

technologies for advanced virtual aircraft products – the VISION project", (ICEAF II), 2nd
International Conference of Engineering Against Fracture, Mykonos, Greece (2011)
16. L. Rentzos, G. Pintzos, K. Alexopoulos, D. Mavrikios, G. Chryssolouris, "Advancing the
interactive context of immersive engineering applications", (ICEAF II), 2nd International
Conference of Engineering Against Fracture, Mykonos, Greece (2011)

134