Академический Документы
Профессиональный Документы
Культура Документы
com
Available online at www.sciencedirect.com
Available
ScienceDirect
ScienceDirect
online atonline
www.sciencedirect.com
Available at www.sciencedirect.com
Procedia CIRP 00 (2019) 000–000
ScienceDirect
Procedia CIRP 00 (2019) 000–000
ScienceDirect www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
52nd
52nd CIRP
CIRP Conference
Conference on
on Manufacturing
Manufacturing Systems
Systems
Predictive
Predictive control
control
28th of
of aaConference,
CIRP Design synchronized
synchronized individual
individual
May 2018, production
production
Nantes, France
a, a,b
Heutmann,T.a,*, *, Schmitt,
Schmitt, R. H.
A new methodology to analyze Heutmann,T. the functional R.andH.a,b physical architecture of
existing products for anforassembly
Engineering (WZL) oforiented product family 30,identification
Fraunhofer Institute for Production Technology IPT, Steinbachstrasse 17, Aachen, 52074,Germany
a
Fraunhofer
a
Institute Production Technology IPT, Steinbachstrasse 17, Aachen, 52074,Germany
Laboratory for
b
Machine Tools and Production RWTH Aachen University, Campus-Boulevard Aachen, 52074, Germany
Laboratory for Machine Tools and Production Engineering (WZL) of RWTH Aachen University, Campus-Boulevard 30, Aachen, 52074, Germany
b
concept. Following, section 3 introduces and details the con- Determination of suitable measure strategies considering
cept to control a synchronized individual production predic- limitations due to tact pallets and tact lines
tively. Section 4 concludes the paper.
2.2. Production-control loop
2. Literature review
Production control is typically implemented by formulating
2.1. Synchronized individual production control loops [6]. The production process represents the con-
trolled system. Principally, manipulated variables are influ-
Transferring the synchronization to job-shop production by enced by disturbance values and thereon controlled by the con-
introducing a tact time has first been achieved by Baecke- trol unit. Deviations to be controlled are identified by a (yet
Heger [12]. Besides that, Gruss [13] introduces a two-stage simple) as-is-comparison. Manufacturing Execution Systems
timing model by formulating macro- and micro-tacts, accord- (MES) are often used in production for this purpose. The nom-
ing to lean production. A macro-tact may represent the lead inal value is provided by reference variables that, in the case of
time of one process step and aims to be a superior control unit; a production, are typically orders and their respective infor-
a micro-tact is one time-limited task within a macro tact [13]. mation (e.g., delivery dates). The actual value is represented by
However, macro-tacts are not equal production wide because the control variable, providing a feedback whether the refer-
they represent the longest time of any process competence cen- ence variable has been achieved. The approaches present in
ter. These are defined as optimized processes with defined task section 2.3 often focus on one control-loop component and its
sequences, results, and standardized interfaces [13]. Namely, specification but find different ways to approach the same
they form the product-individual production-process chain. problem: controlling a job-shop production.
Zwanzig [14] focuses on the timing in an individual production
itself by investigating influence factors, determining the tact 2.3. Job-shop control
time, and developing a procedure for implementation. Never-
theless, none of them can satisfy the requirements of a synchro- Generally, job-shop-control approaches can be classified by
nized individual production due to its special characteristics, the reference variable: either it is a complete reactive approach
which is allocation to tact pallets with fixed tact lines and buffer reacting to disturbances on a given schedule or it is a predic-
possibilities [15]. Therefore, the synchronized individual pro- tive/reactive approach aiming to adjust the production pro-
duction has been achieved by Ziskoven [15] ultimately. gramm with regard to due dates [21]. Based on this, existing
However, this approach only focuses on providing the gen- approaches to control a job-shop production can be classified
eral principle and on an initial scheduling algorithm but not on into five categories, depending on the reference variable, the
providing adaptive scheduling and control methods. Hence, controlled variable, and the correcting variable that results in
Humphrey [16] transfers the theoretical concept in industrial the controller type [22] :
application by developing mapping and scheduling algorithms
for the synchronized individual production. This approach in- Control of inventory (complete reactive)
cludes disturbances like express orders but focuses generally Control of capacity (complete reactive)
on rescheduling when discontinuities appear [17]. Control of load (complete reactive)
When implementing the synchronized individual production Knowledge-based control (predictive/reactive)
in toolmaking, Schuh et al. [18] identify that more than 75 % Control by rescheduling (predictive/reactive)
of tool components to be produced can be considered for a tact
pallet. Aiming to provide a more practical approach, which is However, none of these approaches consider the character-
applicable to more than 75 % of tool components, a principle istics of a synchronized individual production. Consequently,
to combine the synchronized job-shop production with the clas- typical approaches like reallocating one order to another ma-
sical job-shop production has been suggested [18]. However, chine, providing additional machines for rework, or late release
the parallel application of both methods may lead to increased of orders are insufficient because the tact pallet would be
flexibility, but requires a high degree of coordination in order changed. The equation of the sum of the orders’ cycle times
to avoid losses in machine utilization. with the tact time on the new tact pallet could be disproportion-
Other works in the field of the synchronized individual pro- ate at downstream process steps. Furthermore, existing ap-
duction are the optimization of tact lines using cluster analysis proaches of the predictive/reactive category knowledge-based
[19] and the improvement of process quality by focusing on control use root-cause analyses or downtime durations to select
involving employees in decision making processes [20]. measures. A linkage from existing disturbance situations to fu-
Consequently, the works present in the field of synchronized ture similar situation using context information has not been
individual production focus on planning and scheduling but not found in existing approaches.
on predictive controlling. Requirements on a predictive control
approach for a synchronized individual production are: 2.4. Self-optimizing job-shop control
Determination of hierarchical control levels In addition to classic job-shop control, control concepts for
Prediction of disturbances and their effects on each adher- individual production are also used in self-optimizing produc-
ence to delivery level tion planning and control. The aim is that the system decides
independently on goals and measures to achieve them [23].
T. Heutmann et al. / Procedia CIRP 81 (2019) 69–74 71
Author name / Procedia CIRP 00 (2019) 000–000 3
Correcting
variables
the production process. However, this approach also does not
information
cover the characteristics (see section 2.1) that are typical for the
Context
Actual
values
synchronized individual production. Especially the given limi-
tations due to tact pallets and tact lines would require an ad-
justed model. Ultimately, self-optimizing job-shop control still
responds to disruptions and is not proactive as desired, even Production Production
though an underlying expert system evaluates the effectiveness scheduling monitoring
of historical control decisions.
in small and medium companies in areas such as tool manufac- Piston Workshop Element boundary
Cylinder
turing. Furthermore, higher manufacturing costs need to be val-
W
ued in practice against the actual costs of delay.
Due to the nature of a synchronized individual production, Q
T, p
adherence to delivery can be interpreted internally as well as mo mi
externally on in total four levels (Fig. 2). This differentiation is
necessary due to the determination of consequences if an effect Organization/
System boundary
is identified or predicted. However, this is only relevant for management
controlling the synchronized individual production. The exter- mo: outflowing mass flow Q: heat flux
nal customer-delivery date can be split up into internal delivery mi: inflowing mass flow T: temperature
dates for each tact lines, tact time, and single cycle times for p: pressure W: work conducted
each part. These are the adherence to delivery interpretation
Fig. 3. Derivation of influence types using thermodynamic analogy.
levels in a synchronized individual production and be used as
internal reference variables (Fig 2).
A workshop represents the separated space and its elemen-
1 Customer delivery date external tary boundary, the system boundary. Machines, employees, and
work pieces are an integral part of the system as state variables.
2 Tact line delivery date internal Dispositive factors in form of managerial and organizational
Level
Fig. 2. Interpretation levels of adherence to delivery for orders. 3.6. Disturbance detection
3.3. System: Synchronized individual production Employees or machines provide a feedback to the system
about actual cycle times. This is done using of a production-
The controlled system is a synchronized individual produc- data-acquisition system. A terminal to manually record the time
tion with relevant characteristics as described in section 1 and can be provided to the employees. Sensors can be integrated
section 2.1. into machines to acquire the necessary data automatically. Be-
sides that, the product can also be enabled to provide the infor-
3.4. Output mation by the use of radio frequency identification (RFID)
[31].
The primary control variable is the actual delivery date for Based on the information provided, an as-is-comparison of
each order. It is used for comparison with the reference variable the cycle time can identify whether any disturbance affects the
and formulation of the adherence to delivery indicators. schedule (level 1 in Fig. 2). If there is any negative deviation
The applicable evaluation whether an order is on time can from standard, the consequences on the adherence to delivery
only be executed by the tact line delivery date (level 2 in Fig. 2) on level 2 and 3 can be derived. Furthermore, side effects on
since packaging and shipping cannot be influenced by produc- other orders designated to a different tact pallet and their ad-
tion control. Consequently, level 2, 3, and 4 can be referred to herence to delivery can be deduced.
as internal control variables and level 1 as external control var-
iable. 3.7. Disturbance prediction
of cost of delay is appropriate for all orders when it is inter- 𝑤𝑤𝑃𝑃𝑃𝑃,𝑘𝑘 and 𝑤𝑤𝐹𝐹𝐹𝐹,𝑘𝑘 are weighting factors that need to be deter-
preted as risk costs, named 𝐶𝐶𝐷𝐷,𝑅𝑅,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 . Determining the oc- mined by experts for each production-process step. It applies:
curring costs is possible due to the specific figures stated in the
contracts. Therefore, a possibility to address adherence to de- 𝑤𝑤𝑃𝑃𝑃𝑃,𝑘𝑘 , 𝑤𝑤𝐹𝐹𝐹𝐹,𝑘𝑘 ∈ [0,1] (8)
livery on level 2 is achieved. If a disturbance and its conse-
quences affect only the manufacturing process in any produc- The factor 𝐹𝐹𝑃𝑃𝑃𝑃,𝑘𝑘 is determined by intervals and characteris-
tion-process step (level 3 and 4) but not the production-lead tics of the company. As an example, the internal reliability of
time, internal risk costs for a delay 𝐶𝐶𝐷𝐷,𝑅𝑅,𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 appear to any material provision by logistics is characterized by the failure
costs directly related to the customer. Effects on machine utili- rate, including wrong, too early, or too late supplied goods. The
zation, personal costs, or additional waiting time that does not interval may be high for 𝑥𝑥 ≥ 2 (evaluation = 1), medium for
influence the adherence to delivery in this case but increases 2 < 𝑥𝑥 ≥ 0.5 (evaluation = 2), and low for 𝑥𝑥 < 0.5 (evalua-
costs for e.g. buffers may be present. Consequently, the indi- tion = 3). Characteristics and intervals of factor 𝐹𝐹𝐹𝐹𝐹𝐹,𝑗𝑗 need to
vidual severity 𝑆𝑆𝑖𝑖 for each DVT 𝑖𝑖 in form of risk cost of delay be determined by experts in the same way.
𝐶𝐶𝐷𝐷,𝑅𝑅,𝑖𝑖 is defined for all affected orders 𝑗𝑗 as: If the severity and/or the occurrence of one disturbance
value type exceeds a defined threshold, correction measures are
𝑛𝑛
applied to prevent the disturbance’s occurrence generally. The
𝑆𝑆𝑖𝑖 = 𝐶𝐶𝐷𝐷,𝑅𝑅,𝑖𝑖 = ∑(𝐶𝐶𝐷𝐷,𝑅𝑅,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,𝑖𝑖 + 𝐶𝐶𝐷𝐷,𝑅𝑅,𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖,𝑖𝑖 ) (2) threshold is an upper action limit (UAL), which is determined
𝑖𝑖=1 by company experts.
𝑚𝑚
(3) 3.8. Correcting variables
𝐶𝐶𝐷𝐷,𝑅𝑅,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,𝑖𝑖 = ∑ 𝐶𝐶𝐷𝐷,𝑅𝑅,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐,𝑗𝑗
𝑗𝑗=1
𝑚𝑚 Corrective measures represent the controller’s result and can
𝐶𝐶𝐷𝐷,𝑅𝑅,𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖,𝑖𝑖 = ∑(𝐶𝐶𝐷𝐷,𝑅𝑅,𝑠𝑠𝑠𝑠𝑠𝑠−𝑢𝑢𝑢𝑢,𝑗𝑗 + 𝐶𝐶𝐷𝐷,𝑅𝑅,𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡,𝑗𝑗 be applied reactively as well as proactively. Typically, the for-
(4) mer are short-term actions to diminish any effects that have had
𝑗𝑗=1
an influence already. The latter are often long-term strategies
+ 𝐶𝐶𝐷𝐷,𝑅𝑅,𝑛𝑛𝑛𝑛𝑛𝑛−𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢,𝑗𝑗 )
to improve the robustness of the production process against dis-
turbances but can also be ad-hoc activities preventing any fore-
with 𝐶𝐶𝐷𝐷,𝑅𝑅,𝑠𝑠𝑠𝑠𝑠𝑠−𝑢𝑢𝑢𝑢 = Risk cost for necessary set-ups
casted negative influence.
𝐶𝐶𝐷𝐷,𝑅𝑅,𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = Risk cost for necessary transport
Short-term measures are (e.g.):
𝐶𝐶𝐷𝐷,𝑅𝑅,𝑛𝑛𝑛𝑛𝑛𝑛−𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 = Risk cost for resulting non-utilization
Overtime and additional shift
The determination whether a disturbance affects only inter-
Temporary workers and jumpers
nal delivery dates or also customer delivery dates can be
Rescheduling
achieved by risk consequences.
Outsourcing, if a competent and trustworthy partner exists
Besides calculating the severity, the occurrence of a disturb-
ance affecting the production process also needs to be predica-
Long-term measures are (e.g.):
ble. Heutmann, Tils and Schmitt [8] identify relevant data to
consider for each DVT. Then, the occurrence can be calculated
Additional machines
taking the performance of the DVT (see section 3.5) into ac-
Alternative production schedules
count. The underlying assumption is: The higher the perfor-
Late schedule release and late production start
mance, the lower the probability of occurrence of a DVT. The
Safety stocks in form of semi-finished products that are
performance of the DVTs machine, employees, and work piece
equal for some product variants
can be calculated by the formulas of Bilsing [2]. The disposi-
Establishing partnerships for outsourcing on a short notice
tive factors are represented by the organizational performance
𝑂𝑂𝑂𝑂 and can be calculated as:
The selection of one or more appropriate short-term meas-
1 ure(s) is due to its effectiveness and availability (i.e. for out-
𝑂𝑂𝑂𝑂 = × (𝑃𝑃𝑃𝑃 + 𝐹𝐹𝐹𝐹) (5)
2 sourcing). Long-term measures are applied also by their effec-
tiveness but also according to a cost-benefit determination.
with the process quality 𝑃𝑃𝑃𝑃 for each production-process
step 𝑘𝑘 as
4. Conclusion
𝑃𝑃𝑃𝑃 = ∑(𝑤𝑤𝑃𝑃𝑃𝑃,𝑘𝑘 × 𝐹𝐹𝑃𝑃𝑃𝑃,𝑘𝑘 ) , for ∑ 𝑤𝑤𝑃𝑃𝑃𝑃,𝑘𝑘 = 1 (6)
This paper introduces a concept for the predictive control of
a synchronized individual production. Its aim is to increase the
indicator adherence to delivery. As control variables, external
and the fulfilment degree 𝐹𝐹𝐹𝐹 of the defined requirements
delivery dates from orders and internal delivery dates from cy-
as:
cle time, tact time, and tact line are used.
The typical conditions of a synchronized individual produc-
𝐹𝐹𝐹𝐹 = ∑(𝑤𝑤𝐹𝐹𝐹𝐹,𝑘𝑘 × 𝐹𝐹𝐹𝐹𝐹𝐹,𝑘𝑘 ) , for ∑ 𝑤𝑤𝐹𝐹𝐹𝐹,𝑘𝑘 = 1 (7)
tion are considered by determining the manipulated values: ca-
pacity of machines, capacity of personnel, capacities of buffers,
74 T. Heutmann et al. / Procedia CIRP 81 (2019) 69–74
6 Heutmann et al./ Procedia CIRP 00 (2019) 000–000
tact time, allocation to tact pallets, and sequence on one tact Hanser; 2014.
[7] Heutmann T, Tils A, Schmitt R. Time-related uncertainties in individual
pallet. Disturbances that influence the tact time and thus the manufacture. Zeitschrift für wirtschaftlichen Fabrikbetrieb ZWF
lead time are classified into four categories: machine, part, per- 2017;112:877-880.
sonnel and organization/management. The as-is-comparison to [8] Heutmann T, Tils A, Schmitt RH. Beruecksichtigung der Zeitunsicher-
heit für eine robuste, synchronisierte Einzelfertigung. In: Riedel R, edi-
identify any discontinuity is achieved by the use of a time re- tor. Smarte Produktion und digitale Vernetzung. Chemnitz: TU Chem-
cording for the cycle time. Furthermore, risks are quantified nitz IBF, Wissenschaftliche Schriftenreihe des Institutes für Be-
and evaluated to predict time deviations. Based on this, proac- triebswissenschaften und Fabriksysteme, Sonderheft 24; 2018. p. 123-
tive as well as reactive measures are established to handle dis- 132.
[9] Albach H, editor. Industrielles Management. Wiesbaden: Gabler; 1993.
continuities appropriately. Their selection is achieved due to [10] Hirsch BE, editor. CIM in der Unikatfertigung und -montage. Heidel-
determining the effectiveness and efficiency for each measure. berg: Springer; 1992.
Compared to classical job-shop-control approaches, the [11] Wikner J, Rudberg M. Integrating production and engineering perspec-
tives on the customer order decoupling point. In: IJOPM 2005;25:623-
method promises to be more adequate by considering the con- 641.
ditions of a synchronized individual production. Furthermore, [12] Baecke-Heger F. Produzieren im Takt. Markt und Mittelstand; 2008.
the predictive approach considering context information aims 48-51.
[13] Gruss R. Schlanke Unikatfertigung. Dissertation, Cottbus; 2010.
to react knowledge-based before a disturbance affects the [14] Zwanzig F. Taktung der Unikatfertigung am Beispiel des Werkzeug-
schedule. Based on this, the successful industrial application of baus. Dissertation, Aachen; 2009.
the principle is possible. [15] Ziskoven H. Methodik zur Gestaltung und Auftragseinplanung einer
getakteten Fertigung im Werkzeugbau. Dissertation, Aachen; 2013.
[16] Humphrey SE. Mapping and scheduling algorithms for synchronized
5. Outlook individual production. Dissertation, Aachen; 2016.
[17] Schmitt RH, Ellerich M, Humphrey S. Multi-objective allocation of
To validate the approach, two different industrial applica- customized orders to production-line networks. CIRP Annals
2016;65:429-432.
tions from tool and die manufacturing are selected. The first [18] Schuh G, Pitch M, Ziskoven H, Lieb H, Kelzenberg C. Getaktete Fer-
focuses on injection molding tools with the process steps: mill- tigung im Werkzeugbau. Aachen: WZL RWTH Aachen; 2015.
ing, grinding, eroding and assembling. The second company [19] Fels A, Ellerich M, Schmitt R. Cluster analysis for enhancing process
quality in job shop production. In: Silva Gomes JF, Meguid SA, editors.
concentrates on punching tools, forming tools and injection Proceedings IRF2018. Porto: FEUP-INEGI; 2018. p. 1413-1422.
molding tools with the following process steps: cutting, pre- [20] Elser H, Fimmers C, Groggert S, Schmitt RH, Brecher C. Process qual-
grinding, milling, grinding, eroding and assembly. The concept ity improvement through collaboration in synchronized individual pro-
duction companies. Procedia CIRP 2018;67:589-594.
is implemented in a simulation and various scenarios are car- [21] Schwartz F. Stoerungsmanagement in Produktionssystemen. Disserta-
ried out in order to influence the system. Based on historical tion, Hamburg; 2004.
data and event validity tests, common risk types are set to ex- [22] Niehues M, Blum M, Teschemacher U, Reinhart G. Adaptive job shop
treme values to verify the approach. In addition, the simulation control based on permanent order sequencing. Production Engineering
2018;12:65-71.
is used by production schedulers and production employees as [23] Permin E, Bertelsmeier F, Blum M, Buetzler J, Haag S, Kuz S,
well as validated by expert interviews. Results of the validation Oezdemnir D, Stemmler S, Thombansen U, Schmitt R, Brecher C,
will be published in future papers. Schlick C, Abel D, Poprawe R, Loosen P, Schulz W, Schuh G. Self-
optimizing production systems. Procedia CIRP 2016;41:417-422.
[24] Mueller F, Roeber T, Bauernhansl T. Classification of self-optimization
Acknowledgements approaches for manufacturing systems for discrete goods production.
WT Werkstattstechnik 2017;107:154-162.
[25] Schlick C, Stich V, Schmitt R, Schuh G, Ziefle M, Brecher C, Blum M,
This paper is a result of the research project PARSyP − Pre- Mertens A, Faber M, Kuz S, Petruck H, Fuhrmann M, Luckert M,
dictive Analytics for Robust Synchronized Production. It is Brambring F, Reuter C, Hering N, Groten M, Korall S, Pause D,
funded within the BMBF funding measure KMU-innovative: Brauner P, Herfs W, Odenbusch M, Wein S, Stiller S, Berthold M. Cog-
nition-enhanced, self-optimizing production networks. Integrative pro-
Information and Communication Technologies under the grant
duction technology. Cham: Springer; 2017. p. 645-743.
number 01IS16034A-F. [26] Gade PA, Chavan RG, Bhavsar DN. Reduction in set up time by single
minute exchange of dies SMED methodology. IJSTR 2015;5:364-366.
References [27] Schuh G, Stich V, Nyhuis P, Franzkoch B Hempel T, Hering N, Ort-
linghaus P, Potente T, Schiemann D, Troeger K. Cyber physical pro-
duction control. Integrative Produktion, Aachen: Shaker; 2014. p. 121-
[1] Fuchs S. Wertstromorientierte Auftragsfreigabe bei dyamischen 141.
Engpaessen in der Produktion nach dem Werkstattprinzip. Dissertation, [28] Moran MJ, Shapiro HN, Boettner DD, Bailey MB. Fundamentals of en-
Aachen; 2013. gineering thermodynamics. 8th ed. Hoboken: Wiley; 2014.
[2] Bilsing A. Kennzahlengestützte Bewertung der technologischen Leis- [29] Gutenberg E. Grundlagen der Betriebswirtschaftslehre. 5th ed. Berlin,
tungsfähigkeit der Fertigung im Werkzeug- und Formenbau. Disserta- Heidelberg, New York, London, Paris, Tokyo: Springer; 1960.
tion, Aachen; 2007. [30] Dyckhoff H. Produktionstheorie. 5th ed. Berlin, Heidelberg: Springer;
[3] Klocke F, Schuh G, Sauer A, Frick L, Klotzbach C. Werkzeugbau in 2006.
China - Chance oder Bedrohung? Aachen: WZL RWTH Aachen; 2006. [31] Niehues M, Buschle F, Reinhart G. Adaptive job-shop control based on
[4] Schuh G, Stich V, editors. Produktion am Standort Deutschland. FIR an permanent order sequencing. Procedia CIRP 2015;33:127-132.
der RWTH Aachen; 2011. [32] Sellappan N, Palanikumar K. Modified prioritization methodology for
[5] Bornhaeuser M. Reifegradbasierte Werkstattsteuerung. Dissertation, risk priority number in failure mode and effects analysis. IJAST
Stuttgart; 2009. 2013;3:27-36.
[6] Wiendahl HP. Betriebsorganisation für Ingenieure. 8th ed. München: