euro.noise98 Mtinchen PREDICTION OF WHEEURAIL NOISE AND VIBRATION - VALIDATION OF RIM - Rolf J. Oiehl, Muller-BBM, Robert-Koch-Str. I I, D-82152 Planegg, emaii: Rolf.Oiehl@mbbm.de Georg Holzl, Oeutsche Bahn AG, V6lckerstr. 5, 0-80939 Munchen, email: Georg.Hoelzl@bku.db.de INTRODUCTION For the speed range relevant for the majority of today's railroads rolling noise is the predominant source. For the better understanding and support of the development of low noise technologies OB AG and Muller-BBM have developed a simulation tool, called RlM, for the prediction of wheel/rail rolling noise. To broaden the base for decisions this tool is used as an alternative for the TWINS model, owned by ERRI[I]. THEORETICAL BASIS The general setup of the wheel/rail impedance models was proposed in [2] and preceding papers. As described in [3] the Rail/wheel-Impedance Model RIM uses the combined roughness of the running surfaces of wheel and rail as the source of excitation for the system. First the structureborne velocities of the system components (wheel, rail, sleeper) are calculated. Then the pass-by levels are estimated using a simple model for air-borne sound propagation. MEASUREMENTS Measurement set up. In order to validate the prediction model a measurement campaign was realized in 1997. A test train of four coaches was composed. Two test sections were selected in a track, one with smooth rail surfaces, another with rough rail surfaces (slightly corrugated). Vibration sensors on the rail, the sleeper and on the ground and microphones at distances of 25, 7,5 and 3 m from the track were installed to give two measuring sections per sulface quality to measure track responses and pass-by levels. One wheelset was equipped with telemetry based sensors to register the wheel response. Test runs. The test train, consisting of four coaches and a loco, was run over the test sections three times for each speed of 100, 160 and 200 km/h. 271
Characterisation of the track. The track was standard DB AG with ballast, rail UIC 60, sleeper B70 Wand stiff rail pads. Due to the traffic on the line it was only possible to measure the input impedance for the unloaded track. Characterisation of the running surfaces. On every wheel of the vehicles in the test train three tracks were measured, on both rails a single track over a length of 20 m. The roughness spectra of vehicles and track were averaged separately and then added energetically to give a combined excitation roughness. COMPARISON OF MEASURED AND CALCULATED DATA ~ 140 z!!' 130 CD "Cl od g 120 "Cl '" Q) "110 track, measured -track - wheel 100 8 16 31.5 63 125 250 5001000200040008000 Figure I: Calculated impedances of the track and the vehicle and a measurement result for the track, averaged over several postions, including between and above the sleeper. ~ ~-30 CD "Cl ~-40.2 oe-50 o a; > -60 8 16 31.5 63 125 250 5001000200040008000 100 - smooth 100 rough 200 - smooth 200 rough Figure 2: Excitation velocity of combined surface roughness of rail and wheel, measured for the shorter wave lengths, estimated for the rest of the range. The dotted lines show the corrugated surface velocity, the calor the train speed. Impedances. Fig. I compares the calculated impedances for vehicle and track. Due to not having had the possibility to pre-ioad the track for the impedance measurement the model parameters 272
were set to higher stiffness than the measured impedance would indicate, as it can be assumed, that ballast and pads will stiffen with pre-load. Sound and vibration. This stiffening seemed to be quite appropriate as the results for the passby levels at 25 m in Fig. 3 show. Fig. 2 shows the roughness velocities Lv, = L R + 20Ig(w), derived from measured or estimated roughness levels L R for the different train speeds v and sites, using w = 2rrvl A (wavelength '\). The transfer functions Ltf(v,p) (Fig. 3 to 4) were calculated as Ltf(v,p) = Lv,p - Lv, using the velocity Lv or pressure level Lp. The following table shows, that the overall airborne pass-by level at 25 m from the track only deviates by a maximum of 3 db for one case, the average deviation being I to 2 db. I train speed km/h 200 160 100! smooth rail db -1-1 -I I rough rail db 2 2 3 As the roughness measurements did not cover the whole width of the running surfaces of the rail, it can be assumed, that these deviations are partly caused by measurement uncertainty of the excitation. ~ 30 u. '" ~ 20 m '0 c: 10 o 13 c:.2 0 ~ 8 16 31.5 63 125 250 5001000200040008000 30 ~ 20111111111111111111 c: 10.2 13.2 0 =l=1= ig -+- ~ c:.~- = g-10 8 16 31.5 63 --~ 125 250 500 1000200040008000 Figure 3: Transfer function L'f(p) of surface roughness - sound pressu>ee at the train speeds V = 100 km/h (top) and 200 kmlh (bottom) for a measuring distance of 25 m. The thick grey line starting at a frequency of 8 Hz represents the prediction, the thick black lines the measurements for rough (- -) and smooth (-) rails, the thin lines show the additional measurement results for the various sections at distances of 7.5 m and 3 m normalized to 25 m. A comparison of Figs. 3 and 4 shows, that the airborne predictions are beller than structureborne, especially for the higher frequencies, where the levels are underestimated. The measured 273
E.!!! '"E 16 31.5 63 125 250 5001000200040008000 Figure 4: Transfer function L,!(u) of surface roughness - vibration velocity on the rail at a train speed V = 100 km/h. The thick black line starting at 8 Hz represents the prediction, the thin lines show the measurement results for the 4 sections (above and between sleepers, rough and smooth rails). level differences between rail and sleeper, as presented in Fig. 5, fit quite nicely between the two extremes predicted for the fastening position and the average level. ~10" E e 0 m u.. :g-10 ~-20..:.._._ ~ c.~...-. ~-30 ~- - 8 16 31.5 63 125 250 500 1000200040008000 Figure 5: Transfer function of rail vibration velocity - sleeper vibration velocity at a train speed V = 100 kmlh. The thick black lines starting at 8 Hz show the prediction for a "measuring" point on the fastening (-) and for an average velocity on the sleeper (- -). The thin lines show the results for the different measuring points at the fastening and the center of the sleeper. APPLICATION RIM is being and has been used for various prediction and developement tasks in the past, e.g.: I. prediction of noise reduction measures for bridges, 2. developement of structural noise reduction measures for non-ballastcd track. Prediction ofnoise emissions from bridges. In addition to the rolling noise emissions from wheel, rail and sleeper, bridges may, due to the size of the radiating surfaces and the amplification by structural resonances, emit quite considerable noise levels, also known as bridge booming. In the relevant cases this boomig noise is more annoying than the normal rolling noise, although A-weighted pass-by levels will not fully account for this. 274
In order to estimate the effectiveness of noise reduction measures RIM has been used to predict the transfer function of excitation levels on the rail versus vibration levels on the structure. Using these, predictions of the emitted sound levels have been made, based on measurements for the present state. In order to estimate the behaviour of a bridge either FE-models or simpler models describing the input impedance of the structure are added to the ground model. It could be shown, that depending on the frequency range a combination of beam and plate can be used to achieve the same average value as is calculated by time consuming exact FE-modelisation of a steel bridge: for low frequencies the bridge is modelled as a beam, in a medium frequency range the modes of the steel plates between the transversal and longitudinal support have to be taken into account, whereas for higher frequencies it is quite sufficient to use the formula for the infinite plate without any resonances. - simple model ID ~110 - *Yf~g 't III -, -g 100 :::8=1=" ==",1- '1-"0 ~ -t--+,- - FE model 90 1 2 4 8 16 31.5 63 125 250 500 1000 Figure 6: Comparison of calculated impedances: simple beam, plate approximations versus FEmodel. Structural noise reduction measures for non-ballasted track. In order to investigate the reason for the higher noise emissions of non-ballasted track the structural behaviour of both, ballasted and non-ballasted track have been studied. Focal point were the damping and stiffness characteristics of the support of the rail and its fastening. The calculations showed, that the lower damping in the structure and the lower connection impedance for the rail that are inherent to non-ballasted (slab) track are responsible for the higher noise emissions. The theoretical studies led to a proposal of a track construction with booted sleepers with internal damping which is at present under developement within a DB research project. The sleeper boot is designed to allow for the necessary deflection of the rail to ensure sufficient riding comfort. The sleeper is designed to act as a vibration absorber for the rail, thus leading to lower vibration levels on sleeper and rail and in consequence to lower radiated airborne levels. Other applications. Apart from the above described two applications RfM has been used in various studies in order to better understand rolling noise phenomena: Disc braked passenger coaches have the spectral maximum of the pass-by. level above 1000 Hz whereas block braked freight waggons have a maximum at lower frequencies: due to the vehicle speed and the wave number distribution of the exciting roughness spectra, the wheel resonances are less dominant in the emitted noise as the excitation through the corrugated wheels is higher at lower frequencies [4]. 275
To establish criteria for rail grinding according to acoustic requirements a sensitivity study has been conducted [5] to investigate the importance of the wave number distribution of the combined roughness of the running surfaces of wheels and rails. The most important wave length range for conventional vehicles on conventional ballasted track is between I and 5 cm. For track constructions with dominant rail emission it has to be extended to 10 cm. Due to the dependency of the sensitivity on the parameters of the vehicle/track system the requirements could change in the future for different systems. In order to predict the interior and exterior sound levels of trains a prediction scheme has been developed [6,7]. For the estimation of the rolling noise contribution of wheels and rails sound power data has been estimated using RIM. Validation measurements of the specifications for different vehicles showed good agreement with the predictions. For the prediction of the insertion loss of mass spring systems and other vibration mitigation measures ground and slab models have been developed. RIM has been used to supply the track model and the excitation of the system [8]. The setup was used to optimize the combination of the stiffnesses of the different resilient elements in the construction. CONCLUSION RIM is a wheel/rail-impedance model with roughness excitation for the prediction of railroad noise. The model has been validated for passenger coaches on ballasted track in the speed range 100 to 200 km/h. The model has been applied sllccessfully in various studies investigating the rolling noise generation mechanism and its effect on vehicle interior noise, track construction and track substructure. REFERENCES [I] TWINS: Track- Wheellnteraction Noise Software. Theoretical Manual, TNO Institute ofapplied Physics, 1996. and Thompson, D. J.: Wheel-rail noise generation, parts l - V. Journal of Sound and Vibration, 161 :387-482, 1993. [2] Remington, P. J. : WheeUrail rolling noise. parts land fl. Journal of the Acoustical Society of America, 81:1805-1832,1987. [3] Diehl, RJ., G. Hblzl and R. Gbhrlich: Acoustic optimisation of railroad track using computer aided methods. in WCRR'97, Vol E, 421-427, 1997. [4] Diehl, R.J., U.J. Kurze and W. Weil3enberger: Schallschlltzmaj3nahmen an Eisenbahnen. in DAGA'98, to be published J998. [5] Hblzl, G., R.I. Diehl and M. Beier: Anwendung der Rollgeriillschprognose ZlIr ElllWicklllllg eilles Schienenschleifkriteriums. in DAGA'98, to be published 1998. [6] Diehl, R.J. and G.H. Muller: Noise management for high speed trains. in "2nd International Workshop on the Aeroacoustics of High-Speed Tracked Vehicles" in Berlin 1997/04129, to be published 1998. [7] Diehl, R.J. and G.H. MUlier: An engineering model for the prediction of illterior and exterior noise of railway vehicles. submitted for publication to EURONOISE'98. [8) G.H. Muller, R.J. Diehl and M. Dbrle: Assessemelll ofinsertion loss ofmass spring systems for railway lines. submitted for publication to EURONOISE'98. 276