Anda di halaman 1dari 24

Crankshaft Durability of Rover K-Series Engine: Comparison of ENGDYN Analysis with Dynamic Measurements

Roger B. Dailly BMW Group, Birmingham, UK David J. Bell Ricardo Consulting Engineers

Figure 1: MGF with 1.8L K-series VVC engine Abstract This paper describes the technique used to perform a dynamic simulation of the Rover KSeries crankshaft with the aim of predicting multi-axial stresses and Fatigue Factors of Safety using ENGDYN from Ricardo. This production crankshaft had been previously analysed and validated using in-house software techniques and contracted instrumentation work. ENGDYN was used to predict bearing loads and oil film thicknesses across the running range of the engine. Torsional vibration predictions were then compared to actual dynamic measurements of the crankshaft and in-house software results. Multi-axial stresses and Fatigue Factors of Safety were then compared to strain gauge measurements. ENGDYN results indicated good correlation with actual measured results for both torsional vibration and durability of the crankshaft. The results however could also be improved by more accurate FE models of the crankshaft. This paper is a comparitive study and does not analyse in great detail causes for dynamic behaviour of the crankshaft, but does give possible reasons for any differences between the results

5th Ricardo Software International User Conference

1 Introduction
The Rover K-Series engine has now been in production for over 10 years, initially of 1.4 L capacity. Since then 1.1L, 1.6L, 1.8L and 1.8L VVC (Variable Valve Control) have been introduced with an accumulative build of two million engines since 1989. The 1.8L engine is also built under official licence for Lotus and Caterham, and also in the MGF motorsport series. The K-Series is an attractive buy due to its reduced cost and low weight, which is a necessity for sport cars. This report aims to validate Ricardo ENGDYN software with respect to Rover K-Series 1.8 Litre VVC crankshaft durability. The software will be used to output the behaviour of the crankshaft under as realistic conditions as possible. The oil film thickness and bearing load characteristics of the crankshaft through the running range of the engine will be compared to results obtained from in-house software. The torsional and bending vibration output from ENGDYN will then be compared with results obtained from dynamic measurements. Finally, the crank stress and durability results from ENGDYN will be compared to strain gauge measurements at comparative points on the crankshaft. ENGDYN is a computer program used for analysing the dynamics of the engine, and in particular the crankshaft and its interaction with the cylinder block. In this analysis the software will be used to predict the time-domain response of the 3-dimensional vibration of the crankshaft coupled to the block by way of a non-linear oil film. When this loading and motion has been calculated the software can perform a fast Fourier transform to break down the time-domain response into its corresponding frequencies. This allows the results to be post-processed in the frequency domain.

2 Method of Analysis
2.1 Engine Specifications Configuration: in-line 4 Fuel: Gasoline Cylinder bore: 80 mm Piston stroke: 89.3 mm Swept volume: 1.8L Crankpin Peak Power: 107 KW @ 7000 rpm Peak Torque: 174 N/m @ 4500 rpm Engine running range: 750-7200 rpm 2.2 Component Modelling 2.2.1 Crankshaft To perform the analysis within ENGDYN two crankshaft models were created. These included a complete stiffness representation of the crank (excluding the crank nose hub and the flywheel), and a detailed model of the crank from main bearing 4 to main bearing 5, with mesh density increased around the fillets. ENGDYN can however perform crank analysis of any portion of the crank as long as the model incorporates at least two main journal bearings. Features such as bolt holes and oil drilling were omitted on both models, which were meshed using solid tetrahedral elements. The stiffness model of the crank as shown in Fig. 2 was used to generate the mass and stiffness matrices of the crank within ENGDYN. The flywheel and torsional vibration damper assemblies were added within ENGDYN as lumped masses concentrated at the appropriate points. The detailed crank model was used to evaluate the stress and fatigue safety factors around the fillet in the rear bay of the crank, namely web 8. Fig. 3 Shows the stress quality model of the crank . The coupling of the flywheel to the clutch was defined as rigid in ENGDYN. The flywheel

5th Ricardo Software International User Conference

was treated as a lumped mass which did not contribute to the stiffness of the crank. The torsional vibration damper was defined in ENGDYN by the hub and annulus properties. This flywheel and vibration damper are effectively added to the FE model of the crankshaft so that ENGDYN can generate mass and stiffness matrices of the full crank. It is more accurate to define these components in the FE model, but for this analysis these components were not previously modelled. 2.2.2 Cylinder Block The cylinder block was received as an IDEAS model file and required some modification before being read into ENGDYN. The model of the block was very large and it was necessary to create a master-degree-of-freedom set, which represented a condensed mass and stiffness matrix of the engine block. These matrices were derived in MSC/Nastran by static condensation performed in a normal modes solution. 2.2.3 ENGDYN Solution Technique The software previously used by CAE Design Analysis for analysing engine bearing behaviour was Engine Bearing Analysis (EBA). This software represented a rigid cranktrain with appropriate lumped masses modelled within a rigid block. The oil film being modelled using the Booker Mobility Method. For direct comparison it was therefore necessary to solve an indeterminate solution within ENGDYN(1) against a rigid block. For this analysis the oil characteristics and cylinder firing pressure maps across the speed range were also defined. The solution parameters were such that the inlet oil temperature and pressure were 140C and 4 bar respectively. The oil film was modelled such that the rise in temperature, generated from the bearing power and flow rate drop, was calculated at each of

the journal bearings. The centrifugal effects were also taken into account due to the mass and inertia of the crankshaft. 10W30 oil was used along with pressure maps for a full load engine condition. A no-load condition is more detrimental to crank life since the inertia torque is not relieved by the gas torque, however accurate cylinder pressure maps were not available for this load case and hence a full-load condition was used for comparison. Bearing force and eccentricity results were then output and compared with EBA for full load and inertia only running conditions. With the FE models of the crank and block defined in ENGDYN, a dynamic/compliant solution was then performed to evaluate the torsional and bending response of the crankshaft. The dynamic solution generates mass and stiffness matrices of the crank and includes the non-linear gyroscopic effects in the solution. The reactions and journal bearing orbits are then calculated for each pin and journal bearing, with the 3-dimensional vibration behaviour also evaluated. The block was set to compliant, which considered only the stiffness of the block. This means the block was modelled such that the natural mode shapes did not contribute to its interaction with the crankshaft. The results from this solution would then be compared to those from experimentally measured results. A dynamic solution of the crank against a rigid block was also done so that the results could be compared with the Rover Torsional Vibration (TV) program. The TV software uses a mass-elastic system to represent the inertia and stiffness between each web and journal along the crank. 2.2.4 Crankshaft Stresses Upon completion of the bearing load and displacement calculations ENGDYN has the ability to use these results to perform a multiaxial stress combination to resolve the stresses in the desired region. For this crankshaft it

5th Ricardo Software International User Conference

was already known that the fillet in main bearing five closest to web 8 suffered the greatest loading. Failure had occurred in this region during durability tests at 7500 rpm. Since the stress model used in this analysis consisted of webs 7 and 8, then stresses were only resolved in this area. ENGDYN splits the model at cut planes through the journal centres. Each split would extend from the centre of a main journal bearing to the centre of the next immediate pin journal bearing. Taking each split part in turn ENGDYN applies unit force loads in the four orthogonal directions (+Y,Y,+Z,-Z) at both the main and pin journal centres. In addition unit moment loads in the four orthogonal directions are applied at the main journals. For this dynamic solution, which takes into account bending vibration, additional unit moment loads were applied at the main bearing centres. Unit torsional and axial loads were applied at what would be the nose of the crank. In this case the nose is represented by the start of main bearing 4, as shown in Fig. 3. A body load equivalent to a constant angular velocity about the crankshaft axis was also applied so as to include the centrifugal effects of the crankshaft. For the model analysed here, the total number of unit loads amounted to 23. ENGDYN then writes out a file, which can be converted to an MSC/Nastran deck and solved to produce stress per unit loading. The stresses due to quasi-static (a snap shot in the time cycle of crank rotation) bending, torque and vibration loading are calculated such that a number of combined loadcases are created at each crank angle. It is the variation of the bearing loads through the crank cycle that causes the stress to change, and hence a new factor is required so that the unit loads can be multiplied. Each factor also depends on the notch sensitivity and vibration response along the crankshaft. Fatigue Safety Factors (FSFs) are then calculated. The results were postprocessed in Ricardos Safplot software that enables FSFs to be plotted against engine

speed and Goodman diagrams to be created. The resulting file containing all of the factored stresses for each fillet can then be converted and read into IDEAS for post-processing the stress and FSF plots. For this solution the FFS were calculated for the speed range 5000 rpm to 7000 rpm.

3 Results & Discussion


3.1 Normal Modes Table 1 shows the frequencies and mode shape descriptions for a free-free crank with conventional flywheel attached as produced by the Lanczos method using MSC/Nastran. The flywheel was attached to the crank palm with a 6 degree-of-freedom spring with stiffness equivalent to that of the flywheel. The installed modes were calculated with the crankshaft fitted with the torsional vibration damper, flywheel and masses equivalent to the con-rod big-end mass plus piston mass added to the centre of the pin journals. The entire crank was then grounded with linear springs of 5x108 N/mm stiffness at all four of the main journal bearing centres. Natural Frequency 175 Hz 224 Hz 358 Hz 509 Hz 547 Hz 644 Hz 666 Hz 883 Hz Mode Shape Description Vertical Bending Lateral Bending 2 Vertical Bending Torsional Vibration 2nd Lateral Bending Flywheel Axial Crank Palm Vertical Bending Vertical Bending
nd

Table 1: Unrestrained modes of vibration

5th Ricardo Software International User Conference

Table 2 shows that the torsional vibration damper created two smaller torsional vibrations at 321 Hz and 541 Hz. Without the TV damper the torsion mode would create a larger peak and in this case was excited at 509 Hz. Natural Frequency 213 Hz 238 Hz 321 Hz 349 Hz 394 Hz 478 Hz 541 Hz 644 Hz 778 Hz 786 Hz Modes Shape Description

Crank Nose Bending Crank Nose Bending 1st Torsional Vibration Crank Palm Bending Crank Palm Bending 2nd Vertical Bending 2nd Torsion Vibration Flywheel Axial Swashing of Flywheel Swashing of Flywheel

of the oil film whereas the EBA calculation performs a more simple calculation. A similar comparison was made for an inertia only condition at an engine speed of 7500 rpm. Fig. 5a/b show results for main bearing 4 for an inertia only engine condition. As the no load pressure maps were not available the ENGDYN solution was run with only the inertia loading of the crankshaft. This would account for some differences in the bearing eccentricities as EBA is a no load solution. However the maximum bearing load compares exactly.

3.3 Torsional Vibrations The ENGDYN results for a dynamic solution of the crankshaft supported in a compliant block are shown in Fig. 6. Node 1 represents the node at the centre of the hub on the crank nose. The figure shows the response of this node in the rotational direction along the crank axis. The main orders of vibration are plotted and these are relative to a rotating axis set. The speeds at which these resonant peaks occur compare quite well to those obtained through experimental measurement as shown in Fig. 7. The amplitudes of these peaks depends very much on the damping coefficient of the damper rubber and the overall cylinder damping within the engine. For this analysis the damping coefficient of the damper rubber was calculated to be 1.77 N.m.s/rad and the cylinder damping was set to 1000 Nms/rad (default in ENGDYN). The torsional vibration results are tabulated in Table 3 and shows how the predicted values compare to the measured and in-house results. It can be seen that the amplitudes of the predicted harmonic peaks compare very well with the measured results. This indicates that the damping values used in ENGDYN represent the real system. However there are some discrepancies in the frequency at which the harmonic peaks occur. The main differences are in 3.5 order and the first 6.0th

Table 2: Installed modes of crankshaft

3.2 Bearing Loads & Oil Film Eccentricities Fig. 4a illustrates the ENGDYN solution for the characteristics of main bearing 5 under a full load condition at 4000 rpm. Fig. 4b shows how the EBA results for the same bearing compare. It can be seen from the graphs that the ENGDYN results compare well with those obtained from EBA. The only significant differences were those found between the values for oil flow rate and oil viscosity. This difference occurs because the results from ENGDYN include a thermal balance solution

5th Ricardo Software International User Conference

order peak. This indicates that the FE model of the crankshaft does not represent the real crankshaft accurately enough. Improved results can be obtained by incorporating actual FE models of the torsional vibration damper and flywheel to more accurately model these components in the solution. Harmonic Order 4.0th Order Speed (rpm) Freq. (Hz) Amp. (Deg.) 6.0th Order (I,II) Speed (rpm) Freq. (Hz) Amp. (Deg.) 3.5 Order Speed (rpm) Freq. (Hz) Amp. (Deg.) 5000 292 0.10 5625 328 0.10 3000 & 5000 300 & 500 0.07 & 0.06 3400 & 5125 340 & 512 0.09 & 0.07 4400 293 0.15 4750 317 0.15 ENGDYN Results Measured Results

3.4 Bending Vibration & Strain Gauge Measurements The motion of the node representing main bearing 5 (closest to crankshaft palm) was analysed. The results provided a comparison of bending vibration to the strain gauge measurements. Fig. 9a shows the axial displacement of this node in the X-direction. The graph shows that 2.0 order (relative to the crank) vibration is dominant at high engine speed and reaches a maximum around 7000 rpm. This amplitude equates 0.185 mm 2.0 order and 0.228 mm in the time domain. At a lower engine speed the crank is excited at 5400 rpm by 2.5 order and 6100 rpm by 1.5 order. A similar result was obtained for the movement of the same node in the vertical Y-direction as shown in Fig. 9b. Inspection of the results relative to the fixed datum show that the 1.5 and 2.0 order resonances correspond to the forward flywheel whirl mode whilst the 2.0 order resonance corresponds to the reverse whirl mode of the flywheel. A photograph of the instrumented crankshaft is shown in Fig. 10a. The location of the strain gauge is illustrated in Fig. 10b. The precise location of the gauge corresponds to the area of highest stress in the fillet, found from previous analysis. Areas of high strain will coincide with areas of high stress and hence low factors of safety. The strain gauge measures predominantly axial and vertical strain caused by bending of the fillet, the results of which are shown in Fig. 11. Fig. 12 compares the 1.5 and 2.0 order measured strain amplitudes of Fig. 11 with those derived from the ENGDYN analysis. This shows good correlation of the resonant frequencies for each of these harmonic orders. Inspection of the measured 1.5 order strain shows two (possibly three) resonances indicating that there appears to be a vibration absorber mechanism present that effects this order. This is not apparent from the predictions in which a single higher amplitude resonance of 152 Hz (relative to crank) is predicted. It is

Table 3: Comparisons of measured and predicted torsional vibrations. Fig. 8b shows the ENGDYN results for a dynamic/rigid solution. The dominant orders of vibration are plotted and show that 4.0th order vibration is greatest around 4400 rpm. The sixth order peaks occur at 3000 rpm and 5100 rpm. These results are confirmed by the TV software, the results of which are shown in Fig. 8a. The magnitude of the peaks are greater in the ENGDYN results because more accurate damping values were used in the solution.

5th Ricardo Software International User Conference

thought that this vibration absorber mechanism may be due either to the effect of the oil film at the adjacent main bearing, the flexibility of the crankshaft web counterweight or more likely the flexibility of the flywheel which was not included in this model (but can be included in ENGDYN if required.) The correlation of the 1.5 order strain amplitude therefore requires some further investigation. The predicted 2.0 order strains correlate well with the measurements although the predicted amplitude is lower than the measured value. 3.5 Crankshaft Durability The stresses due to quasi-static and vibration loading for every 10 through a cycle of 720 were calculated for each engine speed under full load. These stresses were then combined using an Alternative Goodman Criterion to produce the fatigue safety factors through the desired speed range, as shown in Fig. 13. These results are shown for the fillet at the rear of web 8, closest to the crankshaft palm. Under full engine load at 7000 rev/min the maximum stress was found at 570 of crank rotation, with a value of 198 MN/m2. The stress distribution at this condition is shown in Fig. 14a and the maximum stress corresponds to a fatigue safety factor of 1.283. Similarly at 5000 rev/min as shown in Fig. 14b the maximum stress occurs at 20 crank rotation. It can be seen therefore at higher engine speeds the shear stress due to the gas torque rotates the high stress region in the direction of crank rotation. The ENGDYN factor of safety results compare very well with those obtained through experimental measurement and follow a similar trend. The general decrease in fatigue safety factor with increasing engine speed is due to the corresponding increase in 2.0 order vibration as indicated previously by both the experimental and analytical results. For engine speeds between 5000 and 5250 rev/min the safety factors are over-predicted by ENGDYN, but are under-predicted between 6000 and 6500

rev/min.. The discrepancy at the lower engine speeds is not easily explained since the measured and predicted strains for the dominant 1.5 order resonance at these speeds, as shown in Fig.12, correlate well. However the predicted 2.0 order strains at these speeds are lower than the measured values. Further work is required to explain the differences in fatigue safety factor at these speeds. The discrepancy in fatigue safety factors at speeds between 6000 and 6500 rev/min is consistent with the comparison of measured and predicted strains as shown in Fig.12. As previously discussed, this showed that the 1.5 order strain amplitudes were over-predicted by ENGDYN.

4 Conclusions
The results have shown that ENGDYN could be used with confidence to predict the bearing load and oil film eccentricities of pin and main journal bearings. The mathematics behind the oil flow rate calculation within ENGDYN is more accurate than the EBA solution. With respect to the torsional vibrations ENGDYN predicted slightly lower frequencies for 6.0 and 3.5 order vibration when compared to measured data. However, when compared to in-house software the frequencies matched for all orders of vibration. The analysis did show however that the damping values used in the model were a good approximation. The amplitudes of the predicted peaks were close to the measured values. The ENGDYN results describing the motion of main bearing 5 compared well with the experimental results, in that both showed a rapid increase in 2.0 order vibration at speeds above to 6000 rpm. This is the typical response for a four cylinder crankshaft, with 2.0 order vibration dominant at high speeds. The predicted strains and fatigue factors of safety show good agreement with experimental

5th Ricardo Software International User Conference

measurement and follow a similar trend. The most significant discrepancy is the over prediction of the 1.5 order strain amplitude which results in lower predictions of fatigue safety factor. Further work is required to investigate this difference.

5 References
(1) Ricardo Software: Engine Dynamics Simulation ENGDYN Users Manual, Revision 1.2 Perkins Technology Consultancy: Predication of Durability for a 1.8L I4 Crankshaft Using Three Flywheel Models, March 20, 1998 Perkins Technology Consultancy: An Investigation into 1.8L K-Series Flywheel Whirl Activity, June 16, 1998

(2)

(3)

5th Ricardo Software International User Conference

Figure 2: FE Model of crankshaft for Mass and Stiffness Matrix Formulation

Y X

Figure 3: FE Model ofth Rear Bay of Crankshaft for Stress and Fatigue Analysis
5 Ricardo Software International User Conference

Figure 4a: Oil film thickness and load eccentricity at main bearing 5 from ENGDYN

5th Ricardo Software International User Conference

10

Figure 4b: Oil film thickness and load eccentricity at main bearing 5 from EBA

5th Ricardo Software International User Conference

11

Figure 5a: Oil film thickness and load eccentricity at main bearing 4 from ENGDYN

5th Ricardo Software International User Conference

12

Figure 5b: Oil film thickness and load eccentricity at main bearing 4 from EBA

5th Ricardo Software International User Conference

13

Figure 6: Torsional vibration at nose of crankshaft as predicted by ENGDYN for a dynamic solution of the crankshaft in a compliant cylinder block

5th Ricardo Software International User Conference

14

Figure 7: Torsional vibrations measured at TV damper on running crankshaft(2)

Figure 8a: Torsional vibrations predicted by in-house TV software Method 5th Ricardo Software International User Conference

15

Figure 8b: Torsional vibration at nose of crankshaft as predicted by ENGDYN for a dynamic solution of the crankshaft in a rigid cylinder block

5th Ricardo Software International User Conference

16

Figure 9a: Axial displacement of node 25 at front fillet on main bearing 5 as predicted by ENGDYN for a dynamic solution of the crankshaft in a compliant block

5th Ricardo Software International User Conference

17

Figure 9b: Vertical displacement of node 25 at front fillet on main bearing 5 as predicted by ENGDYN for a dynamic solution of the crankshaft in a compliant block

5th Ricardo Software International User Conference

18

Gauge 1 and Gauge 2 (backup)

Figure 10a: Photograph of Crankshaft Instrumented with Strain Gauges (2)

Pin bearing 4 Strain gauge location at predicted high stress region

Main bearing 5, closest to crankshaft palm.

Figure 10b: Web 8 Closest to Crankshaft Palm Showing Location of Strain Gauge

5th Ricardo Software International User Conference

19

Figure 11: Strain gauge results for web 8 fillet on main bearing 5(3)

Figure 12: Comparison of Predicted and Measured Strain Results

5th Ricardo Software International User Conference

20

Figure 13: Fatigue Factor of Safety resolved for the worst stress in the fillet at each crankshaft speed

5th Ricardo Software International User Conference

21

Figure 14a: Stress plot around fillet, showing maximum stress at 570 crank angle at 7000 rpm full load

Figure 14b: Stress plot around fillet, showing maximum stress at 20 crank angle at 5000 rpm

5th Ricardo Software International User Conference

23

5th Ricardo Software International User Conference

24

Anda mungkin juga menyukai