PAKTPASeminarsmall

1、slide 1 pak pak transfer path analysis pak tpa methods, software modules, background information, application examples pak slide 2 transfer path analysis i il hr introduction with transfer path analysis the relation between the source(s), the transfer and the receiver is analysed. slide 3 pak transf

2、er path analysis pak transfer path analysis overview tpa methods modules workflows background of crosstalk cancellation (ctc) and operational tpa theory benefits limitations tpa results sensor placement application examples, schematical content pak slide 4 transfer path analysis time domain tps netw

3、ork transfer functions: frequency domain fir filter impact or shaker reciprocal (volume velocity source) .csv (mount stiffnesses) simulation ctc result target contributions: throughput data superposition in time domain: groups for subsystem analysis save result as new pak measurement sources: operat

4、ional throughput data forces, accelerations, sound pressure, . transfer path synthesis pak slide 5 transfer path analysis graphics / analysis: fft / aps orders . 20 30 40 50 60 70 80 db tps results/pp_ example ctc am m 95 ex sb airco sb engine sb gearbox sb wheels ab intake ab engine ab gearbox ab e

5、xhaust psyn 20 30 40 50 60 70 80 db 0100200300400500600700800900 1000 hz 2000 3000 4000 5000 1/min 20 30 40 50 60 70 80 db 012345678910 ord 2000 3000 4000 5000 1/min 20 30 40 50 60 70 80 db 1 fft crosstalk cancellation (ctc + amm): determination of transfer functions save as pak-measurement time 2 f

6、rf 3 save as pak-measurement (throughput data) 4 time 5 tps network: calculates contributions receiver operational tpa operational measurement: wot partial load run-down pak slide 6 transfer path analysis operational measurement: wot partial load run-down graphics / analysis: fft / aps orders . 20 3

7、0 40 50 60 70 80 db tps results/pp_ example ctc am m 95 ex sb airco sb engine sb gearbox sb wheels ab intake ab engine ab gearbox ab exhaust psyn 20 30 40 50 60 70 80 db 0100200300400500600700800900 1000 hz 2000 3000 4000 5000 1/min 20 30 40 50 60 70 80 db 012345678910 ord 2000 3000 4000 5000 1/min

8、20 30 40 50 60 70 80 db 1 frequency artificial excitation measurement: impact shaker speaker . time 2 frf 3 save as pak-measurement (throughput data) 4 time 5 tps network: calculates contributions receiver impact tpa - crosstalk compensated ctc (without amm) pak slide 7 transfer path analysis no com

9、pensation of crosstalk frfs are taken directly from measurement operational measurement: wot partial load run-down graphics / analysis: fft / aps orders . 20 30 40 50 60 70 80 db tps results/pp_ example ctc am m 95 ex sb airco sb engine sb gearbox sb wheels ab intake ab engine ab gearbox ab exhaust

10、psyn 20 30 40 50 60 70 80 db 0100200300400500600700800900 1000 hz 2000 3000 4000 5000 1/min 20 30 40 50 60 70 80 db 012345678910 ord 2000 3000 4000 5000 1/min 20 30 40 50 60 70 80 db 2 frequency artificial excitation measurement: impact shaker speaker mount stiffnesses . time 1 frf 3 save as pak-mea

11、surement (throughput data) 4 time 5 tps network: calculates contributions receiver impact tpa - crosstalk not compensated slide 8 pak transfer path analysis pak transfer path analysis overview tpa methods modules workflows background of crosstalk cancellation (ctc) and operational tpa theory benefit

12、s limitations tpa results sensor placement application examples, schematical content pak slide 9 transfer path analysis operating structure transfer functions (for y=x.h): hi = xt,i-1 * yt t: observation i: channel x-1: pseudo-inverse observations fft blocks output yt obervations input xt,i principl

13、e pak slide 10 transfer path analysis hz y1 y2 y3 yn x11 x1m x21 x2m x31 x3m xn1 xnm simultaneously measured data acceleration, sound pressure, m references1 response n observations (over time, rpm, impact, ) h1 hm =* -1 y = x * h h = x-1 * y (operational) transfer characteristics matrix conditionin

14、g: overdetermination varied excitation (wot, partial, rundown, .) principle slide 11 pak transfer path analysis how can we calculate the pseudo inverse of x? the singular value decomposition (svd) identifies linear independant deflection shapes. any actual observation can then be described as a line

15、ar combination of these shapes. the decomposition into deflection shapes makes the calculation of the pseudo inverse easy. svd: when x = u s v+ (where u and v are unitary matrices and s diagonal with decreasing singular values) then x-1 = v s-1 u+ pseudo inverse x1 x2 pak slide 12 transfer path anal

16、ysis comparable to an eigenvalue decomposition the measured data are separated into singular values. this decomposition is a transformation into a space of linear independent (orthogonal) principal components. 1st principal component = largest singular value 2nd principal component = next singular v

17、alue single fft result principal component analysis (pca) singular value decomposition (svd) lets call the 1st pc bending and the 2nd pc torsion. each measurement point can then be described by a certain amount of bending and an amount of torsion. ch1 ch2 rotated coordinate system pak slide 13 trans

18、fer path analysis as a least squares approach the svd searches for the best set of singular values to get the best fit of the measured data noise included. noise components are eliminated and not part of the inversion. principal component analysis (pca) singular value decomposition (svd) x 1 x 2 x 3

19、 x n t1 t2 t3 tn 50% 20% 10% 1% yj reference principal components response singular values (cumulated) principal components accelerations, sound pressure eliminate noise components reference: “method of transfer path analysis for vehicle interior sound with no excitation experiment” kousuke noumura,

20、 junji yoshida, honda r&d co., ltd., japan h = x-1 y pak slide 14 transfer path analysis benefits: crosstalk is automatically included since only reactions to any kind of excitation are measured (source side and receiver side) correct separation of airborne and structure borne components no artifici

21、al excitation is needed (typical problems like a bad coherence due to a vibrating plate nearby the hammer do not occur) fast results the operating state is described rather than the idle structure (the result is not distorted by wrong temperatures, .) correct positioning of sensors is not as crucial

22、 as for impact tpa operational tpa operating system output yt input xt,i hi = xt,i-1 * yt pak slide 15 transfer path analysis limitations: all coherent transmission paths need to be measured simultaneously to separated them correctly. no information about forces the quality depends on the energy to

23、noise ratio good results are obtained for example - up to 2 khz for an engine tpa to the drivers ear - up to 5 khz at an engine test bench small contributions are less accurate. the otpa works best for identifying the main contributions. sensor placement and interpretation of results needs engineeri

24、ng knowledge. operational tpa operating system output yt input xt,i hi = xt,i-1 * yt slide 16 pak transfer path analysis pak transfer path analysis overview tpa methods modules workflows background of crosstalk cancellation (ctc) and operational tpa theory benefits limitations tpa results demonstrat

25、ed in pak software sensor placement application examples, schematical content slide 17 pak transfer path analysis pak transfer path analysis overview tpa methods modules workflows background of crosstalk cancellation (ctc) and operational tpa theory benefits limitations tpa results sensor placement

26、application examples, schematical content slide 18 pak transfer path analysis sensor placement all measured signals are responses to existing physical excitations one sensors measures typically a superposition of reactions to different sources depending on the “distance” to the source. if a source i

27、s not detected by any sensor then its contribution cannot be reproduced (e.g. wind). its possible to place more than one sensor to capture one contribution (e.g. mics in engine compartment). the total contribution is the sum of the single contributions. the combustion for example is the source for s

28、tructure borne engine paths as well as for airborne sound. at some points like a carbody connection for example several sources are superposed. the analysis will thus give a contribution analysis of these points rather than an information about the original sources. different hierarchical interface

29、levels must not be mixed. o v e r v i e w exc.1exc.2exc.3 x magnitude slide 19 pak transfer path analysis source contribution the sources can be separated because the sensors are placed close by the sources. sensor placement path contribution here the contributions from the positions where the senso

30、rs are placed can be regarded as path contributions. cs = cp,i cs cs cs cp cp cp cp cp cp cp cp slide 20 pak transfer path analysis a missing path here the source contribution is distributed to the measured paths. the paths contributions are different from the regular case but their sum is still cs.

31、 sensor placement mixed tpa levels the sum equals the contribution of the source. the single cs cannot be interpreted correctly. cp cp cp cp cp cp cp cp cp cp cp cp cs cs slide 21 pak transfer path analysis street measurement/3rdgearwot run1 ex 29.06.2007 11:17:44pak3rdgearwot run1 29.06.2007 - 11:1

32、7:44 0100200300400500600 hz 10 20 30 40 50 60 70 db( a) 3rdgearwot run1 3rdgearwot run1: driverear left tps amm 95: psyn tps amm 95: sb engine tps amm 95: sb gearbox tps amm 95: sb wheels 0100200300400500600 hz 10 20 30 40 50 60 70 db( a) 3rdgearwot run1 3rdgearwot run1: driverear left tps amm 95: p

33、syn tps amm 95 wo wheels: psyn wheels 210-240 hz sensor placement incoherent sources which are not captured by any sensor without the wheel sensors the road excitation can not be reproduced. (incoherent source) slide 22 pak transfer path analysis street measurement/3rdgearwot run1 ex 29.06.2007 11:1

34、7:44pak3rdgearwot run1 29.06.2007 - 11:17:44 0100200300400500600 hz 10 20 30 40 50 60 70 db( a) 3rdgearwot run1 3rdgearwot run1: driverear left tps amm 95: psyn tps amm 95: sb engine tps amm 95: sb gearbox tps amm 95: sb wheels 0100200300400500600 hz 10 20 30 40 50 60 70 db( a) 3rdgearwot run1 3rdge

35、arwot run1: driverear left tps amm 95: psyn tps amm 95 wo gearbox: psyn gearbox (sb) 300 hz sensor placement same source coherent path contributions without the gearbox sensor the egine contribution can be reproduced. it is fully captured by the remaining egine mount sensors. (coherent contributions

36、) slide 23 pak transfer path analysis street measurement/3rdgearwot run1 ex 29.06.2007 11:17:44pak3rdgearwot run1 29.06.2007 - 11:17:44 0100200300400500600 hz 10 20 30 40 50 60 70 db( a) 3rdgearwot run1 3rdgearwot run1: driverear left tps amm 95: psyn tps amm 95: sb gearbox tps amm 95: gbmt front tp

37、s amm 95: gbmt rear 0100200300400500600 hz 10 20 30 40 50 60 70 db( a) 3rdgearwot run1 3rdgearwot run1: driverear left tps amm 95 one gearbox: psyn tps amm 95: sb gearbox tps amm 95 one gearbox: sb gearbox tps amm 95 one gearbox: gbmt front measured with 2 gearbox sensors the summed contribution of

38、the two sensors is the same as the contribution when measured with only one sensor. sensor placement multiple sensors can be used for a (sub-)system slide 24 pak transfer path analysis pak transfer path analysis overview tpa methods modules workflows background of crosstalk cancellation (ctc) and op

39、erational tpa theory benefits limitations tpa results sensor placement application examples, schematical content pak slide 25 transfer path analysis drivers ear scenario 1 interface- / source-points: engine points (sb+ab) target: drivers ear sb: structure borne ab: airborne pak slide 26 transfer path analysis interface- / source-points: cabin (sb+ab) drivers ear scenario 2(a) target: drivers ear pak slide 27 transfer path analysis drivers ear scenario 2(b) interface- / source-points: engine (sb only) target: dominant source