Вы находитесь на странице: 1из 111

SCUOLA DI INGEGNERIA INDUSTRIALE

E DELL’INFORMAZIONE
Master Degree in Mechanical Engineering
Internal Combustion Engines and Turbomachinery

Noise and Vibrations Measurements and


Analysis in High Power Diesel Generating Sets

Supervisor: Marco TARABINI


Co supervisors: Igor ZANINI, Paolo RIMONDO

Carlo MARAZZA
863449

Academic year 2016/1017

1
2
“Con l’istruzione si sconfigge l’ignoranza che è alle radici della povertà e della fame”
Rita Levi Montalcini

“Le cose di cui nessuno sa e che non lasciano delle tracce, non esistono”
Italo Svevo, “La coscienza di Zeno”

3
Acknowledgements

I consider 2017 as one of the best years of my life: abundant in many


important changes that strengthened my personality and enriched my
experience. I want to thank all those who made all this possible: my uncles
Michele and Edoardo that allowed me and my cousins to live together and
become more independent; Rodolfo and Alberto that shared the apartment
with me and made the stay more than fantastic. A warm thanks to all the staff
of Intergen that supported and helped me: the electrician group, in particular
Fabrizio, with whom I shared the cold control room, Michele and his
kindness, Maurizio and his huge knowledge about engines and, above all,
Paolo that took me under his wing; Davide, for making days less boring;
Igor, without whom this thesis would not exists; and then Elena, Riccardo,
Luca and the heads of the technical office Guglielmo and Franco and all the
others.
A big thanks also to professor Marco Tarabini, for providing me this great
experience, for being always present when needed and above all for his
patience.

Thanks to my parents Fiorenza and Francesco and my sister Evelina, for


being the best I could hope for. Thanks to my aunt Minette and my granny
Giuliana for cheering me up, always.
Last but not least, the greater thanks is reserved for those who make life to
have sense: more than anything else, all my friends.

4
Summary
INDEX OF GRAPHS 9

INDEX OF TABLES 12

ABSTRACT 13

ESTRATTO 14

1. INTRODUCTION 15
1.1. About the company 15
1.1.1. Generating set test 16
1.2. Engines types and description 17
1.2.1. Use 17
1.2.2. Components 17
1.2.3. Tests 18
1.2.4. Technical specifications 19

2. METHOD 21
2.1. Measurements 21
2.1.1. Accelerometers 21
2.1.2. Microphones 21
2.1.3. Frequency domain analysis 22
2.2. Internal combustion Engines 23
2.2.1. Four-stroke Diesel engines 23
2.2.2. Piston induced vibrations 25
2.3. Vibrations 26
2.3.1. Surface vibrations 26
2.3.2. Plane wave propagation: particle velocity and pressure relation 28
2.3.3. Acoustic emission of an infinite plate 29
2.3.4. Acoustic emission of a finite plate 31
2.4. Sound 32
2.4.1. The use of Decibel 32
2.4.2. Frequency 34
2.4.3. Sound emitted by fans 35
2.4.4. Sound emitted by turbochargers 36
2.4.5. Sound emitted by alternators 37
2.4.6. Sound sources comparison in generating sets 38
2.4.7. Sound insulation 38

3. MEASUREMENTS 40
3.1. Description of the instruments 40
3.1.1. Input module NI-9234 40
3.1.2. Microphone BSWA MP201 40
3.1.3. ICP preamplifier BSWA MA201 41
5
3.1.4. Brüel & Kjær DeltaTron Accelerometers Type 4508 B 41
3.2. Measurements procedures 41
3.2.1. Vibrations 42
3.2.2. Noise 46
3.3. MeasLAB 48
3.3.1. MeasLAB Data Acquisition 48
3.3.2. MeasLAB Data Processing 49
3.4. Grid for turbocharger 50

4. RESULTS OF MEASUREMENTS 51
4.1. Noise 51
4.1.1. Background noise: indoor testing room 51
4.1.2. Background noise: outside court 53
4.1.3. Change of load 59
4.1.4. Transient 60
4.2. Vibrations 67
4.2.1. Ambient conditions 67
4.2.2. Statistics 68
4.3. Noise and vibration frequency analysis on engines 72
4.3.1. General frequency domain behaviour of engines 72
4.3.2. Detailed analysis of spectrum 74
4.3.3. Comparison between vibrations spectra 83
4.3.4. Comments on spectra of sound 84
4.4. FRF between noise and vibration 85
4.5. Higher Frequencies 91

5. CONCLUSIONS 96

6. APPENDIX A: MATLAB CODES DESCRIPTION 99


6.1. Intergen_Matlab - Sound 99
6.2. Intergen_Matlab - Vibrations 99
6.3. Functions 100

7. APPENDIX B: ISO STANDARDS 104


7.1. Vibrations – ISO 8528-9 104
7.2. Sound – ISO 8528-10 105

8. BIBLIOGRAPHY 109
8.1. References 109
8.2. Sources for pictures 111

6
Index of figures
Figure 1.2.1 - vibrometer CSI1910 and phonometer HD2010 ................................................... 16
Figure 1.3.1 - Vibration damper ................................................................................................. 17
Figure 1.3.2 - Matlab Engine Scheme ........................................................................................ 18
Figure 1.3.3 - MTU engine based generating set ....................................................................... 18
Figure 1.3.4 - Cummin QSK60-G4 under test ........................................................................... 19
Figure 1.3.5 - Red hot turbine and exhauts gas system of a MTU engine ................................. 19
Figure 2.1.1 - Microphone [m] ................................................................................................... 21
Figure 2.1.3 - Leakage representation and window effect [l] .................................................... 22
Figure 2.1.4 - 16th of November, STFT of engine transient during load change ...................... 23
Figure 2.2.1 - Counterweight and inertial force position ........................................................... 25
Figure 2.2.2 - Scheme of crank system in ICE .......................................................................... 25
Figure 2.2.3 - 1st and 2nd ord. forces .......................................................................................... 26
Figure 2.3.1 - plates internal actions .......................................................................................... 27
Figure 2.3.2 - Sound emission from vibrating plate................................................................... 30
Figure 2.3.3 - Modes of vibrating for finite plate [a] ................................................................. 31
Figure 2.4.1 - Equivalence between sound pressure and sound power and Decibels [b]........... 33
Figure 2.4.2 - Octave and 1/3 octave bands [c].......................................................................... 34
Figure 2.4.3 - A-weighting filter ................................................................................................ 35
Figure 2.4.4 - Turbocharger noise spectrum [d]......................................................................... 37
Figure 2.4.6 - Enclosure structure and insulation strategy [e].................................................... 39
Figure 3.1.1 - Input module NI-9234 [f] .................................................................................... 40
Figure 3.1.2 - Microphone BSWA MP201 [g] .......................................................................... 40
Figure 3.1.3 - ICP preamplifier BSWA MA201 [g] .................................................................. 41
Figure 3.1.4 - Accelerometers .................................................................................................... 41
Figure 3.2.1 - System of refernce ............................................................................................... 42
Figure 3.2.2 - Predisposition of instruments for measurement .................................................. 43
Figure 3.2.3 - Channel position of the sensor: channel 0 accelerometer 4, channel 1
accelerometer 5, channel 2 microphone ..................................................................................... 43
Figure 3.2.4 - Positioning of sensors and picture with reference number (V5, V3) .................. 43
Figure 3.2.5 - Thermal image of the exhaust pipe: Tmax=459,7°C........................................... 45
Figure 3.2.6 - Bubbles of air below the tape .............................................................................. 45
Figure 3.2.7 - Positioning of the accelerometers on a resonant plate ......................................... 45
Figure 3.2.8 - positioning of the microphone for measurement outside a container .................. 47
Figure 3.3.1 - MeasLab first board: sensor selection ................................................................. 48
Figure 3.3.2 - MeasLab boards for sampling frequency, time window, filters, saving .............. 49
Figure 3.4.1 - Grid that should be used for measuring............................................................... 50
Figure 4.1.1 - Logistic of the various container in the external court of the company .............. 53
Figure 4.1.2 - Measurement taken with open doors ................................................................... 54
Figure 4.1.3 - Measurements taken between container .............................................................. 54

7
Figure 4.1.4 - Position of doors and radiator .............................................................................. 54
Figure 4.1.5 - position of the containers in the court.................................................................. 54
Figure 4.1.6 - 24th October, outdoor test with closed door (34 point spaced by 1m) ................ 55
Figure 4.1.7 - 24th October, outdoor test with open door (12 point spaced by 1m) .................. 55
Figure 4.1.8 - Logistic of containers in external court, close to control room ........................... 56
Figure 4.1.9 - Points where the noise was measured .................................................................. 56
Figure 4.2.1 - 1 radiator, 2,4 engine supports, 3 engine stand, 5 alternator body ...................... 68
Figure 4.3.1 - Points of measure for sound ................................................................................ 72
Figure 4.4.1 - 2nd October, accelerometers and microphone position for measure 9 and 10 .... 85
Figure 4.5.1 - Accelerometer along direction y and z on the front plate .................................... 92
Figure 4.5.2 - Accelerometer along direction y and z on the back plate .................................... 92
Figure 4.5.3 - Position of the accelerometer on the engine block .............................................. 93
Figure 4.5.4 - Position of the accelerometer on the radiator ...................................................... 93
Figure 4.5.5 - Microphone position far from plate ..................................................................... 93
Figure 4.5.6 - Microphone close to plate.................................................................................... 93
Figure 4.5.7 - position between turbochargers and intake manifold .......................................... 94
Figure 4.5.8 - Position close to fan on the right of intake system .............................................. 94
Figure 4.5.9 - position to the left of intake system ..................................................................... 94
Figure 7.1.1 - point of measure for vibration according to ISO 8528-9 [j] .............................. 104
Figure 7.2.1 - Positioning for noise measurement [k] .............................................................. 106

8
Index of graphs
Graph 3.2.1 - Velocity and displacement from integration of acceleration along z ................... 46
Graph 3.2.2 - Velocity and displacement from integration of acceleration along x .................. 46
Graph 4.1.1 - Sound pressure and SPL in dB of background noise in testing room .................. 51
Graph 4.1.2 - One third octave band spectrum in dB of background noise in testing room ...... 52
Graph 4.1.3 - One third octave band spectrum in Pa of background noise in testing room ....... 52
Graph 4.1.4 - Sound pressure and SPL of background noise indoor, ventilation fans running . 52
Graph 4.1.5 - One third octave band in Pa of background noise indoor, ventilation fans on .... 53
Graph 4.1.6 - One third octave band in dB of background noise indoor, ventilation fans on .... 53
Graph 4.1.7 - 12th May, 100% load, position 2; rms of SPL=113,8 dB, max SPL=125,7 dB .. 59
Graph 4.1.8 - 12th May, 100% load, position 1; rms of SPL=112,8 dB, max SPL=124,7 dB .. 59
Graph 4.1.9 - 12th May, 50% load, position 2; rms of SPL=112,4 dB, max SPL=124,0 dB .... 59
Graph 4.1.10 - 12th May, 50% load, position 1; rms of SPL=109,8 dB, max SPL=123,1 dB .. 59
Graph 4.1.11 - 12th May, 50% load, position 3; rms of SPL=114,1 dB, max SPL=125,7 dB .. 60
Graph 4.1.12 - 12th May, 50% load, position 3; rms of SPL=112,3 dB, max SPL=125,8 dB .. 60
Graph 4.1.13 - Transient from 100% to 50% of the load ........................................................... 61
Graph 4.1.14 - Transient 100-50% measured by microphone; before cursor, RMS=6,173Pa;
after cursor RMS=4,674Pa (reduction of 2,4dB from 109,7dB to 107,3dB) ............................. 61
Graph 4.1.15 - Transient 100-50% measured by accelerometer; Before the cursor:
RMS=21,54m/s2; after the cursor RMS=23,12m/s2 ................................................................... 62
Graph 4.1.16 - STFT of vibration measured from 100% to 50% load; amplitude scaled with dB
to have a better colours definition .............................................................................................. 62
Graph 4.1.17 - STFT of noise measured from 100% to 50% load;............................................ 63
Graph 4.1.18 - Transient 0-100% measured by microphone; before cursor RMS=6,7Pa
(110,5dB), after 4,52Pa (107,8dB) ............................................................................................. 63
Graph 4.1.19 - STFT of vibration measured from 0% to 100% load; ........................................ 64
Graph 4.1.20 - STFT of noise measured from 0% to 100% load;.............................................. 64
Graph 4.1.21 - Transient 0-100% measured by accelerometer; before cursor RMS=20,43m/s2;
after cursor RMS=25,9m/s2 ....................................................................................................... 64
Graph 4.1.22 - Vibrations while the engine is stopping ............................................................. 65
Graph 4.1.23 - Noise while the engine is stopping .................................................................... 65
Graph 4.1.24 - STFT of vibrations while the engine is stopping ............................................... 65
Graph 4.1.25 - Focus on STFT of vibrations in 0-1000Hz range with stopping engine ............ 66
Graph 4.1.26 - Focus on STFT of noise in 0-1000Hz range with stopping engine ................... 66
Graph 4.2.1 - Example of vibration measurement where full scale at 500 m/s2 is reached ....... 69
Graph 4.2.2 - Number of samples per each vibration range, along x, y and z (132 samples) .... 70
Graph 4.2.3 - Gaussian PDF obtained for 132 samples along x, y and z ................................... 70
Graph 4.2.4 - Gaussian PDF obtained for 49 samples along y .................................................. 70
Graph 4.2.5 - Number of samples per each range of vibrations, along x (40 samples) ............. 70
Graph 4.2.6 - Gaussian PDF obtained for 40 samples along x .................................................. 70

9
Graph 4.2.7 - Number of samples per each range of vibrations, along y (49 samples).............. 70
Graph 4.2.8 - Number of samples per each range of vibrations, along z (43 samples) .............. 71
Graph 4.2.9 - PDF obtained for 43 samples along z .................................................................. 71
Graph 4.3.1 - 29th May, meas. 2, 4 and 6 .................................................................................. 72
Graph 4.3.2 - 2nd October, meas 3, 9 and 11 ............................................................................. 73
Graph 4.3.3 - 6th June, meas. 1 and 3 ........................................................................................ 73
Graph 4.3.4 - 25th September meas. 2, 4 and 5.......................................................................... 73
Graph 4.3.5 - 28th September, meas. 1, 5 and 8......................................................................... 73
Graph 4.3.6 - 28th September, sound 16 .................................................................................... 74
Graph 4.3.7 - 28th September, sound 16, zoom on lower frequencies ....................................... 75
Graph 4.3.8 - 28th September, vibration 9 4z ............................................................................ 75
Graph 4.3.9 - 28th September, sound 17 .................................................................................... 76
Graph 4.3.10 - 12th June, sound in point 1 ................................................................................ 77
Graph 4.3.11 - 12th June, overall vibration in point 1 ............................................................... 77
Graph 4.3.12 - 28th September, sound 16 .................................................................................. 77
Graph 4.3.13 - 28th September, vibration 16 5y, far from radiator............................................ 78
Graph 4.3.14 - 28th September, vibration 1 5y, close to radiator .............................................. 78
Graph 4.3.15 - 2nd October, vibration 9 5y, far from radiator ................................................... 79
Graph 4.3.16 - 2nd October, vibration 1 4y, on radiator ............................................................ 79
Graph 4.3.17 - 2nd October, sound 1 ......................................................................................... 79
Graph 4.3.18 - 25th September, sound 13 .................................................................................. 80
Graph 4.3.19 - 25th September, vibration 1 4y, on radiator ....................................................... 80
Graph 4.3.20 - 25th September, vibration 11 5y, far from radiator............................................ 80
Graph 4.3.21 - 2nd October, noise from fans on the roof .......................................................... 81
Graph 4.3.22 - 3rd May, fan noise ............................................................................................. 81
Graph 4.3.23 - 24th October, open container 12 ........................................................................ 82
Graph 4.3.24 - Comparison engines 29th May (sound 3 right) /12 June (sound 100% 2) ......... 82
Graph 4.3.25 - 2nd October, vibration 6 .................................................................................... 83
Graph 4.3.26 - 2nd October, vibration 4 .................................................................................... 83
Graph 4.3.27 - 2nd October, sound 6 ......................................................................................... 84
Graph 4.4.1 - Noise, whole spectrum ......................................................................................... 86
Graph 4.4.2 - Overall vibration obtained as 𝑉 = 𝑥2 + 𝑦2 + 𝑧2, whole spectrum..................... 86
Graph 4.4.3 - Normalized FRF gives clearer spectrum: peaks in 750-1250Hz range disappear,
since they are present as sound, but not as vibration. Order of magnitude is the same .............. 87
Graph 4.4.4 - FRF: most of lines are linked to base excitation frequency; in 750-1250Hz range
there is overlapping. It will be shown later that it is turbocharger operation range ................... 87
Graph 4.4.5 - Focus on range 0-1500Hz for sound. Frequency peaks from fan are evident ...... 88
Graph 4.4.6 - Focus on range 0-1500Hz for sound .................................................................... 88
Graph 4.4.7 - Normalized FRF. A kind of periodicity pf peaks seems to be present ................. 88
Graph 4.4.8 - Inverse of the normalized FRF (vibrations/sound) .............................................. 89
Graph 4.4.9 - Fan components are present, their link with vibration is small but still present .. 89
10
Graph 4.4.10 - 2nd October, sound 9, 118 dB ........................................................................... 90
Graph 4.4.11 - 2nd October, sound 4, 119,2 dB ........................................................................ 90
Graph 4.4.12 - Sound 4 close to radiator (black), ventilation fans (red), alternator fan (blue) .. 90
Graph 4.4.13 - Sound 9 far from radiator (black), ventilation fans (red), alternator fan (blue) . 90
Graph 4.4.14 - 2nd October, sound 9 filtered by aerodynamic noise......................................... 91
Graph 4.4.15 - 2nd October, sound 4 filtered by aerodynamic noise......................................... 91
Graph 4.5.1 - Secondary vibration amplitude along y ............................................................... 92
Graph 4.5.2 - High vibration amplitude along z ........................................................................ 92
Graph 4.5.3 - Secondary vibration amplitude along x ............................................................... 92
Graph 4.5.4 - Very low vibration amplitude along z ................................................................. 92
Graph 4.5.5 - High vibration amplitude along x ........................................................................ 92
Graph 4.5.6 - Very low vibration amplitude along y ................................................................. 92
Graph 4.5.7 - Negligible vibration along z ................................................................................ 93
Graph 4.5.8 - Small vibration along z ........................................................................................ 93
Graph 4.5.9 - Sound measured far from plate ............................................................................ 93
Graph 4.5.10 - Sound measured close to plate ........................................................................... 93
Graph 4.5.11 - Noise spectrum on the left of turbochargers ...................................................... 94
Graph 4.5.12 - Noise spectrum between turbochargers ............................................................. 94
Graph 4.5.13 - Noise spectrum between intake and fan............................................................. 94

11
Index of tables
Table 1.3.1 - Some technical data of an engine.......................................................................... 20
Table 2.4.1 - Noise origin in turbocharger [d]............................................................................ 36
Table 3.1.1 - Microphone BSWA MP201 datasheet [g] ............................................................ 40
Table 3.1.2 - ICP preamplifier BSWA MA201 datasheet [g] .................................................... 41
Table 3.1.3 - Brüel & Kjær DeltaTron Accelerometers Type 4508 B datasheet........................ 41
Table 4.1.1 - SPL measured around the container with open and closed doors ......................... 55
Table 4.1.2 - Measured background noise with and without fans and noise measured with
engine on. A, B and C are used for logarithmic average calculation (ISO 8528-10) ................. 57
Table 4.1.3 - In green are evidenced the lower values measured on the right side of the
container; in blue there are the higher one on the left side ......................................................... 58
Table 4.1.4 - Noise comparison between Intergen data and those taken on manufacturer place;
values 1-8 were taken at the base: in second column are kept the same of Intergen, in the third
they are corrected by 2dB ........................................................................................................... 58
Table 4.2.1 - Vibration (rms) measured in point 1, 2, 3, 4 and 5 for various generating sets and
statistical results (mean, standard deviation, kurtosis, skewness) .............................................. 68
Table 4.2.2 - Statistical results from vibrations taken on generating sets in different points ..... 71
Table 7.1.1 - Reference level of vibration [j] ........................................................................... 105
Table 7.2.1 - Corrective coefficient K1 depends on difference with background noise [k] ..... 107

12
Abstract

This work aims to analyse the noise and vibration generated by different generating sets
manufactured by Intergen. The different systems were analysed from a general point of
view, with the aim of understanding if there are similarities or general situations.

The macro-areas of interest were mainly two: the investigation of vibration and noise
alone at first, and then studying the effect of one on the other; the second is the
observation of background noise in measurement when performed either inside or
outside the buildings.

Seven engines were tested: these came from different manufacturer and had different
dimension, power and geometries. They also run at different speed: 1500rpm or 1800
rpm, depending on the country of destination. The tests were performed both inside a
soundproof testing room and in the court outside the buildings. The positioning of
microphones and accelerometers were chosen having as basis the standard ISO 8528-9
and 8528-10. Nevertheless, depending case by case, many other point were selected, in
order to point out resonance effects and obtain a more generalized analysis.
Data analysis was performed both in time and frequency domain: in time domain it was
possible to obtain a statistical distribution of vibrations and noise level; with the
investigation on the spectra they were detected the main responsible for sound emissions
by linking vibrations and sound profiles.
The final results depicted how the engines emits sound and vibration in the same range
of frequency (depending on their rotational speed). Main sources of noise emissions were
detected in the diesel engine and the radiator’s fan, if present. Other responsible of
secondary importance were found in the turbocharger, in the intake air system and
alternator.

Concerning the background noise, an important role inside is the one given by ventilation
fans; outside the effect of surrounding obstacles is able to change the output of sound
measurement up to 3dB.

13
Estratto

Questo lavoro vuole analizzare il rumore e le vibrazioni emesse da differenti gruppi


elettrogeni assemblati da Intergen. I differenti sistemi sono stati analizzati da un punto di
vista generico, con lo scopo di capire se ci sono similarità o situazioni generali.

Le macro-aree di interesse sono state principalmente due: l’investigazione di vibrazioni


e rumore prima da sole e poi studiando gli effetti delle une sull’altro; la seconda è stata
l’osservazione del rumore di sottofondo nelle misurazioni ottenute sia dentro che fuori
agli edifici.

Sono stati testati sette motori: questi provengono da diversi produttori e hanno differenti
dimensioni, potenze e geometrie. Girano anche a differenti velocità: 1500rpm e 1800rpm,
a seconda della nazione di destinazione. I test sono stati fatti sia all’interno di una sala
prove insonorizzata che nel cortile all’esterno dell’edificio. Il posizionamento del
microfono e degli accelerometri è stato scelto avendo come base le norme ISO 8528-9 e
8528-10. Tuttavia, caso per caso, molti altri punti sono stati selezionati, in modo da
evidenziare gli effetti di risonanza e ottenere un’analisi più generalizzata.
L’analisi dei dati è stata fatta sia nel dominio del tempo che in quello delle frequenze:
nel dominio del tempo è stato possibile ottenere una distribuzione statistica di vibrazioni
e livello di rumore; con l’investigazione sugli spettri sono stati evidenziati i principali
responsabili delle emissioni sonore, legandole ai profili delle vibrazioni.
I risultati finali hanno mostrato come i motori emettano suono e vibrazioni nello stesso
range di frequenze (dipendendo dalle loro velocità di rotazione). Le principali fonti di
emissioni di rumore sono state individuate nel motore diesel e nella ventola del radiatore.
Altri responsabili di secondaria importanza sono il turbocompressore, il sistema di
aspirazione e l’alternatore.

Per quanto riguarda il rumore di sottofondo, un ruolo importante al chiuso è quello dato
dalle pale di ventilazione; all’esterno l’effetto degli ostacoli attorno è in grado di
cambiare l’output delle misurazioni sonore fino a 3dB.

14
1. Introduction

The problem of sound and vibration emission from big generating sets is quite complex,
since there are many sources of excitation. Vibration can be generated not only by the
intrinsic properties of internal combustion engines, that is, combustion and inertial
components. Other sources are the unbalance of the many rotating parts: crankshaft,
camshaft, gears, turbocharger, radiator fan, alternator; then there are the continuous hits
of the cams on the valves, of the valve on their siege, the grinding of gears; the noise
from the huge masses of air in movement: at the intake mufflers, inside turbocharger, the
cooling air though the radiator, the cooling air of the alternator, and so on. Moreover,
vibration generated by a relatively small component as it could be the turbocharger, can
be amplified by nearby structure that goes in resonance. Each single element which is
vibrating is responsible for noise emission on a wide range of frequencies and
amplitudes.

Another issue, more linked to the industrial world than to the academic one, is that
manufacturers are very jealous of their products, and information that are just a little bit
out of the standard datasheet given at commercial level, are very difficult to find.
The complexity and the width of the topic is such that the literature concerning diesel
engine vibrations, fans noise, generating set insulation, turbocharger properties and so
on, is very rich. Despite the problem has been treated from many point of views, the great
part of them are very specific and focused on a particular subject. The followed approach
is, on the contrary, to look at the whole system with a more general view.

1.1. About the company


Lomagna, where the company is located, is 45 minutes far from Milan. IML is a private
firm, it has more than 150 employees, split in two divisions: the historical one, called
IML (Industrie Meccaniche Lombarde) Motori, which is the concessionary of Deutz
engines here in Italy. They assemble diesel engines of medium power for trucks, tractor,
small generating sets, etc. Recently Deutz bought them. The other part, Intergen,
assemble huge plants for electricity production, required by customers coming from all
over the world. The huge diesel engines come already assembled from Germany (MWM,
MTU), United Kingdom (Cummins), USA (G.E.), etc. Usually they have more than
1MW of power, up to 5MW. They are coupled with alternators (from Marelli Motori),
and then they are provided with the control system, which is created ad hoc. As a matter
of fact, everything is done accordingly to the customer desires, so the level of
15
personalization is very high. These systems are sold all around the world, mainly they go
to isolated places, like oil platforms in the middle of the sea or hospitals in Africa.

1.1.1. Generating set test


Each plant is tested inside the company, which has two large soundproofed rooms, with
systems for heat rejection (a large fraction of the power goes into heat, but the room is
isolated, so temperature rises very quickly inside), combustion products filtering and
rejection systems, and water cooling systems for the larger engines. During the tests, the
load is provided to the generator with a resistive load bank, situated outside the building.

After each test performed within the company, a representative of the customer and an
inspector come and verify that everything is in order and respectful of international
standards. Among all the ISO standards which must be respected, there are also the ones
regarding sound and vibrations. In this field, the reference ones are ISO 8528-9 [1] and
ISO 8528-10 [2]. Those standards specify the formulas to use for computing the final
value of vibration and sound (to be compared with the allowed limits), the conditions
and the points where to effectuate the measurement and the distances.

The instruments used by the company are a


vibrometer CSI 1910 and a phonometer
HD2010. In order to accomplish with the
current ISO standard, instruments are
calibrated every three years, and are adequate
for the internal use of the company, but not
enough for a deeper analysis. The analysed
Figure 1.1.1 - vibrometer CSI1910 and phonometer
data can be downloaded on a Personal
HD2010 computer running windows XP, but it is not
possible to download the time history of the measured signal.

The problems they often have are of two natures. The first is understanding the frequency
range, both of noise and vibrations, in which the generating sets emit. In particular, they
were wondering whether it was possible to associate noise to the correspondent vibration
source. This is required to evaluate if it is possible to improve the soundproof containers
and the damping system on the basement of the engine, mainly to see whether it is
possible to reduce the costs.

The second issue regards the measurement performed outside the building, in the open
space: they wanted to know which, between the various obstacles present, were the more
influential: the shed, the trees or the containers.

16
1.2. Engines types and description
1.2.1. Use
In the months spent in the company, many engines passed through, all different one from
the other in terms of manufacturer, powers and dimensions. They can be easily divided
into three main categories: small, medium and large engines. They are all built in the
same way and necessary for the same use: emergency electrogene groups. They are
adopted mainly on sea platform for oil industry, but also in hospitals placed in critical
regions or inside huge commercial ships. 90% of these engines goes abroad, in regions
as Middle East or Sud Asia.

1.2.2. Components
All the emergency groups tested are constituted by a Diesel engine (MTU, MWM,
Cummins, G.E., …), linked to an electrical generator (mainly from Marelli Motori). For
small-medium engines, there is also a huge radiator mounted on the front: it is an
important source of fluid dynamic noise, due to the big fans rotating at high speed. For
more powerful engines the cooling system sends the refrigerant to a radiator placed
outside the testing room or on the roof of the cooldntainer, so there is not this component
of noise.

All the groups are mounted on a base made of strong soldered


H bars, which lay on different damper as the one in the figure.
They avoid that the great quantity of vibrations are
transmitted to the ground. Due to the huge dimensions,
vibrations are anyway felt, but they are very low.
Figure 1.2.1 - Vibration damper
The turbocharger can be placed in different ways: the most
used solution is having one compressor and one turbine per each bench of the engine. In
the smaller plants, they are placed aside, and it is quite easy to measure their sound by
using a grid. Larger engines have the turbocharger placed on the top and it is not easy to
be reached both with the microphone and the accelerometers (also due to high
temperatures which could damage the instruments).

17
A common scheme of the engine is this one, obtained with MATLAB and reproducing
the structure of most engine, used later to clarify where the measurements were taken
and at which level of amplitude.

Figure 1.2.2 - Matlab Engine Scheme Figure 1.2.3 - MTU engine based generating set

The number of cylinder varies between 12 and 20, depending on model, dimension,
power, but most had 16 cylinders. Being engines to be coupled with asynchronous
generators, they must produce power at the same frequency of the grid, which means that
the velocity of rotation is 1500 rpm for those that go in countries with a grid at 50Hz, or
1800 rpm for those going in counties where it is 60Hz. This is an important factor in
vibration and noise entity.

1.2.3. Tests
Depending on the dimensions, the engines run in the testing room with (small) or without
(medium) the radiator, or they are tested outside (large) inside a container in the court of
the company.

The final test of the engine is performed in presence of the customers and of inspectors
(who verify that all the norms are respected). The phases of the test consist in running
the engine at 0% load for half an hour, then 25% for other 30 minutes, and then depending
on the requests at 50% and 75% for one hour or half, finally four hours at 100% load and
one hour at 110%. During the four hours at 100% the measurements of sound and
vibrations are performed according to standards ISO 8528-9 and 8528-10.

In the final part of the test, the engine is subjected to strong and impulsive changes of
load in order to see how it reacts to sudden change that could happen in real applications.

In the period at the beginning of summer, the smaller and medium engines were tested,
with power between 0,8 and 1,5 MW. As previously mentioned, the radiator is part of
these engines and the test is done inside the sound proof testing room: walls and floor
18
are such that no reverberation should exists, simulating the open field condition. The
measures taken here are the one that are closer to reality.

An example of this category is the Cummins model QSK60-G4, tested at the end of May
for TANAP (Trans Anatolian Natural Gas Pipeline Project):

Figure 1.2.4 - Cummin QSK60-G4 under test


The picture does not take the entire group since it cuts the generator, but it gives an idea
of the dimensions and of the disposition of the parts (such as the turbocharger placed in
a not easy to reach position).

1.2.4. Technical specifications


In the following pages there are some technical
information about the engine, kindly given by
the technical office and contained in the
datasheet of the engines. Datasheets are very
detailed and contain information concerning
losses of power due to friction, heat exchanged
with the ambient (both cooling system and
exhaust gases), but cannot be made public. Here
some generic information, able to give an order Figure 1.2.5 - Red hot turbine and exhauts gas
system of a MTU engine
of comparison.

19
Table 1.2.1 - Some technical data of an engine

These are not information useful for the purpose of this thesis, but they are in any case
interesting. The high power involved are felt mainly during the phases of data acquisition
in terms of heat rejected. Even if outside the testing room there are 10°C, inside it is
anyway very hot.

20
2. Method

The number of subjects touched requires for a brief summary of basic knowledge at least
for the main four topics: measurement, internal combustion engines, vibration and sound.

2.1.Measurements
2.1.1. Accelerometers
Accelerometers principle of works consist in monitoring a little mass displacement. The
acceleration imposed, makes the mass inside the sensor to move consequently, oscillating
around its position of equilibrium. This movement can be registered in different way:

- piezo accelerometers exploit the piezoelectric properties of some ceramic


materials. When the element is hit, a difference of voltage is generated on its two
opposite sides and sent to the acquisition board. With proper calibration, voltage
and acceleration are associated univocally. Those accelerometers are highly
linear and are used in the frequency range going from 2Hz up to 5000Hz;
- LVDT (Linear variable Differential Transformer) accelerometers works thanks to
magnetic properties of the mass, which must be in ferromagnetic material. The
element is suspended on elastic support which allows its movement back and
forth inside windings. The variation of magnetic field generates a current which
is then associated to the vibration. They can be used in the range 1-10000 Hz.
- Capacitive accelerometers, also known as MEMS (Micro Electro-Mechanical
System), exploit the variation of capacity of a plane condenser. The mass is
mounted on a comb of thin lamellae that works both as condenser and elastic
elements. The change in voltage is registered and associated to acceleration.
These accelerometers have to be linked to a source of voltage in order to give an
output. The accelerometers used for measurements are of capacitive type.

2.1.2. Microphones
Microphones are accelerometers able to measure the
oscillation of air pressure. The most used ones have a
capacitor placed between the solid body of the sensor
and a membrane. The membrane feel the variation of
pressure by contracting or expanding: those
infinitesimal displacements are registered by the Figure 2.1.1 - Microphone [m]
21
condenser which sends the voltage output to the acquisition board. As MEMS, a
microphone can not work without a voltage source. Older microphones were magnetic:
membrane lead to the displacement of a magnet inside a coil. Electrical current is
generated and associated with sound.

2.1.3. Frequency domain analysis


Fourier Transform
The spectrum in frequency domain is obtained by applying the Fourier Transform:
+∞ +∞
−2𝜋𝑖𝑓𝑡
𝐴0
𝐹(𝑓) = ∫ 𝑓(𝑡)𝑒 𝑑𝑡 ; 𝐹(𝑓) = + ∑ 𝐴𝑛 cos(2𝜋𝑓𝑛 𝑡 − 𝜑𝑛 ) (2.1.1)
−∞ 2
𝑛=1

The signal in time domain is made by the superposition of infinite sinusoidal signals:
with the Fourier it is possible to separate all these harmonic waves.

Leakage
If the time window for which the signal is sampled is not long enough, there is the risk
of getting leakage error: it occurs when the time interval is not multiple of the one of the
signal, which is then cut and just as a fraction of it is taken. This has an effect on
frequency resolution: the peaks contained in the spectrum are the ones that are multiples
of a certain elementary frequency, defined as the inverse of the time window for which
the signal is sampled. If it is 10s, the frequency resolution will be 0,1 Hz. It means that
all the frequency multiples of 0,1Hz will be present in the spectrum. The amplitudes of
those which are not present will be split in the surrounding ones. The way in which this
division occurs depends on the window used. There are many types of windows, the two
main ones are the rectangular and the Hanning windows.

Figure 2.1.2 - Leakage representation and window effect [l]


Power spectrum
It is defined as:
𝑛=𝑁 2
1 2𝜋𝑖𝑘𝑛
𝑃𝑆 = [ ∑ 𝑓(𝑛)𝑒 − 𝑁 ] (2.1.2)
𝑁
𝑛=0

22
Being the square of the modulus of the discrete Fourier transfer function (units of
measure are squared), it is able to separate the interesting peaks from the noise that can
emerge depending on the time window used for the acquisition.
STFT
It may happen that the sampled signal changes in time not only in terms of amplitude,
but also the frequency varies. It is possible to see evolution of frequency in time, but
doing so in a standard 2D graph it is not possible to visualize the amplitudes. To solve
this problem, the STFT (Short Time Fourier Transform) is used: it allows to obtain a
colourmap that has on horizontal axis the time, on the vertical the frequencies and by
mean of a scale of colour, the amplitudes are visualised as follow:

Figure 2.1.3 - 16th of November, STFT of engine transient during load change

[19] IC engine vibration signals have two essential features. First, there are many
frequency response components generated by the excitation. Second, some excitations
are close in the time domain, even simultaneous. These may cause response components
of the same frequency. The STFT can be used to identify these excitation events.

2.2. Internal combustion Engines


2.2.1. Four-stroke Diesel engines
[4] [5] In Four-Stroke engines each piston requires four strokes to complete the cycle
that lead to power production. The four phases are:

- Intake stroke: the piston moves from Top Death Centre (TDC) to the Bottom
Death Centre (BDC), sucking air inside the combustion chamber through the
opened valves, which closes few instants after the BDC is reached;
- Compression stroke: the piston returns at the TDC: both valves are closed, and
pressure is increased. During the last part to cover, fuel is ignited by the injector.

23
At the end of this phase combustion starts: in spark ignition system, a spark
ignites the mixture, while in Diesel engine it is enough the high pressure reached.
- Power stroke: the combustion of fuel lead to a sudden increase in temperature
and a consequent increase in pressure which push the piston down. It is the stroke
in which power is given to the shaft. When the piston is close to the BDC, the
exhaust valve opens, and the exhaust gas start to exit.
- Exhaust stroke: the piston returns to the TDC forcing out the burned mixture and
preparing for a new cycle. The exhaust valve closes few instants after the intake
valve has opened.
Four-stroke engines have a power stroke each 2 round of the crankshaft; Diesel engines
autoignition of the mixture is associated with very high pressures and noise, which
increase even more after combustion is started: compression ratio (ratio between
maximum and minimum volume of the combustion chamber delimited by the piston) is
in the range 12-24. For these reasons, the vibrations are stronger in CI engines than in SI
ones. Apart from inertial contributions, the effect of combustion is strong. Inertial effect
will be present once per round (12,5 Hz for engines rotating at 1500rpm, 15Hz for ones
running at 1800rpm), while combustion vibration will be present once per two rounds
(15 Hz for engines rotating at 1500rpm, 30Hz for ones running at 1800rpm). If an engine
has more than one cylinder, the firing order has to be such that the behaviour of vibration
is balanced by piston moving in opposite directions, but also the intake and the exhaust
phases must not have interferences between cylinders.

Increasing the dimensions means increasing the volume and power, but also the moving
masses. For this reason, big engines must run at low speeds, or the stresses on rod and
crankshaft would reach unbearable levels. Since huge engines are used for power
production, they have to be efficient, but also the fuel cost has not to be excessive. For
this reason, Diesel engines are used: their combustion process is slower than the one of
SI, consequently they are more suitable for these applications. Moreover, having high
compression ratio, they are more efficient than SI engines. Finally, the torque provided
is higher because they run at low speeds, and are easy to couple with generators.

The use of turbocharger is required in Diesel engines to increase the efficiency: energy
contained in the exhaust gases is exploited with a turbine which drives a compressor: it
increases fresh air pressure in order to reach higher compression ratio and efficiency. The
velocity of rotation changes depending on use, dimensions and load. For smaller engines
it can reach 110000rpm, for bigger ones it runs at 40-50000 rpm.

24
2.2.2. Piston induced vibrations
[6] With reference figure 2.2.2, schematizing a mono-cylinder engine:

Figure 2.2.2 - Scheme of crank system in ICE Figure 2.2.1 - Counterweight and inertial force
position
- 𝑚: mass of the piston;
- 𝛼: angle of rotation (𝛼̇ constant);
- 𝜀𝐷 : eccentricity;
- 𝑇𝑚 : torque;
- 𝑚𝐷 𝜀𝐷 𝜔𝐷2 : inertial force of the counterweight;
- 𝑅: constraint forces;
- 𝐹𝑖𝑛 : inertial force of the piston;
Considering the system equilibrium, the crank gives an internal force, plus there are the
inertial forces that loads the supports of the engine. Vertical and horizontal equilibrium:

𝑅𝑉𝐿 + 𝑅𝑉𝑅 = 𝐹𝑤𝑒𝑖𝑔ℎ𝑡 + 𝐹𝑖𝑛 + 𝑚𝐷 𝜀𝐷 𝜔𝐷2 cos 𝛼


{ (2.2.1)
𝑅𝑂𝐿 = 𝑚𝐷 𝜀𝐷 𝜔𝐷2 sin 𝛼

Being 𝜆 the ratio between the crank and the rod, equal to 0.2-0.3, acceleration is
approximated with the two terms

𝑐̈ ≅ −𝑟𝜔𝐷2 cos 𝛼 − 𝑟𝜆𝜔𝐷2 cos 2𝛼 (2.2.2)

that are the first and second harmonic component. System 2.2.1 becomes:

̅ (𝑟𝜔𝐷2 cos 𝛼) − 𝑚
𝑅𝑉𝐿 + 𝑅𝑉𝑅 = 𝐹𝑤𝑒𝑖𝑔ℎ𝑡 − 𝑚 ̅ (𝑟𝜆𝜔𝐷2 cos 2𝛼) + 𝑚𝐷 𝜀𝐷 𝜔𝐷2 cos 𝛼
{ (2.2.3)
𝑅𝑂𝐿 = 𝑚𝐷 𝜀𝐷 𝜔𝐷2 sin 𝛼

The counterweight is dimensioned to reduce the force generated by piston movement.


They are called forces of first and second order:

- 𝑚𝐷 𝜀𝐷 = 𝑚
̅ 𝑟 (counterweight equal to first order force):
𝑅𝑉𝐿 + 𝑅𝑉𝑅 = 𝐹𝑤𝑒𝑖𝑔ℎ𝑡 − 𝑚 ̅ (𝑟𝜆𝜔𝐷2 cos 2𝛼)
{ (2.2.4)
𝑅𝑂𝐿 = 𝑚̅ 𝑟𝜔𝐷2 sin 𝛼

25
Vibration transmitted to the support are associated with the lower forces of
inertia, due to the contribution of 𝜆 which reduces the entity and the frequency of
rotation will be higher, being associated with the second harmonic. Anyway, on
the external support, damping systems are required: the stiffer are the spring, the
higher is the reduction. With the counterweight, it is possible to reduce only the
vertical vibrations, while the horizontal ones are still present.
̅𝑟
𝑚
- 𝑚𝐷 𝜀𝐷 = (counterweight equal to one half first order force):
2
𝑚
̅𝑟 2
𝑅𝑉𝐿 + 𝑅𝑉𝑅 = 𝐹𝑤𝑒𝑖𝑔ℎ𝑡 − 𝑚 ̅ (𝑟𝜆𝜔𝐷2 cos 2𝛼) − 𝜔 cos 𝛼
{ 2 𝐷 (2.2.5)
𝑚̅𝑟 2
𝑅𝑂𝐿 = 𝜔 sin 𝛼
2 𝐷
In this case also the horizontal force is reduced, but there is a force equal to one
half of the original one, rotating counter clockwise, moreover there are both
harmonics.

To eliminate almost completely these components of excitation usually


a counterrotating shaft is added, so that the counterrotating forces of
inertia are balanced. It is possible to represent the first and second
harmonic component, one as function of cos 𝛼, the other as function of
cos 2𝛼. In systems with more than one cylinder, like a 4 cylinders
engine, 2 will have the crank positioned at 𝛼, the other two at 𝛼 + 𝜋/2.
Figure 2.2.3 - 1st
This will help to reduce the vibrations, since each piston moving up, is and 2nd ord. forces
eliminating the contribution of another piston moving down, and the same is true for the
position of the counterweight. Increasing the number of cylinders, also these phenomena
of mutual annihilation of are increased.
Notice that if the number of pistons in the same line is odd, then there will always be a
component not well balanced; moreover, increasing the length of the engine (to make
space for more cylinders), torsional vibrations will increase. [17] According to an
analysis on an analysis performed Wärtsilä Diesel engines for power production,
common values of vibration are in order of magnitude of 10-30 m/s2 (acceleration) in the
frequency range of 2Hz to 1000 Hz, depending on where the measurement is taken.

2.3. Vibrations
2.3.1. Surface vibrations
[7] The thickness of thin plane surfaces has a value different order of magnitude lower
than the other two dimensions.
26
Thin plates vibrations behaviour is calculated on the basis of five main hypothesis:
- 1. Shear stresses are not considered.

- 2. Displacement of w in z direction are small with respect to thickness.


- 3. Points on middle plane move along z, the other move also along x and y for rotation.

- 4. Normal stresses 𝜎𝑥 and 𝜎𝑦 cancel each other on the middle plane and along the
thickness they are proportional to distance z from middle plane.

- 5. Stress component 𝜎𝑧 is negligible (true for simple flection, good approximation if


normal loads are presents, thanks to thin plate hypothesis and small deformations w).
Considering a square thin plate, internal stresses are linked with internal bending
moments. These can be explained considering an elastic beam which is subjected just to
its weight and fixed at its extremities. The beam will curve under its weight and the point
of maximum displacement will be in the centre. If another beam is put close to it and
another and so on, they will follow the same profile. If they are soldered together, a plane
plate is obtained, shaped like a U. If also the other two edges are fixed, then the same
things can be said for the beams along the direction perpendicular to previous ones. But
since they are all linked one to the other, a beam which is inflected will exchange with
the next one a torque. This provokes its rotation, overlapped to the already existing
flection. Since this happens in all infinitesimal points
of the plate, those moments cause the state of stress
which lead to the displacement of the plate (maximum
in the centre, minimum on the fixed edges).

Internal actions on an infinitesimal element of a flexed


plate, are flexing moment 𝑚𝑥 and 𝑚𝑦 , torque 𝑚𝑥𝑦 and
distributed load 𝑝(𝑥, 𝑦), plus the shear stresses.

The bending and torsional moment can be obtained by


integrating the stress functions. Considering the
equilibrium of rotation around x and y and the vertical Figure 2.3.1 - plates internal actions
equilibrium, the shear stress and distributed load can be obtained. It is possible to get the
equations that describe the vibrating plane velocity and displacement:
𝑛𝜋 𝑛𝜋
𝜆1,2 = ±𝛾1 , 𝛾1 = √1 + 𝑘; 𝜆3,4 = ±𝑗𝛾2 , 𝛾2 = √𝑘 − 1; (2.3.1)
𝑎 𝑎
Wave length 𝜆 depends on the wave number k. Plate’s natural frequency and modal
shapes 𝜓𝑛,𝑚 (𝑥, 𝑦) are:

27
2 𝐵
𝜔𝑛,𝑚 = [(𝑘𝑥𝑛 )2 + (𝑘𝑦𝑚 ) ] √ ; 𝜓𝑛,𝑚 (𝑥, 𝑦) = sin(𝑘𝑥𝑛 𝑥) sin(𝑘𝑦𝑚 𝑦) (2.3.2)
𝜌𝑠

Where the wave numbers kxn and kym are function of n and m:
2𝜋 𝑛𝜋 2𝜋 𝑚𝜋
𝑘𝑥𝑛 = = ; 𝑘𝑦𝑚 = = ; (2.3.3)
𝜆𝑥𝑛 𝑎 𝜆𝑦𝑚 𝑏

It is possible to assume that flexional wave in the plane will propagate only along x
direction, so that 𝑘𝑦 = 0 and with only one wave number 𝑘𝑥 = 𝑘𝐵 . Under this
assumption deformation and velocity are defined as:

𝑤 = 𝑤(𝑥, 𝑡) = 𝜓(𝑥)𝐺(𝑡) = 𝐺̅ 𝑒 𝑗(𝜔𝑡−𝑘𝑏𝑥) (2.3.4)


𝜕𝑤
𝑣= = 𝑗𝜔𝐺̅ 𝑒 𝑗(𝜔𝑡−𝑘𝐵 𝑥) (2.3.5)
𝜕𝑡
𝑣 is the velocity of vibration normal to the plate, moving along x. It is like a wave
propagating along a string: if we put together an infinite number of infinitesimal string,
one equal to the other (same length, material…) so that they vibrate in phase, with the
same kind of wave and no time delay, this is the equation describing its behaviour.

2.3.2. Plane wave propagation: particle velocity and pressure relation


The simplest case is plane wave propagation: it can be said that at a certain time instant,
the pressure is constant in each plane perpendicular to x:

𝑝(𝑥, 𝑡) = 𝐺1 (𝑐𝑡 − 𝑥) + 𝐺2 (𝑐𝑡 + 𝑥) (2.3.6)

x is propagation direction normal to the wave, p the pressure, 𝐺1 and 𝐺2 are generic
functions depending on boundary conditions: 𝐺1 is a progressive wave travelling along
positive direction of x, while 𝐺2 is a regressive wave, travelling in the opposite direction.
An observer travelling at the same speed of the wave 𝐺1 (sound velocity), will see the
pressure as not dependant on time. In the particular case in which 𝐺1 is sinusoidal
(progressive harmonic wave), the solution to the equation (2.3.6) is:

𝑝(𝑥, 𝑡) = 𝑃̅ cos[𝑘(𝑐𝑡 − 𝑥)] (2.3.7)

𝑃̅ is the peak pressure, k is a constant linked to the period (or frequency) of the wave,
that has the following speed, frequency and pulsation:
𝜆 𝑐 𝑐
𝑐= ; 𝑓= ; 𝜔 = 2𝜋 ; (2.3.8)
𝑇 𝜆 𝜆

28
From (2.3.7) it emerges that the pressure is an harmonic function both of time (period T)
and space (period 𝜆). k is a kind of spatial frequency, called wave number. Evidencing
the periodicity of pressure in space, it is possible to get k:
2𝜋 𝑟𝑎𝑑
𝑃̅ cos[𝑘(𝑐𝑡 − 𝑥 + 𝜆)] = 𝑃̅ cos[𝑘(𝑐𝑡 − 𝑥) + 2𝜋] → 𝑘 = [ ] (2.3.9)
𝜆 𝑚
Consequently, the link between wave number and pulsation will be:
2𝜋
𝜔= 𝑐 = 𝑘𝑐 (2.3.10)
𝜆
Equation (2.3.7) can be rewritten as

𝑝(𝑥, 𝑡) = 𝑃̅ cos(𝜔𝑡 − 𝑘𝑥) = 𝑅𝑒(𝑃̅ 𝑒 𝑖(𝜔𝑡−𝑘𝑥) ) (2.3.11)

This mono-harmonic relation is the basis for acoustic comprehension, since according to
the analysis of Fourier, whichever signal can be seen as an overlap of single harmonic
contributions as said also before. Moreover, the system is linear and so the superposition
theorem can be used. This mean that (2.3.11) is the generic harmonic contribution of the
solution of any wave. The relation between sound pressure and particle velocity is:
1
ux (𝑥, 𝑡) = 𝑃̅𝑒 𝑖(𝜔𝑡−𝑘𝑥) (2.3.12)
𝜌0 𝑐

The important consideration emerging from (2.3.12) is that pressure and velocity are in
phase: in absence of reflecting surfaces or obstacles, pressure and velocity reach their
maximum values at the same time, leading to the maximum transmission of sound power.

2.3.3. Acoustic emission of an infinite plate


[8] Plate’s vibration is a back and forth movement: to avoid discontinuity, the air close
to the interface has to move accordingly, following the surface movement. This quick
behaviour generates a sequence of rarefaction and compression waves in air, which move
at the speed of sound and in phase with the vibrating surface. The sound pressure field
and the flexional wave at a certain time instant must have the same dependence on x:

𝑝(𝑥, 𝑧) = 𝑝0 𝑒 −𝑗𝑘𝐵 𝑥 𝑒 −𝑗𝑘𝑧 𝑧 (2.3.13)

The flexional wave has velocity 𝑐𝐵 along x direction and generates a pressure wave with
sound velocity 𝑐. It is a plane wave (since the plate is infinite), its front is parallel to 𝑦
and the direction of propagation in 𝑥-𝑧 plane depends on 𝜆 and 𝜆𝐵 ratio.

𝑐 = 𝜆𝑓 ≠ 𝑐𝐵 = 𝜆𝐵 𝑓 (2.3.14)

29
Figure 2.3.2 - Sound emission from vibrating plate
The wave number is:
2𝜋 2𝜋
𝑘𝐵 = ; 𝑘= ; 𝑘𝑧2 = 𝑘 2 − 𝑘𝐵2 (2.3.15)
𝜆𝐵 𝜆

𝑘𝑧 has to be positive: 𝑘𝐵 < 𝑘 → 𝜆𝐵 > 𝜆

Taking into account the link between pressure 𝑝 and particle’s velocity along 𝑧, 𝑢𝑧
previously found, and ignoring the time dependency, on the interface it has to be imposed
the particle velocity to be equal to vibration velocity 𝑣:
𝑘𝑧
𝑢𝑧 (𝑧 = 0) = 𝑣 → 𝑝 𝑒 𝑗𝑘𝐵 𝑥 = 𝑣0 𝑒 −𝑗𝑘𝐵 𝑥 (2.3.16)
𝜔𝜌0 0
𝑣0 𝜔𝜌0 𝑣0 𝜌0 𝑘𝑐
𝑝0 = = (2.3.17)
𝑘𝑧 𝑘𝑧
Combining (2.3.22) and (2.3.23):
𝑣0 𝜌0 𝑐 −𝑗𝑘 𝑥 −𝑗(𝑘𝑐𝑜𝑠𝛼)𝑧
𝑝(𝑥, 𝑧) = 𝑒 𝐵 𝑒 (2.3.18)
𝑐𝑜𝑠𝛼
Again, valid only when 𝑘𝐵 < 𝑘 → 𝜆𝐵 > 𝜆. When this condition is not verified, it is
possible to see how pressure decreases exponentially with the increase of distance z
(going away from the source). It is the “near-field” condition: the sound emission is
inefficient and confined to a short space near interface.
𝜆 𝑘𝐵
𝛼 = arcsin ( ) = arcsin ( ) (2.3.19)
𝜆𝐵 𝑘

𝑘𝐵 is plate’s flexional wave number, k is the acoustic wave number. Notice how 𝛼 →
90° when 𝜆 → 𝜆𝐵 : looking at figure 2.3.2, the wave front tends to become normal to the
plate: in this condition 𝑐𝐵 = 𝑐. Sound emission becomes efficient when the velocity of
flexional waves is higher than sound speed (supersonic velocity). When 𝑐𝐵 = 𝑐, then

𝐵 4 𝑐 2 𝜌𝑠
𝑐𝐵 = √ √𝜔 = 𝑐 → 𝑓𝑐 = √ (2.3.20)
𝜌𝑠 2𝜋 𝐵

𝑓𝑐 is the critical frequency.

30
The effect of the wavelength is very important in sound propagation: depending on the
ratio between the wavelength of the flexional wave and of the sound wave, the sound
emission will be higher or lower: when 𝜆𝐵 < 𝜆, there is no sound emission. Even if this
conclusion comes from the hypothesis of an ideal infinite plate, it is still valid for any
finite plate. In particular, in a theoretical infinite plate the sound emission is null when
𝑓 ≪ 𝑓𝑐 due to radiation cancellation phenomena (fluid dynamic short-circuits between
compression and rarefaction waves).

2.3.4. Acoustic emission of a finite plate


Consider a thin plate laying on its four edges. Without going too much into details, it has
to be underlined that, contrary to infinite plate, a finite plate has its own n modes of
vibrating and natural frequencies: it means that when excited, the plate will vibrate more
at its resonance points and consequently the acoustic emission will be higher at those
frequency. If for a bar or a string the waves are mono-dimensional, in 2D the plate will
be characterized by peaks and valleys alternated in phase and counter-phase.

Figure 2.3.3 - Modes of vibrating for finite plate [a]

On a finite plate the edges are able to emit sound: to distinguish the cases of sound
emission at 𝑓 < 𝑓𝑐 or 𝑓 > 𝑓𝑐 , they are referred to as “surface modes” when the entire
plate is emitting sound due to its vibration at 𝑓 > 𝑓𝑐 (finite and infinite plates behaves
the same, so what was said before is still true); and as “corner or edge modes” when 𝑓 <
𝑓𝑐 and sound emission is limited to the boundary of the plate. Notice also that the larger
is the size of the plate with respect to sound wave length, the more efficient will be
acoustic emission (the louder will be the volume of sound).

Taking in analysis a simple infinite plane surface, vibrating in phase at velocity


𝑉0 cos(𝜔𝑡) (that is, the emitted sound waves are all parallel to the surface itself, due to
the straight surface moving back and forth in space), the particle’s velocity will be:

31
𝑢(𝑡, 𝑧) = 𝑉0 cos(𝜔𝑡 − 𝑘𝑧) (2.3.21)

Since the waves are plane, sound intensity will be:


2
𝑝𝑟𝑚𝑠 ̅̅̅
𝑝2
𝐼= = ̅̅̅2 = 𝑊 = 𝜌0 𝑐𝑣
= 𝜌0 𝑐𝑢 ̅̅̅2 (2.3.22)
𝜌0 𝑐 𝜌0 𝑐 𝑆

(S vibrating surface, 𝜌0 𝑐 acoustic impedance, ̅̅̅


𝑣 2 mean quadratic velocity of surface
vibration, averaged on the surface itself, 𝜎 acoustic efficiency. 𝑣 and 𝜎 are function of
frequency, and so does 𝑊). Acoustic efficiency can be defined on the basis of sound
power:
̅̅̅2 𝜎
𝑊 = 𝜌0 𝑐𝑆𝑣 (2.3.23)

Consequently
𝑊
𝜎= (2.3.24)
̅̅̅2
𝜌0 𝑐𝑣

- 𝜎 ≪ 1  𝑓 < 𝑓𝑐 , sound emission at low frequency;


- 𝜎 = 1  𝑓 = 𝑓𝑐 , finite and infinite surface have same efficiency;
- 𝜎 → 1  𝑓 > 𝑓𝑐 , sound emission at high frequency;
- 𝜎 > 1  𝑓 ≫ 𝑓𝑐 , it is possible an efficiency higher than 1;

In noise control it is desirable to have 𝑓 < 𝑓𝑐 so that sound is emitted inefficiently: it


mean that the force on the plate should be at a frequency lower than the critical one. It is
possible to act structurally on the plate to increase its 𝑓𝑐 , for example by reducing the
thickness. Reducing thickness will reduce the number of natural pulsation 𝜔𝑛,𝑛 , which
mean having shorter wave length 𝜆𝑥 , 𝜆𝑦 in the field of frequencies of the force acting on
the plate. The wavelength of plate’s modes will be lower than the acoustic wavelength,
leading to a lower efficiency and consequent sound emission.

2.4. Sound
2.4.1. The use of Decibel
The pressure waves commonly generated by a vibrating body, are actually very small if
compared to ambient pressure: due to this fact, human hear evolved to capture sound
pressure waves in a wide range: from 2 ∙ 101 𝜇𝑃𝑎 to 2 ∙ 106 𝜇𝑃𝑎. It is then preferable to
use a logarithmic scale on basis 10 to deal with the width of validity concerning sound:
a reference value is defined for the quantity involved (sound pressure: 20 𝜇𝑃𝑎; sound
intensity: 1 ∙ 10−12 𝑊/𝑚^2; sound power: 1 ∙ 10−12 𝑊). The sound can be expressed
both in sound pressure level, sound intensity level and sound power level:
32
Sound pressure level (SPL):
2
𝑝𝑟𝑚𝑠 𝑝𝑟𝑚𝑠
𝐿𝑝 = 10 log10 2 = 20 log10 [𝑑𝐵] (2.4.1)
𝑝𝑟𝑒𝑓 𝑝𝑟𝑒𝑓

Sound intensity level:


𝐼
𝐿𝐼 = 10 log10 [𝑑𝐵] (2.4.2)
𝐼𝑟𝑒𝑓

Sound power level:


𝑊
𝐿𝑊 = 10 log10 [𝑑𝐵] (2.4.3)
𝑊𝑟𝑒𝑓

The important result of using a logarithmic scale


is that dB are not linear: a sound having 3dB more
than another one (53 𝑑𝐵 vs 56 𝑑𝐵 correspond to
20 ∙ 10−8 𝑊 vs 40 ∙ 10−8 𝑊), is actually two time
more powerful; 10dB correspond to a sound 10
times stronger (53 𝑑𝐵 vs 63 𝑑𝐵 correspond to
20 ∙ 10−8 𝑊 vs 200 ∙ 10−8 𝑊). Concerning SPL,
hearing pressure field in range 20 𝜇𝑃𝑎 - 200 𝑃𝑎
is equivalent to 0 dB (“hearing threshold”) -140
dB (“pain threshold”). Reference level are such
that in free field, sound pressure level and sound
intensity level are the same in dB.
Figure 2.4.1 - Equivalence between sound
Since it is based on a logarithmic scale, dB cannot pressure and sound power and Decibels [b]
be summed directly. If there are two sources of sound, emitting 𝑝1 (𝑡) and 𝑝2 (𝑡), then
final pressure is given by the overlapping of those: 𝑝(𝑡) = 𝑝1 (𝑡) + 𝑝2 (𝑡).

More in general, in case of two non-correlated signal:


2 2 2
𝑝𝑟𝑚𝑠 = 𝑝1,𝑟𝑚𝑠 + 𝑝2,𝑟𝑚𝑠 (2.4.4)

Then:
2 2 2
𝑝𝑟𝑚𝑠 𝑝1,𝑟𝑚𝑠 + 𝑝2,𝑟𝑚𝑠
𝐿𝑝𝑖 = 10 log10 ( 2 ) = 10 log10 ( 2 ) (2.4.5)
𝑝𝑟𝑒𝑓 𝑝𝑟𝑒𝑓
2
More in general, in case of 𝑛 different values of 𝑝𝑖,𝑟𝑚𝑠 :
𝑛
2 𝑛
𝑝𝑖,𝑟𝑚𝑠 𝐿𝑝𝑖 𝐿𝑝1
𝐿𝑝 = 10 log10 (∑ 2 ) = 10 log10 (∑ 10 10 ) = 10 log10 (10 10 + ⋯ ) (2.4.6)
𝑝𝑟𝑒𝑓
𝑖=1 𝑖=1

33
If two equal level are summed, the final total level is 3dB higher than the original ones:
as previously said, 3dB correspond to a double power. On the contrary, if two levels
differ by 10dB, the total level will be just 0,5dB more than the original higher one.

2.4.2. Frequency
The other important quantity in sound analysis is
frequency. Human hear is able to recognise
frequencies in the range 20 𝐻𝑧 - 16000 𝐻𝑧 and
in the form of a single tone (a sound emitted at a
single frequency, like a beep on a computer), a
complex sound, made of different frequencies
(most of sounds belong to this category) or a
random noise (random frequencies, not linked
one with each other).
The spectrum of a sound is obtained by applying
Fourier transform to the sound sampled in
pressure expressed in dB. Since the spectrum of
a sound can be very wide and often it is not only
a single frequency which is excited, but a group Figure 2.4.2 - Octave and 1/3 octave bands [c]
of frequency close one to the other, it is preferable to divide the range 20 𝐻𝑧 - 16000 𝐻𝑧
into bands. Each band contains a well-defined range of frequencies: usually octave bands
or one third octave bands are used. The octave is defined so that the two extremes of the
band are one the double of the other:

𝑓𝑢 = 2𝑛 𝑓𝑙 ; 𝑓𝑐 = √𝑓𝑢 𝑓𝑙 (2.4.7)

Where 𝑓𝑢 is the upper frequency of the band, 𝑓𝑙 is the lower frequency of the band and 𝑓𝑐
1
is the central frequency of the band. If 𝑛 = 1, entire octave bands are used; if 𝑛 = 3, one-
third octave bands are used.

The reference value for dividing the spectrum into octave bands is 1000 𝐻𝑧 which is the
central frequency of one octave band. Starting from this value, all the other bands can be
evaluated, since each central frequency is the double of the previous one. The width of
each band grows with the frequency.

The sum of SPL is not influenced by frequency: if there are sound expressed in dB
belonging to different bands of frequency, they are still summed as previously shown.
However, human hear feels different frequencies in different ways: often a filter should

34
be used to weight the spectrum
according to human hear
sensibility. The most used is
the A-filter, which weighs
more the frequencies in the
range 1000 - 5000 𝐻𝑧 and
less the lower ones.
Figure 2.4.3 - A-weighting filter

2.4.3. Sound emitted by fans


[9] The origin of fan noise is both tonal and broadband. The broadband sound comes
from turbulence at the intake, interaction of turbulent boundary layers and blade surface,
trailing edge, vortex shedding, interaction of boundary layers downstream of the trailing
edge and boundary layer separation. It is mainly concentrated in the higher frequency
region of the spectrum (higher than 1500-2000Hz, up to 4500-5000Hz).
Tonal noise is generated by aerodynamic interaction between the flow and the elements
present on its path. It gives origin to the main peaks in the spectrum, which are the Blade
Passing Frequency and its harmonics. It is stronger at lower frequency (below 500Hz)
and still important in the range 1000-2000Hz [11]. Spectra of fans noise have main peak
at frequency which is the BPF: it depends on the rpm and on the number of blades [10]:
𝑁𝑏𝑙𝑎𝑑𝑒𝑠 𝑟𝑝𝑚
𝐵𝑃𝐹 = (2.4.8)
60
The source of higher fluid dynamic noise is the tip of the blade: it comes from the leakage
of flow, and is linked with the passing of each blade in one point [9]. This gives origin
to vortex, due to the difference in pressure between the pressure and the suction side. The
main parameter on which tip sound depends on is the tip clearance: the larger it is, the
louder will be noise emitted due to larger vortex: sound pressure becomes higher along
the whole spectrum, and an increase in a frequency lower than the BPF is registered.

Sound power level emitted by axial fan can be estimated as:

𝐿𝑊 = 𝐾𝑊 + 10 log10 𝑄 + 10 log10 𝑃 + 𝐵𝐹𝐼 + 𝐶𝑁 (2.4.9)

𝐿𝑊 is SPL; 𝐾𝑊 specific SPL of the fan, given by manufacturer; Q is volume flow rate; P
total pressure; BFI blade frequency increment, which is a correction linked with the blade
passing frequency (BPF); 𝐶𝑁 is the efficiency correction.

To have more silent fans, they have to be operated close to their peak efficiency point,
and it is necessary to reduce variable airflow applications (keeping constant the amount

35
volume of air moved). Sound power emissions of fans space on a wide range of
frequencies, depending on the specific characteristic: backward inclined centrifugal fans
offer best performances in terms of noise, while the axial ones are the worst, because
they produce high specific noise level across a wide range of frequencies. If the casing
is not well designed, it will be also responsible for noise emissions: this noise is
associated not only to vibrations of the structure (the stiffer it is, the lower are sound
emissions), but also with the reverberation of noise generated by the impeller and air
movement, due to section sizes.

In order to reduce fans noise emissions for radiators, there are two possible ways:
- Heavy gauge materials: it is a cheap and effective solution, which act on a wide
range of frequencies. The ticker is the material, the lower will be sound emissions.
- Lagging and cladding: external light gauge in metal, or the use of insulating
material can be used if the previous solution is not enough for keeping the noise
below a certain threshold. There are limited both in terms of space and costs.

2.4.4. Sound emitted by turbochargers


Accordingly to an analysis conducted by Calvo, Diaz, San Romàn [12] the noise emitted
by turbocharger has many origin, as the tab below shows:

Table 2.4.1 - Noise origin in turbocharger [d]

Among these, the main contributor is the whistling, linked to vibration associated with
unbalance of the rotor. The cost for balancing a turbocharger rotor are very high,
consequently it is often considered balanced when the vibration generated are below a
certain threshold. This residual unbalance, provokes vibrations along the exhaust systems
that behaves as an amplifier for acoustic emissions at high frequencies (sound emission
can be increased up to 40 times). As a matter of fact, the noise emitted comes not directly
from the turbocharger, but from the elements that goes in resonance with it.
36
Vibrations due to the unbalance of the rotor are excited by a force proportional to the
square of the rotational speed:

𝐹 = 𝑚 × 𝑟 × 𝜔2 (2.4.10)

Turbocharger at nominal speed can


rotate from 50000 to 100000rpm:
consequently, even if the unbalance
mass m is small and the radius of its
eccentricity is infinitesimal, their
effects has to be increased by 8-10
order of magnitude. Moreover, the
frequency of excitation will be
concentrated mainly at higher bands. Figure 2.4.4 - Turbocharger noise spectrum [d]

[17] Analysis on Wärtsilä engines shown common vibration level on the turbocharger
filter is in the range of 4g acceleration, which is close to the accepted vibration level. If
a turbocharger has a higher acceleration, it is probable that the rotor is unbalanced.

2.4.5. Sound emitted by alternators


A document dated 1966 from the Defense department of the USA [14] constitute a guide
on electric engines and alternator noise generation. Different are the sources of emission:

- Magnetic noise: the magnetic field generated by windings around ferromagnetic


material leads to two main sound emission sources: magnetostriction is an
infinitesimal contraction of the materials linked with the frequency of excitation;
the second are the magnetic forces that have origin in the air gap and radiate in
different directions, with different amplitude and frequencies, depending on the
air gap. Magnetostriction radial forces can reach up to 50% of the radial forces
produced in the air gap magnetic field. The magnetic stress wave works in radial
directions on the stator and on active surfaces of rotor. This causes the
deformation and subsequently vibration and noise [13]. The frequencies at which
forces of vibration acts are synchronous with rotating speed and are evaluated as:
𝑛𝑁𝑟𝑠
𝑓𝑟 = 𝑓1 [ (1 − 𝑠)] (2.4.11)
𝑝
𝑓1 : Supply frequency; 𝑠: Slip factor; 𝑝: Number of poles;
𝑁𝑟𝑠 : Number of rotor slots; 𝑛: Harmonic number
- Rotor eccentricity: as mentioned for turbocharger and engine cinematic, the
eccentricity of rotating elements may lead to vibrations of surfaces that goes in

37
resonance. Nevertheless, for alternator eccentricity has also the effect of changing
the air gap previously mentioned. Quoting [13], there is stronger interaction of
stator and rotor magnetic field at the point where the gap is smaller. Influence of
the static eccentricity manifests as the emergence of side frequency bands, which
are shifted from the mains frequency 𝑓1 of the synchronous frequency 𝑓.
- Aerodynamic noise: alternators have an inside fan used to cool down the wirings.
Noise from fans has already been described
- Mechanical noise sources: the main source of noise comes from bearings, due to
their defects, sliding contacts, rotor unbalance. Temperature, running speed,
lubrication, misalignment are the main actors for bearing noise. Apart from
excited frequencies, noise from bearing is of secondary importance.

2.4.6. Sound sources comparison in generating sets


[15] [16] According to principal manufacturers, noise emission has origin mainly from
engine combustion and vibration (100 dB up to 121 dB) and fans (100 dB up to 105 dB).
Exhaust noise is a major contributor to overall sound levels and, when measured without
an exhaust silencer can be 120 dB(A) or more, depending on the size of the engine. The
sound level can be reduced by up to 40 dB(A) depending on the silencer [15, p4, par2].
All the tested engines had the exhaust system far from the point of measure and insulated,
moreover silencers were used, so it is of secondary importance.
Turbocharger noise has main effects if coupled with the exhaust system. Moreover, its
field of emissions is at high frequency, and again is of secondary importance.
Alternator contributes mainly with its fan, while the electric and induction noise change
with the speed and the load, but are synchronous with the rotation, so their contributions
will be overlapped with the one of the engine. The ranges of sound emitted space in the
range 80-90 dB.

2.4.7. Sound insulation


[15] [16] Generating set are usually closed inside containers the walls of which are filled
with sound-absorbing acoustic foam, effective for controlling high-frequency noise. Fire-
retardant urethane foam and fiberglass are common material used in enclosures.

Cooling air is a significant source of high-frequency noise, but restricting its flow can
reduce generator set cooling efficiency. In indoor installations, high-frequency noise can

38
be reduced by making the air flow to turn upward, so that much of the air flow sound at
ground level is reduced, being directed away from people and other structures.

[16] Rigid materials with significant mass and stiffness reduce the transmission of sound.
Common used material such as sheet steel typical of enclosures, concrete- or sand-filled
block walls or solid concrete walls are typical of indoor installations. It is also important
to eliminate sound paths through cavity and holes in doors or walls, or through access
points for exhaust, fuel or electrical wiring.
Steel and aluminium walls provide at least 10 dB(A) of attenuation for generator sets that
must be located outdoors. When a greater amount of attenuation is needed to meet local
noise ordinances, special sound-attenuating enclosures have to be adopted. Enclosure’s
cost is proportionally related to the level of sound attenuation required. In extreme cases,
it is not uncommon for the cost of the sound-attenuated enclosure to equal the cost of the
generator set. Some enclosures also may negatively impact generator set performance by
limiting ventilation capacity. Careful design from the outset is important to attain noise-
control goals while maintaining generator set performance. Special sound-attenuating
enclosures combine both barrier and absorption noise control strategies.

Figure 2.4.5 - Enclosure structure and insulation strategy [e]

39
3. Measurements

Two very useful guides for the choice of the point of measure, the timing and the
procedure to follow were the works on Diesel engines of Örn [17] and Lindholm [18].

3.1. Description of the instruments


3.1.1. Input module NI-9234
This National Instruments, “C Series, Sound and Vibration Input
Module” has 4 differential analogical input channels, a maximum
sample rate of 51.2 kS/s/channel, an analogical input voltage range
of ±5 V and an excitation current of the IEPE (Integrated Electronic
Piezoelectric) sensor of 2mA. It has a resolution of 24 bits. The
NI‑9234 delivers a wide dynamic range and incorporates software-
selectable AC/DC coupling and IEPE signal conditioning. The Figure 3.1.1 - Input
input channels simultaneously measure signals. Each channel also module NI-9234 [f]
has built-in anti-aliasing filters that automatically adjust to the sample rate.

3.1.2. Microphone BSWA MP201


The BSWA MP201 is a pre-polarized condenser
microphone, calibrated in accordance with IEC
651 Type I Standard. The diaphragm
manufacturing technology eliminates the high
frequency diffraction and ensure the flat
frequency responses.

Table 3.1.1 - Microphone BSWA MP201


datasheet [g]
Figure 3.1.2 - Microphone BSWA MP201 [g]

40
3.1.3. ICP preamplifier BSWA MA201
MA201 current drive microphone preamplifier is developed for pre-polarized
microphones. It is a high-performance preamplifier with low inherent noise and high
input impedance. The preamplifier can be directly connected to any ICP inputs. It is used
for general purposes sound measurements.

Figure 3.1.3 - ICP preamplifier BSWA MA201 [g] Table 3.1.2 - ICP preamplifier BSWA MA201
datasheet [g]

3.1.4. Brüel & Kjær DeltaTron Accelerometers Type 4508 B


Miniature CCLD
Accelerometers Type 4508,
designed to withstand the
rough environment of the
automotive industry. A Figure 3.1.4 -
Accelerometers
combination of high
sensitivity, low mass and small physical
dimensions make them ideal for modal
measurements, such as automotive body and
power-train measurements, as well as for
Table 3.1.3 - Brüel & Kjær DeltaTron
modal analysis on aircraft, trains and Accelerometers Type 4508 B datasheet
satellites.

3.2. Measurements procedures


In order to repeat each time the same steps, following same criteria and trying to use less
time possible, a method to follow was written down. As a matter of facts, time is limited,
41
since the test phase had restricted windows of one hour of duration at different load
conditions. Moreover, even if soundproof headphones are worn, both the sound and the
heat become unbearable after some time. If the measurement was too slow, then there
was not the possibility to gather all the data in the same load condition, and the
measurement remain partial.
In one case, measuring vibrations with 100% load, the worst condition, they passed
suddenly to 0% load, which, means completely different values of vibration and noise
and a waste of the day.

3.2.1. Vibrations
Measurement points
The choice of the measurement points depends from engine to engine. For vibrations, the
points are chosen accordingly to standard ISO 8528-9, furthermore some are chosen
arbitrarily where some acceleration components are considered to be more evident. It is
important to pay attention on where to place the accelerometers: it is better to avoid
flanges, beams, bars, pipes, which can have their own accentuated vibration modes,
substantially different from the ones of the engine block.

For each side, the places where to put the sensors are two on the upper part of the engine,
two below (possibly not on the flanges which link the engine to the base, which lead to
disturbances), and then on the surface, on the back and on the bases of the generator (ref.
ISO 8528-9). It is then necessary to add measurements on the
radiator’s surface and other points, like on the engine head,
near the valves, on vibrating surfaces, under engine’s base.

Having just two accelerometers, for each point two set of


measures are taken: the first containing two directions (x and
y), and the other the third (z).
Procedure Figure 3.2.1 - System of refernce

For the first set of measurements taken at the beginning, the


procedure consists in positioning metallic adhesive tape on the measurement points
(which are chosen before measuring), when the engine is not running. The choice has to
be done so that the three perpendicular directions x, y, z are easily identifiable, according
to ISO 8528-9.
It may happen that the point of measurement has a certain inclination of the relative axis
with respect to the absolute ones: in this case it is necessary to correct this inclination,
measuring it with a goniometer (phone app is used, “Clinometer”, by “Plaincode”).
42
When the engine is started, on the adhesive tape it is pasted
the accelerometer with a drop of Loctite Super Attack, which
takes 20-25 seconds to solidify (less time in the hotter
points). This is done because the engines are painted inside
the company accordingly to the final customer desires:
applying the sensor directly to the engine with the glue may
cause the undesired detachment of the paint. The tape’s
adhesive power is better distributed, and this problem is
avoided.
In a second phase of the measuring, since the use of tape
presented some issues, instead of the described approach, the Figure 3.2.2 - Predisposition of
instruments for measurement
accelerometers were fixed on a neodymium magnet, making
faster the process of measurement and avoiding failure in the acquired measures.
The accelerometers are then linked to the acquisition board, Input module NI-9234; the
board is linked to the computer, and data are acquired using the software “MeasLab Data
acquisition”.

Figure 3.2.3 - Channel position of the sensor: channel 0 accelerometer 4, channel 1


accelerometer 5, channel 2 microphone

Photography and data saving


It is important to take a picture of the
disposition of the instruments in each point,
comprehensive of a reference able to recall
the kind and number of the measurement.
This is done by writing on a paper the
required information, and placing it in a
corner of the picture.
Data saving with MeasLab has to be done
Figure 3.2.4 - Positioning of sensors and picture with
selecting the channel that will be used: each reference number (V5, V3)

43
channel is associated with a sensor, as in the figure below. It is then required to set the
sampling frequency and the buffer length. If a filter is needed there is the possibility of
choosing one.
Finally comes the creation of a new project having title in the form “date – kind of
measure (fans sound/engine sound/vibration…) – name of the engine”; each single
measurement changes the subtitle, so that it contains the date, kind of measurement,
number of the measurement, name of the accelerometer and the direction along which is
placed the sensor. Following these steps in this way, it is easier in the data processing
phase to recognise the signal and linking it to the correspondent picture. An example:
“2905 vibration 3 acc4x acc5y”.

Timing
The measurement of the vibrations in each point has a temporal window of 10-15
seconds, with sampling frequency of 10240Hz. The globally required time to measure in
the three directions of a point takes more or less 2 minutes and 15 seconds and is
comprehensive of:
- Positioning the accelerometers (25s each with glue, 10s for magnets);
- Going back to the computer, taking picture of instrument disposition and starting
measuring (10s);
- Time for measuring and data saving for the first two coordinates (20s);
- Detachment and repositioning one of the two accelerometer (25s each with glue,
10s for magnets);
- Going back to the computer, taking picture of instrument disposition and starting
measuring (10s);
- New measurement for the third coordinate (20s);
The number of points depends on the type and dimensions of the engine. For example,
on one side there could be oil filters, inlet of gasoil pipes, particular geometries which do
not allow the positioning of the accelerometers in one of the three dimensions or other
impediments (which can also be linked to very high temperatures, usually near turbine
and exhaust manifold).

The choice has also to be done depending on the available time for the measurement:
during the phases of the final test, the engine goes half an hour with 0% load, half an
hour at 25% load, one hour at 50% load, one hour at 70% and 2-3hours at 100%. On the
contrary, during ordinary tests in can be continuously switched on and off, for few
minutes, with or without load, making almost impossible to get any data. Obviously, the
presence of the load changes significantly the measured parameters. In total, assuming
to take measures according to standard ISO 8528-9, plus other three measures (radiator,
44
engine head and base), there are 11 measures for each side, which require 50 minutes
only for vibrations.

Also, it is important to keep in mind that the fuel consumption of the engines is very
high, hundreds of litres of gasoil each hour, equal to hundreds of euros. Therefore,
measurements are limited to the phases of final tests: it is possible to ask for 5-10 minutes
more, but the temporal windows are very restricted and often there are also the final
customers or an external inspector who take part to the test, and measuring becomes
difficult.

Problems
Temperature: For the vibrations, mainly at the beginning, there were some practical
problems. The excessive heat in some points made the glue of the tape to weaken, leading
to the detachment of the accelerometer or to the formation of an air bubble below it
(figure 1 and 2).
Moreover, to fix the accelerometer the used glue requires 20-25s to harden: for this
interval of time the sensor has to be hold by hand, but due to the high temperatures in
some points it was almost impossible.

Figure 3.2.6 - Bubbles of air below the tape Figure 3.2.5 - Thermal image of the exhaust pipe:
Tmax=459,7°C

Material: temperature issues were in part solved by using


some magnets to place the transducer. This made easier and
faster the process of measurement, but again there were some
cons: the main one is that not all parts of the engine are
magnetics (some are in aluminium), and no measures could
be taken there. A second problem was that the vibrations in
some points made the magnet to move around the area of
measurement, so during the acquisition, the point of
measurement changed. The result of this is a velocity of the Figure 3.2.7 - Positioning of
the accelerometers on a
sensor which was not null. In order to partially reduce this resonant plate
45
inconvenient, the two magnets are placed so that one keep the other more or less in the
same position: this is what is obtained:

Graph 3.2.1 - Velocity and displacement from integration Graph 3.2.2 - Velocity and displacement from
of acceleration along z integration of acceleration along x
Direction z has a larger magnet and it is not subjected to gravity, so it remains more or
less fixed. On the contrary the other transducer is subjected to a small displacement of
few mm.

3.2.2. Noise
Measurement points
The sound measurement is done according to the standard ISO 8528-10, which requires
to acquire sound in nine points around the plant at a distance of on meter. Since the scope
of the measurement is to establish which parts of the system generate a certain noise at a
certain frequency the measurement of the sound has to be done as close as possible to
the source.
Moreover, the measurement points should be adapted so that it is easier to separate the
different components of it (for example by getting away from the radiator and getting
closer to the generator, it is expected a reduction of the peak of frequency correspondent
to the aerodynamic noise of the fan and an increase of the mechanical one of the
generator, which anyway is low).

Procedure
The configuration for the measurement of sound is similar to the one for vibration, with
the same acquisition board and software. The microphone is (name and tipe of the
microphone). The length of the signal is again 10-15 seconds, at the same sampling
frequency of 10240 Hz.
Three kinds of measurements are required for sound: the first one is the one with the
stopped plant, with only the ambient noise. It is sufficient to do only one measure, taking

46
for assumption that the presence of the engine, radiator and generator do not affect
background noise.

The second one consists in the measure of sound


around the engine, when it is switched off, but with
the suction fans on. Those are placed on the wall
which gives on the outside, and push out the air
exiting from the radiator, granting a recirculation of
air and avoiding an unbearable rise of the room
temperature. In this case, since the fans are placed on
a certain fixed point of the room, different
measurements are required in the points where later
engine noise will be taken. Doing so, in the phase of
data post processing it is possible to eliminate from
the signal the fans component which is not of interest.
Figure 3.2.8 - positioning of the
The third measure is the final one on the plant. The microphone for measurement outside a
sources of noise expected to be found are the radiator container
(aerodynamic noise of the fan), the engine (mechanic, intake and combustion noise), the
turbocharger (two compressors and two turbines) and the generator, and all the vibrating
surfaces.
The microphone is mounted on an arm placed on an easel, which allows to bring the
instrument as close as possible to the source. For each side, the measurements are done:
behind and aside of the radiator (not on the front for safety reasons), two or three on the
side of the engine block (depending on the dimension, also at different height), close to
the turbocharger if possible and aside and behind the generator. In total 7 measures for
each side.

Saving
As it is for the accelerometers, also in this case it is necessary to take a picture for each
position of the microphone and save so that data and picture are easi to be coupled. The
title will be similar to the previous one, for example: “2905 fans sound 3”.

Timing
The time required for sound measurement is less than the one required for vibrations,
around 17 minutes, ad it is divided into:
- positioning the easel with the microphone (5s);
- Going back to the computer, taking picture of instrument disposition and starting
measuring (10s);

47
- Time for measuring and data saving (20s);
In total only 35 seconds. These operations are repeated for each set of measures, requiring
more or less 7 measures for each side, 8 minutes.
Total time required for obtaining good data both for noise and vibration is roundabout 1
hour and 10 minutes.

3.3. MeasLAB
MeasLAB is a program for data acquisition and analysis. It is divided into two parts:
MeasLAB Data Acquisition and MeasLAB Signal Processing.

3.3.1. MeasLAB Data Acquisition


Before starting the program, the acquisition board should be linked to the computer. At
the first use of the program, or each time a new instrument is used, calibration is required.
Going on “Create new MeasLAB Project” a window opens:

Figure 3.3.1 - MeasLab first board: sensor selection

among the various options, with “Sensor Calibration” it is possible to select the
instrument to calibrate, inserting the measurement technique, unit of measure, etc.
Clicking on “CAL” there are two options for calibration of accelerometers or
microphone. At this point it is required to start the device used for calibration, place on
it the sensor and insert in the correct space the reference values provided by the device.
The program will fix the values received from the sensor as the reference, obtaining the
scale factor [EU/V], which in our case are. At this point it is possible to start with the
measurement.

Again, click on “Create new MeasLAB Project” and remain on “Channel Config” and
select the instrument which will be used. Going on, a new window requires to set the
sampling rate, the length of the time buffer, the duration of the buffer, the samples per
channel. After this, another one opens requiring information about the filter: topology

48
Figure 3.3.2 - MeasLab boards for sampling frequency, time window, filters, saving

(Off), type (bandpass), order (8), lower (1) and upper (22) limit and other parameters.
Finally, it is required to select the folder in which the data will be saved and the “Log
settings”, which are the information that the program will take and save from the
instrument. It is enough to take time history and RMS. Resampling frequency and
weighting can be left as the default ones.

In order to start the acquisition process, change page and select “Data Acquisition”: it is
enough to click on the green arrow to start sampling, and on the red square to stop it.
When the sampling ends, the program asks to insert the new name of the measuring set.
It is important to specify each time a new name, for example by changing a number, in
order to not generate problems in the following steps of data processing. Once the
program is arrested, in the selected folder there will be another folder with the name of
the data set and a file.MLP which is the one containing all the sampled data.

3.3.2. MeasLAB Data Processing


To create a new project, click on “New”, “Connect to TDSP Database”, select the folder,
select the .MLP file: the measured set will be loaded on the small window on the upper-
left of the screen. Select all the folders and click on “Download data from…” and all the
measurements are placed in “Imported points”. At this point it is possible to analyse,
export, filter, cut the data. Useful functions can be found in “Advanced”:
- “Time Domain Analysis (TDA)” provides the time history of the signal, the
statistic information about it (RMS, average, max/min…), it offer the possibility
of filtering it and select only portion of it.
49
- “Frequency Domain Analysis (FDA)” offer the same possibilities of TDA plus
the possibility of changing window, looking for peaks, etc. Going on cursor and
selecting “Harmonic cursor” it shows the harmonics of a certain peak.
- “Short Time Fourier Transform (STFT)” gives the colourmap of the signal and
allows to vary the parameters of resolution both in time and frequency domain.
- “Statistics” provides statistical values of each sample (kurtosis, mean, standard
deviation, variance, RMS, the histogram profile, probability distribution)
- “Edit” allows to compute derivatives or to integrate (for example from
acceleration to velocity), to weight the signal, filter it, compute some operations.

3.4. Grid for turbocharger


It worth mention one strategy that was
developed for turbocharger measurement.
At the beginning, when they were testing
the smaller engines that had the
turbocharger on the side at human height,
in order to individuate the emissions from
the turbocharger, I built a grid for the
measurement of noise close to the source,
assuming it to be like a point source
(90x100cm, 6x7=42 points at discrete
Figure 3.4.1 - Grid that should be used for measuring distance one from the other). On a bar are
mounted the supports for the microphone, which are allowed to slide and be fixed in
position. The bar can be positioned at the desired height. The grid itself has a vertical
degree of freedom. When the grid was ready, there were no more engines with the
required characteristics, and the grid was never used.

50
4. Results of measurements

The huge quantity of data collected concerning both noise and vibrations, and the width
of those two fields themselves, requires dividing the analysis into subparts. In particular,
the discussion will start with all that regards the sound field alone; it will follow a similar
part concerning vibrations, that will be presented in the time domain. Finally noise and
vibration will be considered together, dealing mainly with the frequency domain.

4.1. Noise
4.1.1. Background noise: indoor testing room
For the engines tested in the testing room, first of all it was necessary to measure the
background noise when the engine was not running, both with fans switched on and off.
The following measurements represents the sound condition of the room measured at the
centre of it, when the fans are off. Figure 1 shows the time history and the frequency,
figure 2 the 1/3 octave band spectrum and figure 3 the SPL in dB and the RMS of it
(sound level is at 70,17 dB, at very low frequency, with a peak of 87,87 dB):

Graph 4.1.1 - Sound pressure and SPL in dB of background noise


in testing room

Here are provided one third octave band spectra, one with Pa as measured by the
instrument, the other with sound pressure level in dB.

51
Graph 4.1.3 - One third octave band spectrum in Graph 4.1.2 - One third octave band spectrum in
Pa of background noise in testing room dB of background noise in testing room

For more clarity, from now on it is chosen to use only the first type of graph with Pa and
no dB, because the peaks are more evident and the comparison between spectra is easier
(using Pa it is possible to reason linearly, while with dB it is necessary to keep in mind
that the scale is logarithmic). If it worth, also the representation in dB will be provided.

The measurements with the fans on are provided below:

Graph 4.1.4 - Sound pressure and SPL of background noise


indoor, ventilation fans running

Fans introduce a component of sound exactly at 200Hz and its harmonics, and even if
the RMS of the SPL is 95,36 dB, the peak is 107,58 dB, which is quite important, and it
is entirely linked to fluid dynamic noise.

52
Graph 4.1.5 - One third octave band in Pa of Graph 4.1.6 - One third octave band in dB of
background noise indoor, ventilation fans on background noise indoor, ventilation fans on

As a matter of fact, the four big fans have to move a large quantity of air in order to grant
a proper ventilation of the room (as already mentioned, the heat released by the engine
is very high).
Comparing the two cases: focusing on the RMS, the difference of 25,19 dB it is
equivalent to a sound power with the fans which is about 25 times the sound with no
fans. When the engine is on, it is possible to neglect the background noise at 70 dB, but
not the one at 95 dB, in particular there will be a peak at 200Hz.

4.1.2. Background noise: outside court


Some of the engines were tested inside their containers in the space outside the testing
room. The conditions in this case are
different, since even if it should be an open
field, due to logistic reasons the container is
placed close to the flat wall of the shed, with
other containers in the surrounding area and
threes on one side. For these reasons,
corrective actions should be undertaken in
order to consider the effects of those
Figure 4.1.1 - Logistic of the various container in the
obstacles. The measured background noise external court of the company
the day in which the test was performed was 76,4 dB.

According to the ISO Standard 8528-10 (appendix B, chapter 7, tab. 7.2.1) it is possible
to ignore the contribution of background noise if the delta of sound pressure level is
higher than 10 dB: no correction factors are introduced for the test performed outside,
while a correction has to be considered case by case for the engine tested indoor.

53
Bigger engines are tested outside: in the container it is very difficult to take
measurements of vibrations and sound because there is few space available and it is not
possible to enter with the computer. Still, it is possible to take measures with the
microphone outside the container to see the difference of noise emissions inside and
outside the container and also see if there are any effect of the surrounding obstacles. As
a matter of fact, one of the main problem of the company is the measurements of sound
in the open air, because there are always different layout of the objects. The risk is to get
on the phonometer higher values than allowed.

Figure 4.1.5 - position of the containers in the court Figure 4.1.4 - Position of doors and radiator

Figure 4.1.3 - Measurements taken between container Figure 4.1.2 - Measurement taken with open doors

The pictures above give an idea of condition and of dimensions. Due to the little space
available in the inside, the tripod is placed outside the container with the opened doors.
The measurements were taken all around the container, at a distance of 1m from the
container, each 1m starting from the lower edge on the left (see fig. 4.1.6 and 4.1.7) and
at an height of 1,6m. The time available for the test of this engine at 100% was not enough
to allow to take all the measures around the container also with the opened doors, so it
was necessary to take only the ones shown in the following representation:

54
Figure 4.1.6 - 24th October, outdoor test with closed Figure 4.1.7 - 24th October, outdoor test with open
door (34 point spaced by 1m) door (12 point spaced by 1m)

With the help of Matlab the obtained data were normalized and put inside the two
diagrams above, depending on their sound pressure level in dB. From figure 4.1.6, when
the doors are closed, it is possible to obtain three information:
1) The sound is higher in the surrounding of the radiator;
2) The sound is higher on the side where there is the wall of the shed;
3) On the back of the radiator the sound is lower;

The effect of the shed, which is higher than the container and the radiator, is to reflect
the sound back to the container itself, increasing the measured output. The other
container has a lower effect, since it is lower than the radiator, and only a fraction of the
sound pressure waves is reflected. The longitudinal orientation of the radiator and the
open air on the back of the container, explain why the sound registered there is lower.
With the opened door the sound is louder: it has to be considered that here there is the
overlapping of the engine and radiator contribution. Notice also that the SPL seems to be
inverted, because it is higher when the measure is taken far from the radiator, as if there
is no engine-radiator overlapping. Actually, this is caused by the position of the engine
inside the container: it is at left of it, while on the right there is the alternator structure
which acts as a shield.

It is possible to analyse the efficiency of the container and the effect of the wall:

Table 4.1.1 - SPL measured around the container with open and closed doors

The container is able to knock down the noise of averagely 18dB. A similar result was
obtained on another test (the following case), where the reduction was of 19,61dB (from
55
109,73 to 90,12dB). The containers have walls which are full of small holes and filled
with soundproof polyester wool, the thickness of which changes depending on the group
inside.
The effect of the shed is responsible for an increase in the noise which is about 1dB with
respect to container side. It can seem low, but a difference of 1dB, being in logarithmic
scale, means an increase of the 25% of perceived sound. Also, on container side the sound
seems to be the same as the back and the zone where the shed is not present, while an
higher value is expected.

Obstacles effects
Concerning the effect of obstacles in the
surrounding area of measurement, it is worth to
mention a case in which it was required my
presence by the responsible of the company. In
that case I was not able to use my instruments,
since I was working for them with their
phonometer.

They had a problem with the radiator of an


engine they were testing. As in this case, the
Figure 4.1.8 - Logistic of containers in external
radiator was placed on the roof of the container, court, close to control room
on one side there was the testing room, and on the other another container. The limit
required by the customer was of 85dB, while the measured value was of 90dB. Notice
that while I use dB as measured by the microphone, the following are all dBA, which
mean they are filtered to simulate the human ear behaviour.
The following excel paper was created by me for them in order to separate the sound
sources. “Background” refers to the noise of the ambient when all the system is switched
off. “Fans Only” is the sound measured with only the radiator running. The engine is
switched on and “Engine + fans” is obtained.

Figure 4.1.9 - Points where the noise was measured

56
Table 4.1.2 - Measured background noise with and without fans and noise
measured with engine on. A, B and C are used for logarithmic average
calculation (ISO 8528-10)

To obtain the final result, the logarithmic mean of the acquired data is evaluated
according to ISO standard 8528-10:

∑𝑛𝑖=1 100.1𝐿𝑝𝑖
𝐿𝑝 = 10 log10 ( ) (4.1.1)
𝑁

What emerges is that the main responsible for noise emissions is the radiator. Since it
was not possible to switch off the radiator with the engine working, the contribution of
the engine is evaluated considering that:
𝐿𝑝1 𝐿𝑝2 𝐿𝑝𝑛
𝐿𝑝𝑡𝑜𝑡 = 10 log10 (10 10 + 10 10 + ⋯ + 10 10 ) (4.1.2)

Knowing 𝐿𝑝𝑡𝑜𝑡 = 90,12 𝑑𝐵, assumed as contribution of fans noise 𝐿𝑝1 = 89,97 𝑑𝐵 and
of the engine noise 𝐿𝑝2 , this last can be calculated with:
𝐿𝑝𝑡𝑜𝑡 𝐿𝑝1
𝐿𝑝2 = 10 log10 (10 10 − 10 10 ) = 75,4 𝑑𝐵 (4.1.3)

A very low value, which they confirmed being a valid result.

57
Other considerations pointed out were that the presence of the
second container had an important consequence on the data
measured on its side. Looking at the point circled in green,
which are the one taken on the side of the testing room, and
comparing them with the blue, on the container side, it is clear
how these last ones are globally higher, even if the sound
distribution should be symmetric. The difference between one
side and the other was estimated to be 2,46 dB, in particular
the effect was higher on the ground (3,47dB) than at 5-6m
(1,46dB). This is explained by the fact that the containers are
very close one to the other (more than the case analysed
Table 4.1.3 - In green are before), and this effect is more evident at the bottom because
evidenced the lower values the radiator is higher than the second container and
measured on the right side of
the container; in blue there are surrounded by empty space.
the higher one on the left side
To solve the problem, they asked me to go to the radiator’s manufacturer in order to
check if they were able to reduce the noise emissions of at least 5dB. This result was
obtained changing the blades profile and stiffening the vibrating surfaces of the radiator.
The measurement obtained with the same phonometer, in the location of the
manufacturer are:

Table 4.1.4 - Noise comparison between Intergen data and those taken on
manufacturer place; values 1-8 were taken at the base: in second column
are kept the same of Intergen, in the third they are corrected by 2dB

58
The radiator was tested not above the container, but at ground level in the middle of an
empty square, for this reason only measure from 10 to 19 could be taken. The result was
good, thanks to the mean reduction of 5,7 dB. To estimate what could be the results with
the new radiator mounted, a realistic reduction of 3dB was assumed on the measured
values at the bottom of the container.

4.1.3. Change of load


The engines are tested at different load conditions, starting from 0% up to 110% of the
load, for time interval of different length at each level of load. It is interesting to see what
changes in term of sound between one condition and the other. It is expected that
reducing the load, also the noise would reduce consequently.
Engine running at 50% and 100% of the load is considered:

Graph 4.1.10 - 12th May, 50% load, position 1; rms Graph 4.1.8 - 12th May, 100% load, position 1; rms
of SPL=109,8 dB, max SPL=123,1 dB of SPL=112,8 dB, max SPL=124,7 dB

Graph 4.1.9 - 12th May, 50% load, position 2; rms Graph 4.1.7 - 12th May, 100% load, position 2;
of SPL=112,4 dB, max SPL=124,0 dB rms of SPL=113,8 dB, max SPL=125,7 dB

59
Graph 4.1.12 - 12th May, 50% load, position 3; Graph 4.1.11 - 12th May, 50% load, position 3;
rms of SPL=112,3 dB, max SPL=125,8 dB rms of SPL=114,1 dB, max SPL=125,7 dB

The differences between engine at 50% and 100% of the load are of few dB: in position
1 the delta is of 2.92 dB (for 100% load, sound power level is 2.92 more than at 50%);
in position 2 the difference is of 1.4 dB and in position 3 it is 1.82 dB. This is proved
also by other measures. These results are interesting: even if the number of samples is
not enough for a statistical analysis, the mean of these three values is 2,05dB: so if the
engine power doubles, also the sound power level, as average, is double.
To be kept in mind for later discussions, is that the range of frequency at which the noise
is higher is in the band 88-355Hz.

4.1.4. Transient
One phase of the engine difficult to measure are the transients: they are performed in
sequence by changing quickly the load of the engine and observe its response. The
system has to be able to recover from the instantaneous change of load quickly enough
in order to avoid damages. This mean that the engine has to be fast in reaching its nominal
speed of 1500 or 1800rpm. If the load is increased instantaneously, that is, from 0% to
100%, the risk is of being refused by the grid if the frequency are too different; on the
contrary, if the load is decreased, from 100% to 0%, the engine has to be fast, avoiding
to go overspeed or it can easily break, due to huge dimensions of pistons and rods. The
transients are monitored by the ECU of the engine, but their effect has repercussions also
on vibrations and sound.
They obtained this kind of graph (this is for the 100-50 jump), where the red curve is the
voltage measured by the sensor while the blue is the frequency, both function of time.
While the acceleration results increased with the reduction of load, the frequency after 4
seconds returns nominal (50Hz, 1500 rpm).

60
Graph 4.1.13 - Transient from 100% to 50% of the load

In the following the two cases of 0% to 100% and 100% to 50% are provided, represented
in time domain both with sound and vibrations. A STFT (resolution 10ms as time
interval, 4000 samples for frequency, amplitude in dB in order to better point out close
values) will help to see the evolution of frequency and amplitude in time.
The behaviour of the engine previously shown can be found also in these graphs, where
the load passes from 100% to 50%:

Graph 4.1.14 - Transient 100-50% measured by microphone; before cursor, RMS=6,173Pa;


after cursor RMS=4,674Pa (reduction of 2,4dB from 109,7dB to 107,3dB)

61
Graph 4.1.15 - Transient 100-50% measured by accelerometer; Before the cursor:
RMS=21,54m/s2; after the cursor RMS=23,12m/s2

The change in vibration is counterintuitive: a reduction of load lead to a slight increase


in vibrations. Notice that the velocity of rotation remains the same, and so does the
frequency of excitement. What is changing is the power of the engine, consequently the
vibration is expected to reduce.
Similar consideration can be done looking at the two STFT:

Graph 4.1.16 - STFT of vibration measured from 100% to 50% load; amplitude scaled with dB to have a
better colours definition

The STFT above is the one of vibrations: during the change after 12,5 seconds of load
there is a slight reduction of the vibrations frequency, which after few seconds return to
its previous values, while the amplitude remains more or less the same.

62
Graph 4.1.17 - STFT of noise measured from 100% to 50% load;

The second STFT is the one registered with the microphone: there is again a low
reduction of sound emission, mainly at lower frequencies. It is interesting the yellow
curve that is associated with an amplitude around 15Pa at very high frequencies. Notice
how the engine take 4-5 seconds to recover.
Similar behaviour is obtained for the change of load from 0% to 100%:

Graph 4.1.18 - Transient 0-100% measured by microphone; before cursor RMS=6,7Pa


(110,5dB), after 4,52Pa (107,8dB)

63
Graph 4.1.21 - Transient 0-100% measured by accelerometer; before cursor RMS=20,43m/s2;
after cursor RMS=25,9m/s2

As before, the sound is decreased of 3,41dB and the acceleration is increased of 5,47m/s2.
Of course, since the jump in power is higher, both values are increased consequently.

Graph 4.1.20 - STFT of noise measured from 0% to 100% load;

Graph 4.1.19 - STFT of vibration measured from 0% to 100% load;

64
The signature of the change of load is similar to the previous one, both for sound and
vibrations. Also in the STFT it is evident the passage between one load condition and the
other, as it is evident that the engine requires more time than previous case to recover its
nominal speed (5-6 seconds).

Last but not least, it is interesting also to see the behaviour of the engine while it is
stopping:

Graph 4.1.22 - Vibrations while the engine is stopping Graph 4.1.23 - Noise while the engine is stopping

The engine with all its rotating elements, has a great inertia so it takes a lot to stop
completely. The vibrations go to zero after 1 minute, while the noise that remain after
the same interval of time is the one associated to the fans of the radiator placed on the
roof.
In the STFT of vibration it is possible to follow the evolution of frequencies in time: after
10 seconds, when the engine is switched off, the amplitude of vibration at the beginning
(at 12 and 14 seconds) presents a kind of increase, which is the same found on previous
measures at 50% of the load, and then it decreases exponentially.

Graph 4.1.24 - STFT of vibrations while the engine is stopping

65
Components linked with combustion process (the higher ones in amplitude) reduce in the
first few seconds after the switch off. This is caused by some residual fuel in the lines
that allows some last combustion.
Therefore, what remains are only the vibrations associated with the inertia of the pistons.
Those are evident as thin yellow lines that tend to zero frequency: the engine is still
rotating but slowing down: the huge pistons with their rods are still forcing the structure
to vibrate but at lower and lower frequencies. This continues up to the point when the
velocity of the piston is so low that inertia is no more able to force the vibration of the
structure, and the thin yellow lines disappear.

Graph 4.1.25 - Focus on STFT of vibrations in 0-1000Hz range with stopping engine

Concerning the sound, it is more useful to focus on the range 0-1000Hz: the behaviour
is similar to the one of vibration: red lines, associated with combustion reduce and
disappear, leaving space to those linked to local surfaces still vibrating due to vibration
induced by the rotating elements.

Graph 4.1.26 - Focus on STFT of noise in 0-1000Hz range with stopping engine

66
The main difference with the STFT of the vibration are the two lines that remains
constant despite the switching off of the engine. Those are associated with the radiator
and the fan placed on the roof of the container, that remain on, since they are separated
from the engine. This help to individuate their main frequency at 232,5Hz and 597,2Hz.
Almost all the frequency that remain constant are associated with fans and air flows, like
the ones at very low frequencies).

4.2. Vibrations
4.2.1. Ambient conditions
Concerning vibrations, the main problem that can badly affect the measurement are the
resonance frequency of the point where the accelerometer is placed. As a matter of facts,
the excitation frequency of the engine covers a wide range of frequencies, multiples of
25Hz or 30Hz. But since the type of excitation is more like an impulsive one, the effect
is similar to hitting with a hammer a certain surface: the energy given to the element is
widespread across all the spectrum, but there will be higher peaks in correspondence of
resonance. For this reason, it is important to understand which are the pollutant
frequencies in the obtained spectrum and which are the interesting ones, that is, the one
responsible for noise emissions.

One way to avoid this problem, is to select properly the point of measure: for example,
the choice of an element which is stiff, rigid and well-fixed to the engine, so that its
resonance frequency will be higher than the one associated to the sound spectrum range
defined in the previous paragraph. Unluckily, this is not always possible. Three are the
main limits to this approach: the temperature, the material and the geometry of the
engine.

At the beginning, the vibration measurements were done by pasting the sensor on a tape
placed in the point of measure. The glue of the tape was not able to withstand the high
temperatures in some points, so that the accelerometer detached, ruining the
measurement in that spot. Moreover, some points were so hot that not only the
accelerometers are not designed to withstand those temperatures, but also there was the
risk of getting the hand burned.

It was then decided to use magnet: it was not the first choice because there was the risk
of ruining the painting, but placing it carefully this risk could be reduced. The problem
is that it is required a structure which contains enough iron to let the magnet to remain
attached. Many parts of the engines are made of aluminium or cast iron, and vibrations
were impossible to be taken.
67
Last but not least, the geometry: the accelerometers are axial, each of them get vibration
only in one direction. Consequently, to obtain a proper measurement along the three
dimensions, the point of measure should allow the disposition of the sensors along axis
x, y a z. This is not taken for granted, because if we consider also previous issues, the
adequate spots reduce a lot.

4.2.2. Statistics
Vibration measurement is more difficult than sound one, since each engine has its own
geometrical characteristics. Consequently, the exact point in which the sensors were
placed changed case by case, but they were always in the same areas, reassumed in the
following schema.

Figure 4.2.1 - 1 radiator, 2,4 engine supports, 3 engine stand, 5 alternator body

In following tabs are grouped the measured vibration in the indicated points (or nearby):

Table 4.2.1 - Vibration (rms) measured in point 1, 2, 3, 4 and 5 for various generating sets and statistical results
(mean, standard deviation, kurtosis, skewness)

68
The measurement taken from the engine were of course more than these. In this tab there
are those taken in the point previously indicated. The void spaces are due to different
reasons. For example, the engine tested the 3rd of May was the first one in which the new
instrument from Politecnico were used and there was not a strategy behind the
measurement. The last one, tried the 17th of November, was inside a container and there
was not enough space for movements. The 12th of June there was not the radiator. Other
blank points are due to wrong output from the accelerometer which was bad mounted or
other reasons.

The mean gives information about the common value of vibration in a certain area: as
could be expected, the measures on the support of the engine and on its base, are higher
than the ones on the alternator and on the radiator which are far from the engine block
and united with it by a rubber joint (alternator) and the base of the group, made of H-
shaped bars.
With a more general look on all the measurement obtained concerning vibrations, it is
possible to individuate the range of vibrations in which different Diesel engines for high
power production find place.

All the data collected were saved in an excel file. The measurement which were clearly
out of range due to errors in data acquisition or wrong mounted accelerometers, were not
considered. For example, the following time history, associated to measure along z, point
2 of the 28th of September, has values that are out of scale (the full-scale is 500m/s2):

Graph 4.2.1 - Example of vibration measurement where full scale at 500 m/s2 is reached

This problem was noticed when the jack was not well screwed on the accelerometer. By
strengthening it, the problem was solved. Looking at the data, the cut-value is imposed
above 110m/s2.

69
The distribution of the measured data is the following:

Graph 4.2.2 - Number of samples per each vibration Graph 4.2.3 - Gaussian PDF obtained for 132 samples
range, along x, y and z (132 samples) along x, y and z

More precise results can be obtained by looking at the single directions of vibrations:

Graph 4.2.5 - Number of samples per each range of Graph 4.2.6 - Gaussian PDF obtained for 40 samples
vibrations, along x (40 samples) along x

Graph 4.2.7 - Number of samples per each range of Graph 4.2.4 - Gaussian PDF obtained for 49 samples
vibrations, along y (49 samples) along y

70
Graph 4.2.8 - Number of samples per each range of Graph 4.2.9 - PDF obtained for 43 samples along z
vibrations, along z (43 samples)

The Gaussian curves are obtained with Matlab function y=gaussmf(x,[std(x)


mean(x)]: std(x) is the standard deviation of x while mean(x) is the mean. The
information contained in the graphs, are reassumed in the tab below:

Table 4.2.2 - Statistical results from vibrations taken on


generating sets in different points

The number of samples N is enough to make some considerations based on these


statistical results. Direction y has the higher values of vibration, followed by z and x.
According to theory this makes sense: vibration along x axis has not an important source
of excitement apart from the fan. Being engines with 16 cylinders they are well balanced
along z axis, and less along y axis. Moreover, FEM analysis developed on similar
engines, shows how the main excited mode on industrial engines is the torsional one,
which, depending on the shape of the engine, will excite more the y direction [20] [21].

Standard deviation is quite high due to the fact that these results are general, obtained on
the basis of different kind of engines of different power, dimension and geometry.
Moreover, the point of measure changes at each measurement, so it is impossible to get
a stricter range of validity for vibrations entities.

In skewness there are neither negative values nor ones close to zero, this is caused by
some spare values which were not filtered (the maximum level for the filter was set at
110m/s2), but are far from the zone where all values are concentrated.

71
Since the curve is centred in the mean value, the correct value for kurtosis is 3: this means
that measures of y and z are close to normal distribution while x is not.

4.3. Noise and vibration frequency analysis on engines


Each engine has its own signature which depends on many factors: the most important
are its dimensions, the manufacturer but also the point of measurement. For each engine
the acquisitions were taken always at the same height and in the same points. In order to
compare different models, three areas are chosen, common for each engine (figure 4.3.1):

Figure 4.3.1 - Points of measure for sound

4.3.1. General frequency domain behaviour of engines


The one third octave band graphs are plotted below: they belong to 5 engines tested
during this year. They are not only in chronological order, but also according to their
dimensions and power. Moreover, the first two, which are the smaller, run at 1500rpm,
while the others run at 1800rpm. In the measure of the 29 May the microphone had not
the coffin, so there is a disturbance at low frequency due to air. This is evident mainly in
the first graph, since the measure was taken close to the radiator, where the wind effect
is more evident.

Graph 4.3.1 - 29th May, meas. 2, 4 and 6

72
Graph 4.3.3 - 6th June, meas. 1 and 3

The following engines are more powerful and run at 1800rpm:

Graph 4.3.4 - 25th September meas. 2, 4 and 5

Graph 4.3.5 - 28th September, meas. 1, 5 and 8

Graph 4.3.2 - 2nd October, meas 3, 9 and 11

The first thing to notice is that the range of frequency at which the noise is emitted the
most is below 710Hz, and this is true for all the engines. This is due to the forcing
frequency of the engine which are linked with the firing order, the inertial force of the
pistons and the combustion which depends on the velocity of rotation (1500 or 1800rpm).

73
The increase of the frequency is associated with a reduction of the amplitude of the
harmonics. There is also a peak which is higher than the other: as it will be pointed out
later, it is associated with fluid dynamic noise of the fans. It is also interesting to notice
as the sound emissions of the first two engines, that are smaller and slower, are lower
than the ones of the bigger and faster ones (max peak of 0,25Pa for the smaller and 0,35
Pa for the larger, within the range of 88-177Hz). It is then possible to proceed with a
detailed analysis, focusing mainly in the range 0-1000Hz and later add some
considerations for higher frequencies.

4.3.2. Detailed analysis of spectrum


Before showing the results, some considerations should be done:
1) On the point of measurement, the surface is excited by the forcing of the engine.
The frequency can be far from the resonance, so the amplitude of vibration is low,
or vice versa, it is at resonance and consequently the amplitude will be high. This
concept is important, because depending on the point of measure, the result will
be different: for example, a small plate will be excited at a certain frequency at
which it emits sound, but the volume will be low and acute. On the contrary a
larger plate will lead to higher volume but at lower frequency. Keeping this in
mind, we can go on with the analysis.
2) The engines have 16 cylinders: those types of engines are very well balanced,
whichever is the V angle of inclination. Moreover, the impulse is given every
45°, which is a good solution for engines with an even number of cylinders.
This is the spectrum up to 1000Hz of one of the engines tested:

Graph 4.3.6 - 28th September, sound 16

74
Air flow effects
At the lower frequencies in the range 0-40Hz it is possible to notice how the sound seems
to be disturbed: this is caused by the fact that the measure was taken close to the radiator,
where the wind blows on the microphone, and despite the presence of the coffin, a part
of this wind induced noise is registered, together with its harmonics. This was found out
by comparison between different measures on different engines, always in conditions in
which the microphone was invested by an air current.

Graph 4.3.7 - 28th September, sound 16, zoom on lower frequencies

Engine excitation
Considering the measured vibrations (first graph below) and the registered sound (second
graph below, taken in absence of wind so that peaks are not covered by any disturbance),
it is clear how they are multiples (0.5/1.0/1.5/2.0/… Hz) of 30Hz.

Graph 4.3.8 - 28th September, vibration 9 4z

75
This is explained by the fact that the engine has 16 cylinders, 8 for side. If it run at
1800rpm, the frequency is 30Hz. It is a four strokes engine, so there is firing each two
rounds, with a frequency of 15Hz.
Within 720°, the firing order is such that the combustion occurs each 45° (720°/16
cylinders): having 16 combustions each 2 rounds, 8 per round, the frequency will be
8*30Hz=240Hz: in the range 30-240Hz there will be a peak each 30Hz due to excitation
given by combustion.
Moreover, it will be noticed a peak at 30Hz due to the movement of the piston at each
cycle, due to inertia: the piston is moving up and down each entire round: this component
is linked just to the inertial force, since combustion occurs each 2 rounds. Since there are
16 cylinders moving up and down each 2 rounds, due to inertia there will be
16*30Hz=480Hz and all the intermediate frequency, multiple of 30Hz due to inertia: 30,
60, 90… Linked to all the possible combination of pistons movement.

Graph 4.3.9 - 28th September, sound 17


In these large dimension engines, the contribution due to inertia is secondary with respect
to the one due to combustion, moreover, literature shows many tests performed on similar
engines of different manufacturers, pointing out how the vibrations are mainly torsional
[20] [21]. Since combustion occurs each 2 rounds (being a 4 strokes), there will be a peak
at 15Hz, but since there are 16 combustions each 720° (which means each 45° of the
crankshaft) there will be a peak at 240Hz. However, there will be all the intermediate
frequencies multiple of 15Hz since the system feels also all the possible combination of
other power strokes. Contribution of combustion is higher than the one of inertia, and
often they overlap. The same can be said for engines running at 1500Hz, where the
excitation frequency is 25Hz and its multiples:

76
Graph 4.3.11 - 12th June, overall vibration in point 1

Graph 4.3.10 - 12th June, sound in point 1

Fans and radiators


Different is the origin of the higher peaks that appear in the first graph of sound: there
are peaks at 161,9Hz and its harmonics, and they are very high in pressure level

Graph 4.3.12 - 28th September, sound 16


77
But there is no sign of important peaks at the same frequencies in the spectrum of
vibrations, apart from the radiator:

Graph 4.3.13 - 28th September, vibration 16 5y, far from radiator

Graph 4.3.14 - 28th September, vibration 1 5y, close to radiator

They are found also in other engines, more or less at the same frequency. Below one
graph of sound and two of vibration, dated the 2nd of October are provided. Looking
carefully at the vibration ones, for the measure taken on the radiator there is also a peak
at 161Hz, as in the sound. But the other has not this component, since the measure is
taken far from the fan.

78
Graph 4.3.17 - 2nd October, sound 1

Graph 4.3.16 - 2nd October, vibration 1 4y, on radiator

Graph 4.3.15 - 2nd October, vibration 9 5y, far from radiator

79
Again, the same behaviour, but at frequency slightly different (153.9Hz and harmonics):

Graph 4.3.18 - 25th September, sound 13

Graph 4.3.19 - 25th September, vibration 1 4y, on radiator

Graph 4.3.20 - 25th September, vibration 11 5y, far from radiator


80
It is present also on the engine tested outside: here the radiator is placed on the roof of
the container, and it is not made by one big rotating fan but by three small fans. Therefore,
their presence is at higher frequency (the fans rotate faster) but with lower amplitude
(smaller fans), as their peak at 297,5 Hz emerges above the others:

Graph 4.3.21 - 2nd October, noise from fans on the roof

The first measure, taken the 3rd of May, was on a smaller engine that had a fan with 18
blades. Since the noise depends on the number of blades, here it is its spectrum, where
the radiator emits at high frequency (476,6 Hz):

Graph 4.3.22 - 3rd May, fan noise

Using the formula 2.4.9 for the BPF, it is found that the rotational speed of the fan was
2231 rpm (8 blades, 297,5Hz) and 1588,7 rpm (18 blades, 476,6Hz) respectively.

81
On the contrary there is no sign of fans peak when there is no radiator: it is the case of
the engine tested the 12th of June: it was tested inside the testing room, but the radiator
was placed outside the building:

Graph 4.3.23 - 24th October, open container 12

The same engine can be used for a comparison: this last one (represented by the blue
line) is slightly bigger than the other (red line) and has no radiator. They both run at
1500Hz and for the previous reasoning the peaks are each 25Hz and multiples
(0,5/1,0/1,5... Hz).

Graph 4.3.24 - Comparison engines 29th May (sound 3 right) /12 June (sound 100% 2)

Three important differences emerge from the graph:

1) the blue one, being larger, has also higher amplitude of sound pressure (the two
peaks before 200Hz are probably linked to the resonance of the point of measure,
since they are harmonics of the elementary frequency at 25Hz);

82
2) the red has peaks at 162,9Hz and multiples (325,8 Hz, 488,7 Hz) that are not
present in the blue one.
3) the red measure is characterised by a more important disturbance which covers
some peaks: as previously mentioned, this effect is caused by the air flow on the
microphone, since the measure is taken close to the radiator.
Evidences lead to conclude that the components with high amplitudes are linked with the
radiator noise. Moreover, due to the absence of evident peaks at the same frequencies in
vibration spectrum far from radiator, while there are some signs of vibration on radiator
surface, it is possible to say the noise has mainly fluid dynamic nature.

4.3.3. Comparison between vibrations spectra


The measurements below were taken on the front and rear supports of the right side:

Graph 4.3.26 - 2nd October, vibration 4

Graph 4.3.25 - 2nd October, vibration 6

83
Vibrations have the same order of magnitude and more or less the same profile. But while
in measure 4 the x direction has relevant component only at low frequencies and then it
disappears, in measurement 6 on the contrary it has higher amplitudes also at higher
frequencies.

Vibrations along y and z on the other side have similar entities in both the measurement
points, where vibration in y direction prevails above the ones along z.

There is also an interesting peak at 150 Hz which is present in measure 4 but not in
measure 6: this can be an evidence of how the local resonance of the plate affects the
measurement.

4.3.4. Comments on spectra of sound


Together with vibrations results, it is important also to give the ones obtained by sound
measurement. For example, as already mentioned the plates on which the accelerometer
are placed for these last measures have their own behaviour and ways of vibrating which
can exalt or smooth the actual vibration entity, also because they can go into resonance
with other components of particular shapes and geometries. Comparing vibrations with
sound taken close to the accelerometers, it is possible to understand which are the actual
frequency of vibration that give the most important contribution to noise. As a matter of
fact, there could be a large surface which is vibrating less than a smaller one, but which
will emit louder sound due to its dimensions.

Graph 4.3.27 - 2nd October, sound 6

Sound spectrum is not as regular as the one of vibration. They are clear the already
mentioned peaks at 161 Hz, 322 Hz e 483.2 Hz linked with the fan. The predominant
peaks linked directly to the engine are the ones at 75Hz, 135Hz, 240Hz, 315Hz and others

84
with their harmonics, that are present in the vibration spectrum, but do not emerges
particularly.

4.4. FRF between noise and vibration


For a better visualization of the vibration effects on the sound, it is possible to think about
the displacements imposed by the engine on its surface as the input to the system “air”
which generates as output the sound pressure waves, which are perceived as sound. It is
then possible to analyse in the frequency domain the Frequency Response Function
(FRF): it allows to point out which are the main inputs that have an effect on the outputs.
The conclusion obtainable will be a confirmation of what is said up to now.
This type of analysis requires that
both sound and vibration
measurements are taken together, in
the same time instant, so that all the
data can be perfectly overlapped at
the correct frequency. This was done
for the first time the 2nd of October,
so it is possible to focus on this case.
The measurement of sound and
Figure 4.4.1 - 2nd October, accelerometers and microphone
position for measure 9 and 10
vibration are taken in the point
indicated in the pictures.

Notice that having only two accelerometers, two measures were required (the 9 and 10)
in order to get the three dimensions x, y and z. This can lead to partial incongruences,
being the measure along x (number 10) taken separately from y and z. Knowing this fact,
the choice of the measure to take alone fall on x because it is the one with lower
amplitudes of vibration. A second consideration, already mentioned many times, is that
the point where the accelerometers are placed (a squared and thick plate of metal), will
have its own resonance frequency, which will result amplified. Anyway, what is
interesting is to get in general which are the frequency generating the louder sound
concerning the entire engine.
FRF is evaluated as:
𝑜𝑢𝑡𝑝𝑢𝑡 𝑠𝑜𝑢𝑛𝑑
𝐹𝑅𝐹 = = (4.4.1)
𝑖𝑛𝑝𝑢𝑡 𝑣𝑖𝑏𝑟𝑎𝑡𝑖𝑜𝑛

85
For the FRF analysis two ways are possible: the first is to study the effect of each single
direction of vibration, the second is to evaluate the resultant of the vibration’s
components. Also in this case both path can be followed.
In the following two pictures there is the spectra of sound and of the resultant of vibration,
measured up to 5000Hz:

Graph 4.4.2 - Overall vibration obtained as 𝑉 = √𝑥 2 + 𝑦 2 + 𝑧 2 , whole spectrum

Graph 4.4.1 - Noise, whole spectrum

For this particular engine, the frequencies at which the sound contribution seems to be
interesting go up to 1500Hz (also for vibration, but this is not relevant since it can depend
on the plate behaviour with vibrations).

86
Concerning the FRF it is possible to compare sound and vibration directly or using
normalization. In this case I divided each array of amplitudes for its maximum value, so
that the lower and less significant results are filtered out both in vibrations and noise.

Graph 4.4.4 - FRF: most of lines are linked to base excitation frequency; in 750-1250Hz range there is
overlapping. It will be shown later that it is turbocharger operation range

Graph 4.4.3 - Normalized FRF gives clearer spectrum: peaks in 750-1250Hz range disappear, since
they are present as sound, but not as vibration. Order of magnitude is the same

In this wide range the results seem to be similar. Focusing on a more restricted range of
frequencies, for example 1500Hz (in order to well distinguish the different peaks), the
obtained normalized FRF gives a clearer idea of the vibration effects on the sounds. As
a matter of fact, it can happen to have high vibration in the point of measure which are
not emitting a significant level of sound. Without normalization, the effect of these
vibration would lead to wrong interpretation of sound emission. Below the interesting
ranges 0-1500Hz are plotted.

87
Graph 4.4.5 - Focus on range 0-1500Hz for sound. Frequency peaks from fan are evident

Graph 4.4.6 - Focus on range 0-1500Hz for sound

Graph 4.4.7 - Normalized FRF. A kind of periodicity pf peaks seems to be present

88
These last three graphs confirm what was previously said about sound sources of the
engine. It is well clear how sound is emitted mainly at frequency which are multiples of
15Hz, that is, it is linked to motion and combustion inside engine’s cylinders. In order to
better understand the effects of the sources, each vibrating surface should be measured
in order to get its vibrating frequency and analyse its contribution to the sound. It can be
taken as assumption that smaller surfaces emit sound at lower volume.

Moreover, the graphs of the FRF point out how the principal source of sound emission,
the big fan of the radiator, has not any correspondence in the field of vibrations, because
there are no peaks at 161,1 Hz and 322,2 Hz. Actually, looking at the inverse of the FRF

Graph 4.4.8 - Inverse of the normalized FRF (vibrations/sound)

and zooming, two small points emerges at those frequencies, meaning that the noise is
mainly of fluid dynamic nature.

Graph 4.4.9 - Fan components are present, their link with vibration is small but still present

It is then possible to act on the spectrum of sound in order to separate the components of
the fan from all the others. It is possible to take the engine tested the 2nd of October, when
the measure of sound and vibrations were taken contemporaneously. The following two

89
graph shows the spectrum of the measures taken one close and the other far from the
radiator.

Graph 4.4.11 - 2nd October, sound 4, 119,2 dB Graph 4.4.10 - 2nd October, sound 9, 118 dB

This is what is obtained by separating radiator contribution from the others:

Graph 4.4.12 - Sound 4 close to radiator (black), Graph 4.4.13 - Sound 9 far from radiator (black),
ventilation fans (red), alternator fan (blue) ventilation fans (red), alternator fan (blue)

The frequency here presented are 161,2 Hz, 322,3 Hz, 483,4 Hz, 644,5 Hz, 805,6 Hz,
966,6 Hz for the radiator; 194,9 Hz, 384,8 Hz, 779,7 Hz, 974,6 Hz, 1170 Hz for the fans
of the testing room; 314,9 Hz, 622,9 Hz, 944,8 Hz, 1259,6 Hz for the alternator’s fan.
The first harmonic is the one that feel the most the distance from the point of measure,
both for radiator and testing room fans and the alternator, with the only difference that
measure 9 is close to it, while 4 is far, so the reduction is inverted.

90
Graph 4.4.15 - 2nd October, sound 4 filtered by Graph 4.4.14 - 2nd October, sound 9 filtered by
aerodynamic noise aerodynamic noise

The difference between point 4 and point 9 is clearer considering that the change of noise
from the room’s fans remain more or less on the same level, and helps at the first view
to understand the difference between the two measures. In any case it is evident that the
emissions of noise are mainly linked to aerodynamic reasons.

The filtered spectrums are more similar than before, also, the level is sound is closer too.

4.5. Higher Frequencies


Sound emission at frequency higher than 1500Hz is not comparable to what has been
found to be the main responsible for sound emissions, that is, the aerodynamic noise of
fans. It is anyway interesting to give some brief considerations about what was
measured.

Effect of resonance
Vibrations are present also at high frequencies, while sound is not. The sources of high
frequency vibrations are elements running at high speed, as the turbocharger. If the
accelerometer is placed on a stiff surface, its resonant behaviour may affect the
measurement.
Though, high peaks in amplitude were found in some measurement: it is the case of the
engine tested the 28th of September: here, among all peaks at high frequency, there is
one which stands out more than the others:

91
Graph 4.5.2 - High vibration amplitude along z Figure 4.5.1 - Accelerometer along
direction y and z on the front plate

Graph 4.5.1 - Secondary vibration amplitude along y Graph 4.5.3 - Secondary vibration amplitude along x

As it is possible to see, this component is always present, but in direction z it has an


order of magnitude more than in direction x and y. Considering measure taken in other
points, for example the other specular plate:

Graph 4.5.4 - Very low vibration amplitude along z Figure 4.5.2 - Accelerometer along
direction y and z on the back plate

Graph 4.5.6 - Very low vibration amplitude along y Graph 4.5.5 - High vibration amplitude along x

92
On the contrary it is not present on other points (only direction z provided):

Graph 4.5.8 - Small vibration along z


Figure 4.5.4 - Position of the
accelerometer on the radiator

Graph 4.5.7 - Negligible vibration along z


Figure 4.5.3 - Position of the
accelerometer on the engine block
Moreover, the measurement taken with the microphone close to the plate presents a
slight peak, while those taken far from it does not:

Graph 4.5.10 - Sound measured close to plate


Figure 4.5.6 - Microphone close to
plate

Graph 4.5.9 - Sound measured far from plate Figure 4.5.5 - Microphone position
far from plate
These measurements lead to conclude that for that particular frequency the plate where
the accelerometers are placed goes in resonance along z direction. The frequency of

93
vibration is the 96th harmonic of the base frequency, 30Hz (slight error due to probable
wrong calibration of the accelerometer, 2882/30=96,07).

Turbocharger noise
Apart from small rigid surfaces vibrating at higher frequency than larger and thin ones
(like the casing of the radiator or the alternator), the other possible source of high
frequency sound emissions is the turbocharger. From publications [12], it emerges that
the frequency at which the turbocharger emits higher than 800Hz. The sound emissions
of the turbocharger are driven by the flow toward the exhaust gas, pipes and muffler
which goes outside. In most of the noise data collected, the amplitude at these frequencies
are of secondary importance: from the large quantity of measures taken, there are no clear
and univocal clues that lead to the determination of turbocharger noise components.
In an attempt to identify turbocharger, contribute, a measure with the microphone was
done above the engine:

Figure 4.5.8 - Position close to fan


Graph 4.5.13 - Noise spectrum between intake and fan on the right of intake system

Figure 4.5.7 - position between


Graph 4.5.12 - Noise spectrum between turbochargers turbochargers and intake manifold

Figure 4.5.9 - position to the left of


Graph 4.5.11 - Noise spectrum on the left of turbochargers intake system

94
The graphs are all scaled so that the maximum value is 3Pa. The first graph is associated
with a measure done close to the fan, consequently the noise is emitted broadband due to
air passing close. The second measure is taken exactly between the two turbochargers
and the air filters, while the third is on the left of them. The second spectrum shows slight
higher peaks in the range 900-1500 Hz, which is the range of supposed emission for
turbocharger: it is reasonable to associate this part to the air intake and the compressor-
turbine system.
An information missing and of fundamental importance for at least determining the area
of the spectrum associated with turbocharger, is the velocity of rotation, that is not always
given by the manufacturer.

Whistling and howling have origin in the mechanical behaviour of the rotor. In particular,
whistling comes from unbalance of rotating parts, but its effect, it is said, is amplified by
the components of the exhaust systems which are placed out of the testing
room/container. It is expected to be the responsible of the sound emitted in the evidenced
range.
Alternator
Despite the many measurements taken on the alternator, of both noise and vibrations, no
univocal clue leads to associate to this element any peak on the spectra. As literature
suggests, the noise (and vibration) emission are strictly linked with its rotational speed
which is the same of the diesel engine, and more or less at the same frequencies.
Consequently, alternators contributions are covered by the stronger one of the engine, so
that nothing sure could be said, apart from the overlapping between the two.

95
5. Conclusions

Results presented in this thesis evidenced that the analysis of the time histories, spectra
and STFT allowed obtaining useful information concerning noise and vibrations. The
measurement procedure accomplishes the ISO 8528-9 and 8528-10, so that the
comparison of results was easier and direct.

The main goal of the work was obtaining a general overview on the different generator
sets manufactured by Intergen. The work was divided in three main parts, that are the
analysis of noise, vibration and vibro-acoustics (noise+vibration). Measurements were
analysed both in time and frequency domain. The different noise and vibration sources
were identified, and their contribution to the overall level was analysed.
Despite the large amount of vibration sources, the dominant vibration is the one
synchronous with the engine rotation speed. The vibration spectra show relevant multiple
and submultiples of 25Hz or 30Hz. The maximum vibration was 30-40 m/s2 on all the
groups some higher values are present, but they depend on local resonance of the point
chosen. The main direction of vibration propagation was along the y axis. No main
excitation along x axis exists, apart from the radiator fan, consequently x direction is the
less critical one. Also, it has been noticed that the power of the engine has not a particular
effect on vibration measurements, that is, neither higher nor lower values were measured
for engine of different power.

Concerning noise, the louder element is the fan of the radiator, when present. Its
dimension, shape, and rotation speed entail a large air flow in highly turbulent regime.
The noise often exceeds 110 dB with peaks close to 115dB. The noise is not entirely
generated by the fan, as the interaction air-radiator leads to vibrations generating noise
with typical turbulent layer characteristics. Again, another major contributor is the Diesel
engine: large flat vibrating surfaces have a very large radiation efficiency, that together
with the combustion noise and air movement lead to sound pressure levels above 105dB.
The noise and vibration originated by the turbocharger and the alternator were minor. On
the turbocharger it wasn’t possible to take vibration measurements due to temperatures
and the positioning. Nevertheless, despite the negligible contribution to overall noise
emissions, turbocharger’s frequencies were detected in the range 900-1500Hz, as theory
suggest. Nothing in particular could be said about the electric generator: literature says
that range of noise emissions are synchronous with the velocity of rotation both for
mechanical and electrical noise. Since also the engine contribution occurs at harmonics
multiple of its rotational speed frequency, which is the same of the alternator, it covers
96
all the other possible peaks. Due to overlapping it is then difficult to detect the vibration
and noise originated by the generator. Also, from measurements taken close to alternator,
both of sound and vibration, nothing particularly different from what was observed
elsewhere was noticed on the spectrum silhouette. According to experiments conducted
in the past [14], the noise emissions from generator should be around 80-90 dB, which
is of secondary importance with respect to other sources.

A general overview of the background noise was also performed. Indoor, in the testing
room, the contribution of ventilation fans in the location of engine noise measurement is
important (95 dB).
Concerning noise measured around the container tested outside, it was demonstrated the
effect of nearby obstacles on sound measurement. On both considered cases the measure
was affected by 2-3 dB. Soundproof containers are well designed, since they lead to a
noise reduction which is higher than the recommended 10dB: measured reduction was
around 15-20dB.

Even if vibrations are within the limits, further reduction may be obtained by stiffening
the components that are close to resonance at the frequency of excitement. By increasing
stiffness, the peak of resonance is moved toward higher frequencies, far from the main
range of excitement. For example, it is the case of the radiator casing, which is a thin
plate, excited both by the fan and the engine: including a transversal bar will further
reduce both noise and vibrations.
It is suggested to correct all the future sound measurement taken in the testing room, by
eliminating from the reading of the phonometer the sound emitted by fans. Even if noise
measured indoor is quite always in the limit allowed by customers and standards, this
correction will allow to get even better results.
On the contrary, to avoid exceeding the limit imposed by standards and by final customer
outdoor, it is highly recommended to place the container far from huge obstacles such as
other containers or sheds. If this can not be done due to comprehensible logistic and
practical reasons, it must be evaluated the correction of noise on the side with
reverberations by considering the side when the propagation is similar to the free field
one. The reduction can be up to 3dB, depending on the case. Further investigations are
needed in this field and will be the topic of forthcoming studies. Setbacks as the one
occurred with the radiator placed on the roof may be reduced.
In any case, to cut the noise emission for outdoor configurations, it should be always
considered to act on the radiator placed on the roof, being the dominant noise source: as
previously mentioned, the container is very effective in noise insulation. As it was done

97
by the manufacturer, the operation of correction in case of high emissions, are to stiffen
the surfaces of the system to get away from low frequency resonance, and change the
geometry of the blades. If this was not enough, consider insulating with soundproof
panel: though, this is quite complicated, since it must be always allowed passage for
cooling air.
Finally, the use of noise shields preventing from a direct transmission roof/listener could
also be considered. All the results and the considerations are valid for a wide combination
of engines, power, dimension and manufacturer

98
6. Appendix A: Matlab codes description

Matlab has been widely used for many purposes. The total lines of codes are more than
one thousand, spared between many programs and functions. Here some of the main
ones are provided.

6.1. Intergen_Matlab - Sound


It takes as input an Excel file: for the sound, there are four columns (time, pressure(t) in
Pa, frequency, pressure(f) in Pa) for each point of measure. Since putting everything in
a unique Excel file requires too much time with Matlab, so it is more convenient to keep
them separated (one Excel file for each point of measurement). It converts everything in
dB: the vector SPLrms is defined as (Pref=2e-5 Pa):
𝑃𝑎𝑡 2𝑟𝑚𝑠 𝑃𝑎𝑡 𝑟𝑚𝑠
log10 ( 2 ) = 20 log10 ( ) (4.1)
𝑃𝑟𝑒𝑓 𝑃𝑟𝑒𝑓

For the evaluation of the Root Mean Square, the RMS function of Matlab is used: it gives
as many outputs as are the inputs in Pat. Using N=2 in evaluates the global RMS of the
signal, RMSP, used to evaluate the final sound pressure level in dB of the signal. Finally,
it evaluates the value in dB also in frequency domain (vector SPLrms_freq defined as
SPLrms but it uses Paf).

Then it evaluates the characteristic magnitudes of the measurement (time resolution,


period, sampling frequency, frequency resolution and aliasing limit frequency).
It is also possible to evaluate the filtered sound in dBA using the function filterA (see
function below) and the Power Spectral Density using the function periodogram (see
function below). Being a large amount of data, it is convenient to reduce them by using
the function rms_reducer (see function below).

6.2. Intergen_Matlab - Vibrations


It takes inputs from three Excel files, containing four columns (time, amplitude(t),
frequency, amplitude(f)), correspondent to the measurement of the accelerations in the
three directions x, y, z.

99
Due to the fact that the length of the uploaded vectors may be different, the program
checks the lengths of the acquired measurements and cuts the data of the longer vectors
to pair them with the shorter.
As before, time resolution, period, sampling frequency, frequency resolution and aliasing
limit frequency are evaluated. Moreover, for each direction x, y, z, the program prints
the mean of the vibrations, peak value, true and RMS peak-to-peak amplitude, RMS,
standard deviation and crest factor. Notice that being the mean around zero, RMS will
be equal to standard deviation.

The output graphs give different information: as before the PSD, the reduction of data in
time domain and in frequency domain (using Welch).

6.3. Functions
[] = filt_octave(x,f,fs)

It takes as input x, f and fs: x is the signal Paf (pressure in Pa in frequency domain); f are
the values of frequency correspondent to each value of Paf; fs is the sampling frequency.
It defines fxx as the module of x; the central frequency and the boundaries of each
frequency band are also set; with a for cycle the number of bands to be plotted are counted
as function of the sampling frequency: it takes a range of frequency which arrives at
maximum at the limit aliasing frequency and on the basis of this there will be a certain
number N of bands covering that range.

norm is a normalization coefficient, defined as the ratio between 1 (default delta Hz in


freqrange) and the frequency resolution (interval between two values of frequency in the
signal).
A vector band is created: it has N values, equal to the number of bands, and each value
is defined as the RMS (or the max) of each frequency band of the signal. The
normalization coefficient is used to take the correct amount of frequency values of the
signal. For aesthetical reason, the band’s length is increased by one, keeping the last
value equal to one.

For example, the range of frequency 11-22 Hz contained in freqrange, has 22-11=11
values, but maybe the original signal has 100 values between 11 and 22 Hz. It is defined
norm as 1/frequency resolution of the signal, so that it is taken 1*norm values of f for
each single value of freqrange.

Finally, the program plots the band graph: everything is placed in an if cycle that controls
which bands to plot (usually max frequency of the signal is 5120 Hz, so the band limit
100
will be 5680Hz). It is used function bar to plot the band values, axis([XMIN XMAX YMIN
YMAX]) is needed to define the maximum and minimum values of the axis; set defines
what to write on the axis and at which distance).

[xA] = filterA(x, fs)

It takes as input x which is pressure(t) measured and fs which is the sampling frequency.
It evaluates the FFT to bring it to the frequency domain and takes only one half, being
symmetrical.
The five coefficients for the filter are experimental: c1, c2, c3, c4, c5. They are used to
create the A-filter in the frequency domain, defining vector A with the same length as the
input signal vector X in frequency domain. To apply the filter, it is enough to multiply A
times X.
At this point the entire spectrum is reconstructed symmetrical using complex conjugate.
With the inverse Fourier transform (IFFT) everything is brought back to the time domain
and with the function reshape(new signal, size(old signal)), it is shaped as the original one.

The final output is the signal in time domain, expressed in Pa and A-filtered.

[Lp_avg] = logavg(L)

It takes as input a vector L of sounds expressed in Pa and evaluates the logarithmic mean
according to ISO 8528-10 par 13.2.

[] = plotengine(N,dim)

The input N is the resolution (number of point for each segment, use for the cylinder)
and a vector of engine’s dimensions (dim=[engine’s length, engine’s width, engine’s
height, radiator’s width, radiator’s height, radiator’s length, generator’s length,
generator’s radius, base’s width, turbocharger’s radius, turbocharger’s width]).
Everything is modelized with simple figures (cylinders and parallelepiped).
On the basis of data contained in dim, it evaluates start and the end of each edge of each
figure, as a segment which goes from d0 to df with N points. For example the engine
length is:
∆𝑙𝑚
𝑙0𝑚 = 0 − (6.3.1)
2

101
Which means that the starting point for the length of the engine is 0 minus half of the
length so that the engine middle point will be centred in zero, and the other extremity
will be at

𝑙𝑓𝑚 = ∆𝑙𝑚 + 𝐿0𝑚 (6.3.2)

Plot3 uses three vectors, two of which have constant values: O is a vector of “ones” time
the value of start/end of the coordinate range between which it is defined the segment;
the third contains the vector which defines the segment. Assuming plot3(x, y, z), x is the
length, y the width and z the height. The negative values are used because everything is
wanted to be symmetrical with respect to the axis. For the cylinders, the circles at the
extreme are plotted and then linked with a series of segments.

[tpart SPLpart Papart_rms] = rms_sound_reducer(Pa_t Nrms t P_ref)

It takes as input the pressure value in time domain Pa_t, the number of samples Nrms to
which the signal has to be reduced, the time vector t and the reference pressure P_ref.

The amount of data is reduced dividing the pressure vector into N sub-vectors, and for
each of the it is computed the RMS. The resulting vector is N long and contains the RMS
values of each interval. The output is the vector tpart containing the time instants
(evaluated as the mean between the two extreme values limiting each interval), the vector
Papart_rms containing the reduced number of values in Pa and SPLpart containing the
same but in dB.

rms_vibration_reducer works in the same way but it is for vibrations.

[] = plotsound(dimx,dimy,risoluz,N_quadrati,sound)

It takes as input the dimension of the measurement grid used for turbocharger along the
axis, dimx and dimy, the resolution, the number of squares (N_quadrati=8) along y
direction, to which correspond N-1 points of measurement along y and N-2 points along
x. sound is a matrix dimx×dimy containing the RMS of the measured values in each point
of intersection of the grid.
The first step is to evaluate the number of lines to plot for the grid along x and y, risox.
Then it evaluates delta_grid used to define the distance between one point of intersection
and the other, and then plot the coloured squares (it is the distance between the centres
of the squares).

102
The vectors x and y are created and then inserted in meshgrid to create the coordinates of
a rectangular grid. Z is defined as empty because it is not needed yet. C is initialized with
ones: it is the matrix of colours, and since 1 correspond to blue, doing so, a blue
background is created. The vectors a and b are used to localize the points of interception
though a for cycle, they are redefined each time by jumping of a quantity equal to
delta_grid previously defined.

BB is the matrix which refers to the coloured squares around the interceptions: it is
defined by overlapping four squared sub-matrices having delta/2 dimension and
containing the linear values in the range [1, 2] or [2, 1]. These sub-matrices (BB1, BB2,
BB3, BB4) are a kind of quarter of square, containing the colour gradations (1=blue,
2=yellow) visualized on the grid. They are used to assemble the matrices BBA and BBB.
BBA are vertical colour bands on the left and on the right: BB1 and BB2 are symmetrical,
since BB1 goes from blue to yellow from left to centre, while BB2 goes from yellow to
blue from centre to right. BBB are the horizontal lines, and BB3 and BB4 contains the
horizontal colour bands. BB is finally defined as an average between matrices BBA and
BBB. The plotted result will be a coloured square, with gradation passing from yellow to
blue going far from the centre. BBB has dimension delta×delta.
Matrix BB is inserted in C (which contains the blue background) which is cut in order to
give to it a rectangular shape (7x6) and not a squared one.

[P,F]=periodogram(Pa_t,wind,NFFT,fs)

The inputs are the pressure signal in time domain, the type of window and the sampling
frequency and gives as output the PSD. P is the estimation of the PSD, F is the vector of
normalized frequencies at which the PSD is estimated. If NFFT is even, the frequency
are in the range [0, fs/2], if it is odd the range is [0, fs/2). Default window is rectangular:
it is possible to change it, for example writing Hann to have Hanning windowing.

[pxx,w] = pwelch(Pa_t,length_vect,noverlap,N,fs)

It is used to evaluate the PSD with Welch approach and reduce the amount of data (like
the previous function rms_reducer). It takes as input the signal in time domain, the length
of the window vector (each column of Pa_t is divided in a quantity of sections length_vect
long and on these it is applied the Hanning window), the number of overlaps noverlap (if
it is empty, it is automatically set so to have 50% of overlapping), the number of output
points N and the sampling frequency used to evaluate the output w, which is the
normalized frequency vector at which the PSD is estimated.
103
7. Appendix B: ISO standards

7.1. Vibrations – ISO 8528-9


It describes the procedure for measuring and evaluating the external mechanical
vibration behaviour of generating sets. It applies to AC generating sets driven by Internal
combustion engines, for fixed and mobile installations with rigid or resilient mountings.

The measured variables are acceleration, velocity and displacement, expressed as the rms
of the value:

𝑡
∫𝑡 2 𝑣 2 𝑑𝑡
𝑣𝑟𝑚𝑠 = √ (𝑡1 −𝑡 (7.1.1)
2 1)

The accuracy of the measuring system, which should give as output the rms of the
measured variable, must be in the range ±10% in the range 10-1000Hz. Depending on
the sensor, the output signal may be integrated or differentiated to derive the quantities
not directly measured. It is important to well fix the transducer in order to affect not the
measurement.

The recommended points of vibration measurement in generating sets are shown in the
figure 1.1. The measuring points have to be located on the solid engine block and on
solid areas of the generator frame, in order to avoid measuring local structural vibrations.

Figure 7.1.1 - point of measure for vibration according to ISO 8528-9 [j]

104
Even if the excitation frequency for these type of engines is in the range 2-300Hz, it is
required to expand the range up to 1000Hz. Reference levels of vibration are provided
below in Tab 7.1.1.

Table 7.1.1 - Reference level of vibration [j]

According to annex B of ISO 8528-9, those kind of generators, experience higher values
of vibration than those running independently. This is typically due to oscillating masses,
torque variation and pulsating forces linked with pipe-work. These lead to an increase of
the stresses on the support and consequently of the vibration entity. Those type of
vibration are usually higher than the common ones linked to rotating machines, and
remain more constant during the life of the plant.
The standard also precise that with this procedure for evaluation of the vibrations, only
general conclusion can be drawn, and it does not allow to make statements on mechanical
stresses.

7.2. Sound – ISO 8528-10


In this standard, it is defined the measurement method for the determination of airborne
noise emitted by reciprocating internal combustion engine driven generating sets in such
a way that the total of relevant noise emissions. The noise of a generating set is defined
105
as the total noise emitted by that generating set: this includes the surface noise of the
engine and the generator, the inlet noise, exhaust noise and the noise emitted from the
cooling system of the engine and the fan of the generator and also the noise which will,
for example, be emitted from the joining sections and the base-frame. All sources of
engine noise, are evaluated on a similar basis to yield comparable results. When the
exhaust and cooling systems are ducted to a remote site their noise contribution is not to
be included in this part of ISO 8528. Noise emission characteristic value is expressed as
sound power level (SPL). The ISO 8528-10 allows also to evaluate A-weighted SPL and
the SPL expressed with one or one third octave band.
There are two grades of accuracy used to classify the measures: grade 2 is used when
there is negligible background noise level and an acoustic-free field over a reflecting
plane. Two coefficients of correction are used, K2A for environmental contribute, K1A for
background noise. Grade 3 is more severe, and require higher values of correction for
K1A and K2A.

When the measurements are taken, the ambient and air inlet temperatures shall be not
higher than 320 K (47°C). The generating set speed, average electrical power output,
ambient temperature and type of fuel and cetane index used during the test should be
taken into account because of their influence on the noise emission.

The generating set should be installed


on a typical noise-reflecting ground
plane of concrete or nonporous asphalt.
The distance from the source to the next
wall of the test cell should be twice the
distance between source and
microphone.

The standard is also very precise in Figure 7.2.1 - Positioning for noise measurement [k]
giving information regarding the area around the generating set and the distance at which
the measures have to be taken (figure 1.2). It says that the measuring points should be
arranged equidistantly along the measuring surface and completely enclose the noise
area; the number depends on the size of the generating set and the uniformity of the noise
field; the arrangement of measuring points for the accuracy grades 2 and 3 do not differ.

However, since the objective of this thesis is not only the measurement of sound, but the
research of the main vibrations which lead to noise, the measurement points for sound
are the same of vibration ones, and the distance is the minimum one. Of course, in the
area of the inlet air and exhaust gas openings, microphones shall be arranged so that they
will not encounter air/gas movements.
106
In order to obtain a test result for gr ade 2 it is required that the calculated correction
factor K2A≤2 dB; for grade 3 it is necessary to have an environment with the
environmental correction K2A≤7 dB. In many cases due to the reality of the acoustic
properties of the measuring area and the test bed conditions (this occurs predominantly
with larger generating sets) only grade 3 may be obtainable.
Noise generated by air movement at the microphone itself, is classified as background
noise. Background noise is taken to be that noise which is not emitted by the generating
set under test, or that which is emitted by openings or component parts which are not the
object of measurement (in our cases, background noise comes from the suction fans of
the room).

The minimum frequency range has to be in the range 63-8000Hz, even if wider range,
especially at lower frequencies, are recommended.

K1 is evaluated as (according to ISO 3744 and ISO 3746):


1
𝐾1 = 10 log (1 − 100.1∆𝐿) [𝑑𝐵] (7.2.1)

Anyway, as it is possible to see in Tab 1.2, in our case no correction should be introduced

Table 7.2.1 - Corrective coefficient K1 depends on difference with background noise [k]

in the evaluation of SPL, since the difference between the machine in use and the
background noise is much higher than 10 dB, so K1A can be considered 0.
The averaged octave or one third octave sound pressure level in decibel, considering also
corrections for background noise and environmental influence, is evaluated as:
𝑛
1
̅̅̅
𝐿𝑝 = [10 log ( ∑ 100.1𝐿𝑝𝑖 ) − 𝐾2 ] [𝑑𝐵] (7.2.2)
𝑛
𝑖=1

where Lpi is the octave or one third sound pressure level at measuring point i, n is the
total number of measuring points. Knowing the surface of the parallelepiped around the
machine equal to:

107
𝑆 = 2 × 2𝑎𝑐 + 2 × 2𝑏𝑐 + 2𝑎 × 2𝑏 [𝑚2 ] (7.2.3)
it is also possible to calculate the sound power level of the source (according to ISO 3744
and ISO 3746):
𝑆
𝐿𝑊 = ̅̅̅
𝐿𝑝 + 10 log ( ) [𝑑𝐵] (7.2.4)
𝑆0

where S0 is the reference surface equal to 1m2. Vice versa, ̅̅̅


𝐿𝑝 can be evaluated as
(according to ISO 11203):
𝑆
̅̅̅
𝐿𝑝 = 𝐿𝑊 − 10 log (𝑆 ) [𝑑𝐵] (7.2.5)
0

Which mean that ̅̅̅


𝐿𝑝 is the 1m distance surface sound pressure level, being S0=1m2.

108
8. Bibliography
8.1. References
[1] International Standard 1995, ISO 8528-9: “Reciprocating internal
combustion engine driven alternating current generating sets – part 9: Measurement
and evaluation of mechanical vibrations”;
[2] International Standard 1998, ISO 8528-10: “Reciprocating internal
combustion engine driven alternating current generating sets – part 10:
Measurement of airborne noise by the enveloping surface method”;

[3] M. Tarabini, 2016, “Experimental Characterization of Mechanical


Systems - A guide for engineers and technicians”, Dispensa del corso di Misure,
Politecnico di Milano, pp 11, Chap. 2 par 2.2.1-2.2.3, par 2.3.1, par 2.3.3.2, Chap. 3
par 3.1.1-3.1.2, par 3.2, par 3.3.2, Chap. 6, Chap. 9 par 9.2.1;

[4] G. Ferrari, 2016, “Motori a Combustione Interna”, 2nd edition, Società


Editrice Esculapio Bologna, BO, Italy; Chap. 1, Chap 5, Chap 13;

[5] G. A. Pignone, U. R. Vercelli, 2016, “Motori ad alta potenza specifica.


Le basi concettuali della tecnica da competizione”, 3rd edition, Giorgio Nada
Editore, Vimodrone, MI, Italy;
[6] N. Bachschmid, S. Bruni, A. Collina, B. Pizzigoni, F. Resta, A. Zasso,
2016, “Fondamenti di meccanica teorica e applicata”, 3rd edition, McGraw Hill,
Milano, MI, Italy, Chap. 7, 8, 9;

[7] R. Corradi, Handbook for Noise and Vibrations course: “Controllo delle
Vibrazioni e del Rumore (nei sistemi meccanici). Vibrazioni delle lastre piane
sottili”, Dispensa #4, Politecnico di Milano;
[8] R. Corradi, Handbook for Noise and Vibrations course: “Controllo delle
Vibrazioni e del Rumore (nei sistemi meccanici). Onde acustiche”, Dispensa #7,
Politecnico di Milano;
[9] T. Kalmár-Nagy, B. Dezső Bak, T. Benedek, J.Vad, 2015, “Vibration and
Noise of an Axial Flow Fan”, Periodica Polytechnica Mechanical Engineering, pp
109-113, par 1, 3, 5;

[10] C. Burgess, R. Thompson, 2014, “Practical Consideration of Noise from


Fans”, Inter-noise 2014, 43rd International Congress on Noise Control Engineering,
Melbourne, Australia; par 2, 3, 4;

109
[11] R. van de Vondervoort, 2015, “Prediction of Aerodynamic Performance
and Noise Production of Axial Fans”, Master degree thesis, University of Twente; p
16 par 1.3, pp 22-23, par 2.5, pp44-46, par 5.2;
[12] V. Diaz, J. L. San Román, J. A. Calvo, 2006, “Controlling the
turbocharger whistling noise in diesel engines”, International Journal of Vehicle
Noise and Vibration, pp 17-28;

[13] M. Janda, O. Vitek, V. Hajek, 2012, “Induction Motors - Modelling and


Control”, CC, Brno, Chap. 8,”Noiseof Induction Machines”, par 3.1-3.4;

[14] J. S. Campbell, W. A. Andersen, V. B. Hansinger, 1966, “Analysis of


Induction Motor Noise and Vibration”, Defense Documentation Center for Scientific
and Technical Information, Cameron Station Alexandria, Virginia; section 3
“Magnetic Noise”, section 4 “Bearing Noise”, section 5 “Fan Noise”, section 6
“Unbalance Noise”, section 7 “Effect of Load”;
[15] W. Bloxsom, 2013, “Sound science: Understanding and implementing
generator set noise control”, Press release, MTU Onsite Energy, Mankato,
Minnesota, USA; pp 3-4, 6-7;

[16] D. Aeberg, 2007, “Generator set noise solutions: Controlling unwanted


noise from on-site power systems”, Technical information from Cummins Power
Generation Inc.; pp 5-6;
[17] J. Örn, 2014, “Vibration guideline for large diesel engines”, Bachelor’s
thesis, NOVIA University of Applyed Science, Vaasa, Finland; p 15, pp 22-26;

[18] A. Lindholm, 2016, “Diesel Generator Set Vibrations Analysis and


Mitigation”, Report 2016:339, Energiforsk AB; Chap. 6 and 8;

[19] Y. Jin, Z. Hao, X. Zheng, 2011, “Comparison of different techniques for


time-frequency analysis of internal combustion engine vibration signals”, Journal of
Zhejiang University-SCIENCE A, pp 519-531; p522-523;
[20] T. Hara, T. Furukawa, K. Shoda, 1995, “Vibration Analysis of Main
Engine Shaft System by Building Block Approach”, JIME (the Japan Institut of
Marine Engineering), p77-81; p80;

[21] H. Tienhaara, 2004, “Guidelines to engine dynamics and vibration”,


dissertation, Marine News, Wärtsilä Corporation, pp20-25;

110
8.2. Sources for pictures
[a] R. Corradi, Handbook for Noise and Vibrations course: “Controllo delle
Vibrazioni e del Rumore (nei sistemi meccanici). Onde acustiche”, Dispensa #7,
Politecnico di Milano;

[b] www.solerpalau.it/formacion_01_23.html
[c] personal.cityu.edu.hk/~bsapplec/Fire/Octave01.jpg

[d] V. Diaz, J. L. San Román, J. A. Calvo, 2006, “Controlling the


turbocharger whistling noise in diesel engines”, International Journal of Vehicle
Noise and Vibration, pp 17-28;

[e] D. Aeberg, 2007, “Generator set noise solutions: Controlling unwanted


noise from on-site power systems”, Technical information from Cummins Power
Generation Inc.; pp 5-6;
[f] http://www.ni.com/en-gb/shop/select/c-series-sound-and-vibration-input-
module?modelId=122186;
[g] https://www.bksv.com/en/products/ transducers/vibration/ Vibration-
transducers/accelerometers/4508-B;
[h] http://www.bswa-tech.com/web_proDetail. action?proId=227;

[i] https://octopart.com/15875-amphenol+rf-49665991;
[j] ISO 8528-9:1995(E), p3, p10;

[k] ISO 8528-10:1998(E), p7, p10;


[l] M. Tarabini, 2016, “Experimental Characterization of Mechanical
Systems - A guide for engineers and technicians”, Dispensa del corso di Misure,
Politecnico di Milano, p 12, p130;

[m] https://goo.gl/images/rkYsDi;

111

Вам также может понравиться