2. FINITE ELEMENT MODEL

Transcription

1 e a C ~ N -Fy STRUCTURAL DYNAMICS MODELING AND TESTING OF THE DEPARTMENT OF ENERGY TRACTOWTRAILER COMBINATION Richard V Field, Jr?, John E Hurtado*, Thomas G * Experimental Structural Dynamics Sandia National Laboratoiies t Structural Dynamics and Vibration Control Sandia National Laboratories POBox 5800 Auquerque, New Mexico fixed-front pivot combed with a sprg/damper rear mount, and is very common on the road today 2 FINITE ELEMENT MODEL Figure 1 illustrates the MSC/NASTRAN fite element model of the DOE tractorhrailer combation Currently, the cab, sleeper, and trailer are modeled as rigid body elements; each has a concentrated mass element associated with it located at the center-of-mass, and appropriate rigid element connections jo these pot masses to the remader of the vehicle One particular terface is located at the fifth wheel, where the tractor and trailer connect The fifth wheel itself is modeled as a pot mass, cludg the effects of the steel mountg plate that adds significant stiffness to the tractor frame At the top of the fifth wheel is the kg p, which allows relative rotation the X and Y directions between the tractor and trailer Another terface occurs where lear sprgs model the cab and sleeper "hard-mounts" to the truck frame These mounts exhibit stiffness the three translational directions, and are situated a 3-pot and 4-pot configuration, respectively The tractor frame and crossmembers are modeled with a large number of QUAD elements that capture the frame elastic modes NOMENCLATURE g k rms Uf 1 S(f 1 FRF MAC MIF US1 Dampg Unit of acceleration Stiffness Root-mean-square Insertion Loss Factor Power Spectral Density (PSD) Function Frequency Response Function Modal Assurance Criterion Modal Indicator Function Ride Severity Index A detailed view of the steer axle suspension system is illustrated Fig 2, where the leaf sprg is modeled as a simple lear sprg the vertical direction, and "very stiff the 1 INTRODUCTION / Designg trucks that exhibit good ride characteristics has been a challenge to automotive engeers for many years When attemptg to improve ride quality, the response of primary terest is the vibration experienced by the driver and bunk occupants The ride environment of the truck driver is fluenced by road roughness, rotatg tire/wheel assemblies, the drivele, and the enge [6]In this study, only the ride due to external road surface excitation is considered /' Previous methods to improve cab-ride quality cluded alterg the frame bendg stiffness, troducg softer primary suspensions, and improvg tire stability and driver seats [3] The idea of a cab suspension system was first employed heavy truck design more than three decades ago and consisted of an dependent leaf sprg assembly placed at each of the four corners of the cab [4] Sce then, the design of the cab suspension system has evolved to a This work was supported by tbc Unitcd States Dcpartmcnt of Encrgy undcr Cbnwact DEACO-LWAL X90f10 _ SanJia is a niulriprogram laboratory operated by Sandia Corporationa Lockhccd Mart Company, for thc Unitccd Sraics Dep;irtnicni of Energy ~ OSTI POBox 5800 Auquerque, New Mexico ABSTRACT: This study presents a combed analytical and experimental effort to characterize and improve the ride quality of the Department Of Energy tractor/trailer combation The focus is to augment the experimental test results with the use of a high quality computer model The discussion cludes an overview of the fite element model of the vehicle and experimental modal test results System identification techniques are employed to update the mathematical model The validated model is then used to illustrate the benefits of corporatg two major design changes, namely the switch from a separate cab/sleeper configuration to an tegrated cab, and the use of a cab suspension system C Carrie*, and Clark R Dohrmannt 3 Figure I : MSCBVASTRAN model of the DOE fractor/trailer combation -1-,

2 Portions of this document may be illegible ekctronic image products Images are produced from the best available origal document

3 DISCLAIMER This report was prepared as an account of work sponsored by an agency of the United States Government Neither the United States Government nor any agency thereof, nor any of their employees, make any warranty, express or implied, or assumesany legal liabiiity or responsibilityfor the accuracy, completeness, or usefulness of any formation, apparatus, product, or process disclosed, or represents that its use would not frge privately owned rights Reference hem to any specific commercial product, process, or service by trade nami: trademark, manufacturer, or otherwise does not mcesariiy constitute or imply its endorsement, recommeodation,or favorg by the United States Government or any agency thereof The views and opions of authors expressed herem do not necessarily state or reflect those of the United States Government or any agency thereof

4 A/ \fi / V Leaf sprg / performed with them disconnected The tires (each axle contas four) are modeled identical to those on the steer axle, with one exception - an additional sprg can be cluded to model the combed tirehake stiff ness the fore/aft direction This allows the NASTRAN model to simulate the vehicle modal response with and without the brakes applied As before, the front and rear axles were modeled as a series of BEAM elements The suspension systems on the trailer also utilize leaf sprgs, but with a much higher sprg rate Similarly, modal testg was performed with and without the brakes on the trailer, leadg to a tire model identical to that on the tractor drive axles Tiremodel -/ 3 MODAL TESTING v- : Figure 2: Steer axle suspension model remag five directions The leaf sprg is rigidly connected to both the steer axle and the truck frame A lear tire model is used here, consistg of two sprg elements that represent the effective stiffness the vertical and lateral directions Note that this is valid because only small displacements are expected a modal test There is no fore/aft stiffness for the tire because there are no brakes on the steer axle Lastly, the axle model is a series of NASTRAN BEAM elements A low level, contuous-random force excitation signal was put to the vehicle through the electrodynamic shakers at two locations on the vehicle; the 100 Ib shaker was located at the frontlpassenger side of the tractor, and the larger shaker was placed at the passenger side of the trailer, halfway back The signals sufficiently excited the structure while keepg the dynamic response with the lear regime about the static equilibrium Frequency response functions (FRFs) between the applied excitation forces and the measured accelerations were calculated and recorded Modal frequencies, dampg factors, and mode shapes were then estimated from the recorded FRFs Figure 3 contas a detailed view of the Neway AD-246 suspension system used on both the front and rear tractor drive axles Each suspension system utilizes a trailg arm, modeled as a rigid lk, an airbag, modeled as a lear sprg the axial direction only, and anti-roll bar, modeled as a series of BEAM elements The AD-246 also contas two shock absorbers (just -board of the airbags) that connect the anti roll-bar to the frame cross-members These were not modeled, however, sce the experimental modal tests were The modal survey of the vehicle without the brakes applied was performed by releasg the applied brakes and rollg the vehicle forward to overcome any stiction the wheels and beargs Figure 4 shows experimental FRFs for this test Anti-roll bur Front drive ax &x Several modal tests were performed on the vehicle Tests were applied with and without the brakes applied, but only the latter case is discussed here Electrodynamicshakers rated at 50 Ib, 100 Ib, and 250 Ib were used to simultaneously excite the structure The forces were applied the vertical direction, while a total of 101 Endevco 7751 accelerometers monitored the vibrational response of the system \ Tire model t 0 t I 4 Frequency ( H z ) F 8 Figure 4: Experimental FRF Figure 3: Drive axle suspension model

5 080 Updated parameter values were obtaed usg the -house code PESTDY (Parameter Estimation for STructural Dynamics) and lead to the improved MAC and frequency agreement shown Table 2 and Figs Mode Description - Test/Analysis Frequency (Hz) MAC rollltwist #I 109 I rollltwist #2 176 I 176 bounce pitch 319 I counter pitch 416 I bendg 596 I forelaft Frequency (Hz) Figure 5: Experimental MIF configuration, and Fig 5 shows the Mode Indicator Functions (MIFs) Table 2 lists the estimated natural frequencies from the data analysis, along with a brief description of the mode shapes Both the experimental and analytical mode shapes are displayed Figs 6 through 11 Table 2: Results of model update 4 TEST/ANALYSIS CORRELATION In order to obta good agreement between the fite element model and modal test results, it was necessary to change the values of some uncerta parameters the model Six of the parameters that were updated are shown Table 1 The origal parameter values lead to large errors the frequencies of the model In addition, the mode shapes did not correlate well, as dicated by poor itial MAC values These rather large discrepancies can be attributed to a lack of sufficient formation to itially estimate the stiffnesses of the various suspensions and tirelwheel assemblies Parameter Tire vertical stiffness Origal Tire forelaft stiffness Eb 2015, 2580, Tractor leaf sprg stiffness 2050, r 8340, Tire lateral stiffness 1500, rn Air bag stiffness 2100, Trailer leaf sprg stiffness 14000, 7 Figure 6: RolVtwist mode #1, experimental, (a), and analytical, (b), Updated 11750, 4568, , Ei n Eb Eb 1040, , Figure 7: RolVtwist mode #2,experimental, (a), and analytical, (b) Table 1: Parameter changes due to model update -3-

6 Figure 8: Bounce mode, experimental, (a), and analytical, (b) Figure 9: Pitch mode, experimental, (a), and analytical, (b) A few comments concerng the results of the test/analysis correlation are order Because the correlation was performed usg modal test data, the updated parameters are associated with small amplitude motions It is expected that some stiffnesses, such as those of the leaf sprgs, would be reduced over-the-road environments where stiction effects are overcome Efforts are currently underway to obta stiffness and dampg estimates directly from the transient response of drivg the vehicle over a prescribed bump Fally, it is noted that there remas some uncertaty the absolute values of the tire vertical and suspension stiffnesses because they act series 5 DESIGN WORK The design of a next-generation DOE tractor is to be completed the next few years This new vehicle will most likely exhibit significant modifications from the current design One such change may be a switch to an tegrated cab configuration, sce this provides more room to the occupants Another might be the addition of a tractor cab suspension system to improve the vibrational response of the cab this paper, the authors exame the performance of a cab suspension system that is available on the market today In addition, an optimized suspension will also be considered This cab suspension system is optimal the sense that it provides the maximum protection to the driver from vibration due to road surface put For the purposes of this study, the suspension is modeled as a simple sprg-damper system and is placed to the full updated fite element model with new tegrated cab, as shown Fig 12 Furthermore, the stiffness and dampg constants of this system are used as the design variables the optimization problem to be discussed later Lear sprgs the three translational directions model the cab hardmounts to the truck frame the front Figure 10: Counterpitch mode, experimental, (a), and analytical, (b) Figure 1 1: Frame bendg mode, experimental, (a), and analytical, (b) Figure 12: Integrated cab with cab suspension -4-

7 Frequency response and power spectral density analyses can b e utilized to assess the vibratory response of automotive vehicles Below are s o m e results from such a study that simulates the DOE vehicle drivg a t 55 rnpb over a rough road S o m e theoretical concepts a r e presented, and analytical results usg the model are discussed Special emphasis is given to the issue of random loadg and the cross-spectral density matrix that simulates the road put to a n 18-wheel vehicle (4) where keq is the effective stiffness a t the particular station of terest (!e, the collective vertical tire stiffness), and $ l, ( f ), qi2(f), and $ 2 2 ( f ) a r e the Fourier transforms of Eqs (1)-(3) Note that by scalg Eq (4) by the stiffness terms, the put is now a force power spectral density Most road surfaces have irregularities that a r e best described as random processes, and many can be accurately represented under a n additional assumption of stationarity [5]It is therefore relevant to study the vehicle response due to a stationary random rough road put at several key locations of the truck The remag terms volve various degrees of correlation between the different axles Because the vehicle is at a constant speed, this c a n b e accomplished by delayg the put to the tractor front drive axle, as well as those behd it [8],as illustrated Fig 13 It h a s been found that terra roughness can be represented the frequency doma by power spectral density functions of the form To apply this put PSD, knowledge of the cross-spectral density matrix of put forces is required Assumg the vehicle is followg a straight, constant s p e e d trajectory, let W, and W, represent the amplitude s e e n by t h e left and right tires, respectively, due to the excitation of the road For the studies presented here, it is assumed that the left and right tires experience the s a m e road surface, but t h e two a r e not perfectly correlated Defe $,l(t) = $&) = E[W,(T)W2(t (5) Y(Q) = c n -, where R is the spatial frequency, units of cyc/es//engtb, and n and c a r e constants As discussed [5], n = 20 is typical Therefore, for a vehicle travelg at a constant s p e e d of v o, the power spectral density that describes the put displacement s e e n a t the road surface c a n be given by (1) (3) where E [ - ] is theexpectation operator, $,,(z) and $22(2) are the autocorrelation functions, and QI2(z) is the crosscorrelation function representg the teraction between the left and right sides of the vehicle where f is now t h e temporal frequency, units of cyc/es/sec In addition, plots of Y vs R are presented [l], (81and [9] for several road surface types Assumg a rough r o a d terra, c = 324e-6 -cycles Figure 14 vs frequency, where plot (a) is directly from Eq (6) for vo = 55 m p h To estimate the correlation of the terra between the left and right tire tracks of the vehicle, consider the coherency function The schematic shown Fig 13 illustrates the 10 locations which the rough r o a d will excite the vehicle, assumg the dual wheel locations a t stations 3 through 10 act together The complete put power spectra is therefore characterized by a ten-by-ten matrix, where the elements of this matrix volve the Fourier transforms of Eqs (1)-(3) Specifically, the two-by-two blocks on the diagonal of this matrix a r e of the form (7) I I A, 00 Frequency ( H z ) Figure 13: Input stations for the DOE tractor/trailer combation Frequency (Hz) Figure 14: Input autopower and crosspower spectral density functions -5-

8 , Typically, as discussed [9], the coherency function is near unity at low frequencies (Le, high correlation for hills or dips the road surface), and much lower at the higher frequencies (Le, low correlation for cracks the pavement, potholes, etc) Utilizg this assumption, Eq (7) can be used to deterc f ) Fig 14(b) me ~ $ ~ ~shown Design The cab suspension system, placed between the rear of the tegrated cab and truck frame, can be utilized to improve the ride quality of the vehicle The ride severity dex (RSI), a weighted measure of the power of the output spectrum, can be used to gauge this improvement Consider the followg relation where S O u r ( f ) is the output power spectral density function and f E If,,f2] denotes the frequency band of terest As shown, the RSI is a scalar quantity that is simply a weighted sum of the output PSD The sertion loss factor, Lcf), is a quantity dicative of the human body's sensitivity to vibration The higher the loss factor, the more sensitive the human is to the excitation, c, -s R S I, mg-rms Current cab 2042 Current sleeper 2353 z = z = z = z = z = Plots illustratg the possible benefits of an optimized cab suspension are cluded Figs 15 and 16, where it is evident that changg the design of the cab suspension only has a significant effect on the response of the cab the frequency range of 0 to 6 Hz The optimal RSI occurs when the suspension is pushed 24 ches forward of its present location The vertical response Fig 15 illustrate a drastic reduction the bounce and pitch modes when usg the optimal cab suspension In addition, the vertical mode of the suspension reduces frequency from f = 134 H z to f = 100 H z sce, as the location of the cab suspension moves forward, it is required to carry The mimization problem to be solved here can be stated as c,,, suspension, and positive and negative values of z are behd and front of it, respectively Note that the optimal sprgdamper values are to be applied at both the passenger- and driver-side of the vehicle The terms volvg S z ( f ) are scaled by an additional factor of (14)2 = 196 sce vibration the forelaft direction is particularly disturbg to humans [2] k, c, z Table 3: RSI for current design (cols 1 and 2) vs tegrated cab with optimal suspension For the studies cluded here, only the vibration at the driver seat is considered, and the ride severity dex is given by m RSI, subject to km,cm5 k, c 5 k,,,, k, Bounce C a b suspension 'itch Counter P i t c h (10)? I where k, c, and z refer to the stiffness, dampg, and location of the cab suspension system, respectively, and the bounds on the stiffness and dampg ensure physically practical solutions Because varyg the location of the suspension system requires re-meshg the fite element model, however, Eq (10) is solved for a fixed location, z This location is then varied discreetly along the tractor frame to provide a qualitative estimate of the optimal location ' 0 d -New design -Existg vehicle -- The optimization capabilities of MATLAB [7] are utilized to determe optimal values of the cab suspension system at various locations along the frame The results are shown Table 3, where z is the location of the suspension system, referencedto the location of truck's third frame crossmember (where a typical cab suspension would be located due to structural considerations) z = 0 is the typical location of the 2 4 Frequency ( H z ) 6 Figure 15: Vertical response at the cab of existg vehicle and new design -6-8

9 7 ACKNOWLEDGEMENTS The authors would like to acknowledge the help of Rob Pirtle and Hoden Farrah of Marmon Motors and Ken Vande Brake of Lk Manufacturg, Inc 8 REFERENCES [I] Baum, JM, JA Bennett and TG Carne, Truck Ride -- -New design -Existg vehicle 4 6 Frequency ( H z ) Figure 76: Fore/aft response at the cab of existg vehicle and new design [2] 8 [3] [41 more of the cab weight When considerg the fore/aft response of the cab, shown Fig 16, the optimal suspension aga demonstrates reduction the power of the bounce, pitch, and counter-pitch modes [5] [6] As a general observation, it appears that the RSI will contue to decrease as the location of the cab suspension is moved forward There are physical limits, obviously, sce movg the suspension forward requires that it support a greater static load due to weight of the cab Cab stresses are also an issue and, therefore, the location of the cab suspension cannot be chosen to mimize the ride severity dex alone [7] [8] 6 CONCLUSIONS [9] The prelimary dynamic model of the DOE tractor/trailer combation correlates well with the characteristics seen durg the experimental modal tests As the model is refed, it is clear that it will be an tegral tool determg the design of the next-generation tractor Any possible benefits to be gaed from the use of available passive cab vibration suppression methods will be researched In addition, the computer model can be used to exame any improvements utilizg active vibration control techniques Possible design modifications to the vehicle, namely the conversion to an tegrated-type cab with a cab suspension system, have been discussed With the addition of the tegrated cab with nomal cab suspension, a 44% reduction ride severity can be achieved over the current cabkleeper configuration, which is hard-mounted to the truck frame Furthermore, once an optimized cab suspension is utilized, the RSI can be reduced by an additional 18% Fally, the results shown here dicate that the cab suspension system should be located as far forward as structural considerations will allow -7- Improvement Usg Analytical and Optimization Methods, General Motors Research Laboratories, GMR-2324-R, 1978 Evaluation of Human Exposure to Whole-Body Vibration, International Organization for Standardization, I S /1-1985(E) Flower, W, Analytical and Subjective Ride Quality Comparison of Front and Rear Cab Isolation Systems on a COE Tractor, Society of Automotive Engeers, NO780411, pp , 1978 Foster, AW, A Heavy Truck Cab Suspension for Improved Ride, Society of Automotive Engeers, No , pp ,1978 Gillespie, TD, Fundamentals of Vehicle Dynamics, Society of Automotive Engeers, 1992 Gillespie, TD, Heavy Truck Ride, Society of Automotive Engeers, The Thirty-First L Ray Buckendale Lecture, SP-607, 1985 Grace, A, MATLAB Optimization Tooox, The MathWorks, 1995 Healy, AJ, An Analytical and Experimental Study of Automobile Dynamics with Random Roadway Inputs, ASME Journal of Dynamic Systems, Measurement, and Control, December, pp , 1977 Soong, TT and M Grigoriu, Random Vibration of Mechanical and Structural Systems, Prentice-Hall, Inc, 1993