I have focused my research on the relationship between the transmission error of a hypoid bevel gear pair and the noise, vibration, and harshness performance of an automotive rear-axle main reducer. A hypoid bevel gear is one of the most important transmission components in a rear-drive vehicle. It reduces speed, increases torque, and changes the direction of motion. Its meshing quality directly affects the NVH behavior of the main reducer and the whole vehicle. In practical production, a hypoid bevel gear pair usually undergoes cutting, heat treatment, and lapping. Lapping is the final modification process, and it can effectively control and reduce transmission error. The forward-side and reverse-side transmission error values of the hypoid bevel gear pair after lapping are closely related to the NVH performance of the main reducer under acceleration and coast conditions. My objective is not simply to make transmission error as small as possible. Instead, I aim to identify a suitable range of transmission error for a hypoid bevel gear pair and to optimize the lapping process so that the gear pair remains within that range after manufacturing.
I began by establishing a vibration model for a hypoid bevel gear pair meshing system. This model helps explain how manufacturing errors, time-varying mesh stiffness, and tooth surface contact produce internal excitation. The excitation then propagates through the gear pair, the bearings, the housing, and the axle structure, and finally appears as airborne noise and structural vibration. I treat the hypoid bevel gear pair as a single-degree-of-freedom torsional system with equivalent mass, damping, mesh stiffness, and displacement excitation. The basic motion equation I use is
$$M\ddot{x}+C\dot{x}+K(t)\left(x-e(t)\right)=W$$
where \(M\) is the equivalent mass of the hypoid bevel gear pair, \(C\) is the viscous damping coefficient, \(K(t)\) is the time-varying mesh stiffness, \(x\) is the relative displacement along the contact line, \(e(t)\) is the manufacturing error excitation, and \(W\) is the static normal load. The equivalent mass and static load can be written as
$$M=\frac{J_1J_2}{J_1r_{b2}^2+J_2r_{b1}^2}$$
$$W=\frac{T_1}{r_{b1}}=\frac{T_2}{r_{b2}}$$
Here \(J_1\) and \(J_2\) are the rotary inertias of the pinion and the gear, \(r_{b1}\) and \(r_{b2}\) are the base radii, and \(T_1\) and \(T_2\) are the torques applied to the two members. I separate the mesh stiffness, the relative displacement, and the static load into constant and fluctuating parts:
$$K(t)=K+\Delta k,\qquad x=X+\Delta x,\qquad W=KX$$
Substituting these expressions into the motion equation and neglecting higher-order terms gives
$$M\Delta\ddot{x}+C\Delta\dot{x}+K\Delta x=-\Delta k X+\Delta k\Delta x-\Delta k e+K\Delta e$$
This equation shows that the dynamic behavior of a hypoid bevel gear pair is driven by three major internal excitations: a time-varying stiffness excitation, a manufacturing-error excitation, and a combined meshing-impact excitation. The measured transmission error of a hypoid bevel gear pair after lapping belongs to the manufacturing-error category. Therefore, transmission error is not merely a geometric index. It is an excitation source that can directly influence the NVH performance of the main reducer. I also note that installation error and load deflection must be considered. A hypoid bevel gear tooth surface is usually a local conjugate surface, and a properly designed transmission error curve can absorb small mounting and load-induced deviations. If transmission error is too small, the gear pair may become too sensitive to installation error. If transmission error is too large, the meshing impact and noise may increase. This is why the relationship between transmission error and NVH is not strictly monotonic.
The transmission error of a hypoid bevel gear pair can be defined as the difference between the actual angular position of the driven gear and the ideal angular position calculated from the gear ratio. I use the following expression:
$$TE=\left(\phi_2-\phi_2^{(0)}\right)-\frac{z_1}{z_2}\left(\phi_1-\phi_1^{(0)}\right)$$
In this equation, \(\phi_1\) and \(\phi_2\) are the rotation angles of the pinion and the gear, \(\phi_1^{(0)}\) and \(\phi_2^{(0)}\) are the initial rotation angles at the start of measurement, and \(z_1\) and \(z_2\) are the tooth numbers of the pinion and the gear. For a hypoid bevel gear pair, the transmission error measurement is commonly performed by single-flank testing. In single-flank testing, only one tooth flank is in contact at a time, and the angular displacement of the input and output shafts is recorded by high-resolution encoders. The measured deviation is then converted into a transmission error value. This method is close to the actual meshing condition of a hypoid bevel gear pair under light load, and it is very sensitive to tooth profile conjugation. In my study, I record the forward-side transmission error and the reverse-side transmission error separately. The forward side corresponds to the drive side under acceleration, and the reverse side corresponds to the coast side.
A transmission error curve can be represented as a harmonic series:
$$TE(\theta)=\sum_{n=1}^{N}A_n\cos\left(n\theta+\varphi_n\right)$$
where \(A_n\) and \(\varphi_n\) are the amplitude and phase of the \(n\)th harmonic, and \(\theta\) is the angular position of the hypoid bevel gear. The first-order harmonic is usually the dominant component for NVH evaluation, but higher-order components can also produce distinct noise peaks. In practical production, I record the first-order transmission error because it is stable, repeatable, and closely related to the overall meshing quality of the hypoid bevel gear pair. The main geometric parameters of the hypoid bevel gear pair I studied are listed in Table 1.
| Parameter | Pinion | Gear |
|---|---|---|
| Tooth number | 10 | 43 |
| Transverse module | 4.186 | 4.186 |
| Pitch diameter | Not applicable | 180 mm |
| Midpoint spiral angle | 45°16′ | 27°3′ |
| Spiral direction | Left hand | Right hand |
| Addendum | Not applicable | 1.33 mm |
| Whole tooth height | Not applicable | 8.82 mm |
| Pitch deviation | ±0.018 μm | ±0.02 μm |
| Cumulative pitch error | 0.036 μm | 0.063 μm |
| Offset | 25 mm downward | 25 mm downward |
| Mean pressure angle | 21°15′ | 21°15′ |
| Normal backlash | 0.06 to 0.18 mm | 0.06 to 0.18 mm |
| Contact patch along tooth length | 50% to 70% | 50% to 70% |
| Contact patch along tooth height | 55% to 75% | 55% to 75% |
I also studied how tooth profile error changes during lapping. I used a high-precision coordinate measuring machine to record the helix angle error, pressure angle error, profile error, and cumulative pitch error before and after lapping. The changes are summarized in Table 2. The helix angle change was less than 1.5 arcmin, and the pressure angle change was less than 4 arcmin. The profile change was less than 4 μm. However, the cumulative pitch error changed by about 11 μm for both the pinion and the gear. This was much larger than the profile change. Therefore, among the geometric errors I examined, the pitch error had the strongest influence on the transmission error of the hypoid bevel gear pair. Lapping significantly reduced the transmission error value, which confirms that lapping can correct the meshing performance of a hypoid bevel gear pair.
| Geometric error item | Typical change before and after lapping | Observation |
|---|---|---|
| Helix angle error | Less than 1.5 arcmin | Small change |
| Pressure angle error | Less than 4 arcmin | Small change |
| Profile error | Less than 4 μm | Small change |
| Cumulative pitch error | About 11 μm | Dominant change |
| Transmission error | Strong reduction after lapping | Lapping is effective |
The pitch error of a hypoid bevel gear pair can shift the actual meshing point along the contact path. If the pitch deviation along the normal direction is \(\Delta t\), then
$$\Delta t=t_2-t_1$$
where \(t_1\) and \(t_2\) are the pitch values of two mating teeth. A positive or negative pitch deviation shifts the contact point earlier or later. This breaks the symmetry of the ideal parabolic transmission error curve and creates a composite curve with unequal amplitudes. The result is uneven load sharing, edge contact, and meshing impact. Even if the overall transmission error value is within a certain limit, a local pitch error can still produce a sharp impact and a high-frequency noise peak. This explains why a hypoid bevel gear pair with a small transmission error value can sometimes generate a poor NVH response. The transmission error measured by a rolling inspection machine is an accumulated angular error over many teeth. It does not fully describe the meshing behavior of every single tooth. Therefore, I must evaluate the hypoid bevel gear pair together with its contact pattern, pitch error, and installation condition.
The noise generation mechanism of a main reducer is closely related to the internal excitation of the hypoid bevel gear pair. The gear pair acts as a vibration source, and the housing, bearings, shafts, and axle structure act as transmission paths. The radiated noise is the final result of the structural response. The sound pressure level is commonly expressed as
$$L_p=10\log_{10}\left(\frac{p^2}{p_0^2}\right)$$
where \(p\) is the measured sound pressure and \(p_0\) is the reference sound pressure, usually \(2\times10^{-5}\) Pa. Sound power and sound intensity are also useful descriptors:
$$W=\frac{P}{t}$$
$$I=\frac{W}{S}=wc$$
In these expressions, \(W\) is sound power, \(P\) is radiated energy, \(t\) is time, \(I\) is sound intensity, \(S\) is area, \(w\) is energy density, and \(c\) is the speed of sound. Human hearing is not equally sensitive to all frequencies, so I use A-weighted sound level, written as dB(A), to evaluate the NVH performance of the main reducer. A-weighting attenuates low-frequency and middle-frequency components and emphasizes the frequency range to which the human ear is most sensitive. For the rear-axle main reducer in my study, I set the following NVH criteria: the peak noise under fifth-gear acceleration should be below 60 dB(A), and the peak noise under fifth-gear coast should be below 65 dB(A), allowing for a normal experimental tolerance of about 2 to 3 dB(A).
I first conducted road tests to observe how different transmission error values affect the main reducer NVH performance. I used a commercial NVH acquisition platform with microphones and speed sensors. The microphone was placed near the rear seat, close to the main reducer. The speed sensor recorded the rotational speed of the powertrain. Because the fifth-gear ratio of the transmission was 0.799 and the pinion had 10 teeth, the meshing order of the hypoid bevel gear pair was
$$c=\frac{10}{0.799}=12.52$$
This means that the main reducer noise appears mainly at the 12.52 order. The road test conditions are listed in Table 3.
| Main reducer condition | Engine speed range | Purpose |
|---|---|---|
| Fifth-gear acceleration | 2000 to 3600 rpm | Drive-side meshing and acceleration NVH |
| Fifth-gear coast | 3600 to 2000 rpm | Coast-side meshing and coast NVH |
To make the road test meaningful, I controlled other variables carefully. I ensured that the contact patterns of all hypoid bevel gear pairs were located in the middle-to-toe region and had good consistency. I also controlled the machining accuracy of the differential case and the reducer housing. Five special control parameters were defined: runout of the gear mounting surface, coaxiality of the bearing positions, bore diameter, offset, and perpendicularity between the pinion axis and the gear axis. The tolerances and improved pass rates are shown in Table 4.
| Special control item | Component | Required value | Pass rate before improvement | Pass rate after improvement |
|---|---|---|---|---|
| Mounting surface runout | Differential case | 0.04 mm | 89.2% | 100% |
| Bearing position coaxiality | Differential case | 0.016 mm | 56.2% and 57.7% | 100% |
| Bore diameter | Reducer housing | 72 mm with negative tolerance | 87.3% | 99.1% |
| Offset | Reducer housing | 25 ± 0.02 mm | 95.1% | 100% |
| Perpendicularity of axes | Reducer housing | 0.04 mm | 67.6% | 100% |
After the contact pattern and housing accuracy were controlled, I selected ten hypoid bevel gear pairs and assembled them into main reducers for road testing. The gear pairs had different transmission error values, but their contact patterns were all acceptable and consistent. The measured forward-side transmission error, reverse-side transmission error, fifth-gear acceleration peak noise, and fifth-gear coast peak noise are listed in Table 5.
| Gear set | Forward-side TE (μrad) | Fifth-gear acceleration peak noise (dB(A)) | Reverse-side TE (μrad) | Fifth-gear coast peak noise (dB(A)) |
|---|---|---|---|---|
| 1 | 17 | 68.67 | 15 | 68.49 |
| 2 | 10 | 60.30 | 7 | 63.09 |
| 3 | 7 | 71.39 | 5 | 70.87 |
| 4 | 22 | 72.62 | 14 | 71.43 |
| 5 | 16 | 69.54 | 10 | 67.55 |
| 6 | 13 | 60.34 | 8 | 65.46 |
| 7 | 15 | 67.51 | 16 | 69.52 |
| 8 | 17 | 66.87 | 17 | 70.84 |
| 9 | 20 | 70.61 | 15 | 67.45 |
| 10 | 11 | 61.36 | 13 | 68.91 |
The road test results show that a smaller transmission error does not automatically produce better NVH performance. For example, one hypoid bevel gear pair had a forward-side transmission error of 7 μrad, but its acceleration peak noise was 71.39 dB(A). Another hypoid bevel gear pair had a forward-side transmission error of 10 μrad, and its acceleration peak noise was 60.30 dB(A). This confirms that the relationship between transmission error and main reducer NVH is not monotonic. The hypoid bevel gear pair is locally conjugate, and the main reducer assembly inevitably contains small installation errors. A certain amount of transmission error is needed to absorb these errors and to avoid edge contact. If transmission error is too small, the contact pattern may become very sensitive to load and mounting variation. If transmission error is too large, the meshing impact increases. Therefore, I concluded that there is a suitable range rather than a single minimum value.
Because road testing is expensive and limited by route, driver, weather, and vehicle availability, I used a rear-axle NVH bench to obtain more data. The bench included a drive motor, an eddy-current loader, a silent enclosure, a temperature control system, and the same NVH acquisition platform used in the road test. The test conditions were set to simulate fifth-gear acceleration and fifth-gear coast. The bench parameters are summarized in Table 6.
| Bench parameter | Drive motor | Eddy-current loader |
|---|---|---|
| Power | 132 kW | 132 kW |
| Rated speed | 3150 rpm | 750 rpm |
| Maximum speed | 7000 rpm | 3000 rpm |
| Voltage | 380 V AC | 380 V AC |
| Oil temperature control | 30 to 140 °C | 30 to 140 °C |
| Background noise | ≤25 dB(A) | ≤25 dB(A) |
| Interior noise | ≤35 dB(A) | ≤35 dB(A) |
Before using the bench data, I verified its reliability. I selected three hypoid bevel gear pairs that had already been tested on the road and tested them again on the bench. The comparison is shown in Table 7. The peak noise values and the speeds at which the peaks occurred were close to the road test results. Therefore, I considered the bench test reliable enough for comparative analysis.
| Gear set | Road acceleration peak noise (dB(A)) | Bench acceleration peak noise (dB(A)) | Road coast peak noise (dB(A)) | Bench coast peak noise (dB(A)) |
|---|---|---|---|---|
| 1 | 68.67 | 66.07 | 68.49 | 67.41 |
| 2 | 60.30 | 59.95 | 63.09 | 61.82 |
| 3 | 71.39 | 63.01 | 70.87 | 62.06 |
For order tracking, I used the rotational speed signal as the reference. The relationship between meshing frequency, order, and speed is
$$c=\frac{60f}{n_1}$$
$$f=\frac{zn_2}{60}$$
$$c=z\frac{n_2}{n_1}$$
where \(f\) is the meshing frequency, \(n_1\) is the reference shaft speed, \(n_2\) is the gear speed, and \(z\) is the tooth number. For the hypoid bevel gear pair, the order is fixed at 12.52 in fifth gear. This allows me to separate the main reducer noise from other noise sources. I also used FFT and digital order tracking. The Fourier transform pair is
$$F(f)=\int_{-\infty}^{+\infty}f(t)e^{-j2\pi ft}\,dt$$
$$f(t)=\int_{-\infty}^{+\infty}F(f)e^{j2\pi ft}\,df$$
For the discrete Fourier transform, the radix-2 FFT relation can be written as
$$X(k)=X_1(k)+W_N^kX_2(k)$$
$$X\left(k+\frac{N}{2}\right)=X_1(k)-W_N^kX_2(k)$$
where \(W_N^k\) is the twiddle factor. The frequency resolution, sampling frequency, and Nyquist frequency are
$$\Delta f=\frac{1}{T}=\frac{1}{n\Delta t}$$
$$f_{\text{sample}}=\frac{1}{\Delta t}$$
$$f_{\text{nyquist}}=\frac{f_{\text{sample}}}{2}$$
To perform order tracking with a varying speed, I approximate the shaft angle as a quadratic function of time:
$$\theta(t)=a_0+a_1t+a_2t^2$$
When three reference pulses arrive at times \(t_1\), \(t_2\), and \(t_3\), the angle increments are equal. The arrival times can be obtained from
$$t_k=\frac{1}{2a_2}\left[-a_1+\sqrt{a_1^2-4a_2\left(a_0-k\Delta\Phi\right)}\right]$$
where \(k\) is the interpolation index and \(\Delta\Phi\) is the constant angle increment. This digital order tracking method avoids the aliasing and averaging problems that occur when the speed changes rapidly. I used it to extract the 12.52-order noise from the main reducer under acceleration and coast conditions.
With the bench test, I collected thirty hypoid bevel gear pairs with different transmission error values. Their forward-side transmission error ranged from about 3 to 27 μrad, and their reverse-side transmission error ranged from about 3 to 18 μrad. The corresponding main reducer peak noise values under simulated fifth-gear acceleration and coast are listed in Table 8.
| Gear set | Forward-side TE (μrad) | Acceleration peak noise (dB(A)) | Reverse-side TE (μrad) | Coast peak noise (dB(A)) |
|---|---|---|---|---|
| 1 | 17.05 | 68.67 | 15.54 | 68.49 |
| 2 | 10.67 | 59.30 | 7.67 | 63.09 |
| 3 | 18.32 | 68.39 | 16.21 | 71.87 |
| 4 | 22.56 | 69.62 | 14.44 | 70.43 |
| 5 | 16.06 | 69.54 | 10.02 | 67.55 |
| 6 | 13.99 | 60.34 | 8.98 | 65.46 |
| 7 | 15.27 | 64.51 | 16.90 | 69.52 |
| 8 | 17.55 | 66.87 | 17.57 | 70.84 |
| 9 | 20.08 | 69.61 | 15.00 | 67.45 |
| 10 | 8.23 | 55.36 | 13.00 | 66.47 |
| 11 | 11.67 | 56.80 | 3.87 | 65.89 |
| 12 | 19.19 | 63.29 | 11.34 | 67.27 |
| 13 | 26.63 | 70.54 | 5.56 | 63.88 |
| 14 | 9.17 | 56.73 | 12.64 | 68.44 |
| 15 | 10.88 | 58.34 | 10.54 | 65.89 |
| 16 | 12.44 | 59.16 | 4.88 | 64.78 |
| 17 | 7.32 | 56.87 | 6.36 | 64.01 |
| 18 | 6.58 | 65.45 | 8.34 | 65.08 |
| 19 | 21.22 | 71.19 | 3.15 | 68.34 |
| 20 | 5.31 | 63.28 | 7.88 | 63.18 |
| 21 | 14.69 | 63.45 | 4.45 | 64.19 |
| 22 | 13.20 | 62.08 | 12.14 | 66.78 |
| 23 | 8.34 | 58.61 | 13.67 | 67.46 |
| 24 | 10.13 | 59.22 | 5.66 | 63.07 |
| 25 | 11.05 | 58.18 | 4.25 | 62.77 |
| 26 | 22.51 | 68.73 | 6.17 | 63.39 |
| 27 | 23.03 | 65.41 | 7.15 | 64.14 |
| 28 | 24.89 | 66.07 | 9.57 | 66.92 |
| 29 | 25.17 | 70.66 | 10.89 | 67.22 |
| 30 | 7.87 | 60.18 | 12.05 | 68.55 |
I fitted the bench data with a smooth spline model. The fitting error was evaluated by the sum of squared errors:
$$\mathrm{SSE}=\sum_{i=1}^{n}\left(y_i-\hat{y}_i\right)^2$$
The smoothing parameter was adjusted until the fitted curve was smooth and the SSE remained small. The forward-side and reverse-side fitted curves showed clear valleys. When the forward-side transmission error of the hypoid bevel gear pair was between 8 and 12 μrad, the corresponding fifth-gear acceleration peak noise was below 60 dB(A). When the reverse-side transmission error was between 4 and 8 μrad, the corresponding fifth-gear coast peak noise was below 65 dB(A). These ranges are the suitable transmission error ranges for the main reducer NVH performance in my study. The result is summarized in Table 9.
| Condition | Suitable hypoid bevel gear transmission error range | Main reducer NVH criterion |
|---|---|---|
| Forward side under fifth-gear acceleration | 8 to 12 μrad | Peak noise below 60 dB(A) |
| Reverse side under fifth-gear coast | 4 to 8 μrad | Peak noise below 65 dB(A) |
After identifying the suitable range, I focused on how to make the hypoid bevel gear pair remain within that range after lapping. Lapping is the final finishing process. It removes a small amount of metal from the tooth surface, improves surface finish, corrects minor geometric errors, and reduces transmission error. The lapping process uses a lapping liquid containing abrasive particles. When the hypoid bevel gear pair rotates under a controlled braking torque, the abrasive particles are pressed and rolled between the tooth flanks. The cutting action depends on the normal force, the relative sliding speed, and the number of lapping cycles. The normal force can be approximated from the applied torque and the contact radius:
$$F=\frac{T}{R}$$
The tangential and normal components are
$$F_1=F\cos\beta$$
$$F_t=F\sin\beta$$
$$F_n=F_1\cos\alpha=F\cos\beta\cos\alpha$$
$$F_r=F_1\sin\alpha=F\cos\beta\sin\alpha$$
where \(\alpha\) is the pressure angle and \(\beta\) is the spiral angle. The cutting depth of the abrasive particles can be estimated by
$$a_p=\sqrt{\frac{F_{n0}}{\pi\sigma_s}2\tan^2\varepsilon}$$
where \(F_{n0}\) is the average normal force on a single abrasive particle, \(\sigma_s\) is the yield strength of the gear material, and \(\varepsilon\) is the half-cone angle of the active cutting part of the abrasive particle. This expression shows that the cutting depth increases with the normal force and decreases with material strength. In practice, the normal force on the abrasive particles is influenced by the tooth side clearance, the spindle speed, and the lapping cycle number. Therefore, I optimized these three parameters.
The tooth side clearance is a key parameter for the meshing of a hypoid bevel gear pair. If the clearance is too large, the gear pair may produce impact and noise. If the clearance is too small, the tooth surfaces may wear rapidly or even seize. The spindle speed controls the relative sliding speed between the tooth flanks. A higher spindle speed usually produces finer cutting and a smoother surface, but an excessively high speed may cause over-lapping and damage the tooth surface. The lapping cycle number controls the total cutting time. Too few cycles leave the tooth surface insufficiently corrected, while too many cycles remove too much material and worsen the transmission error. I designed a three-factor, three-level orthogonal experiment. The factors and levels are listed in Table 10.
| Level | Factor A: tooth side clearance (mm) | Factor B: spindle speed (rpm) | Factor C: lapping cycles |
|---|---|---|---|
| 1 | 0.05 | 1200 | 2 |
| 2 | 0.06 | 1800 | 5 |
| 3 | 0.07 | 2400 | 8 |
The orthogonal array and the measured transmission error results are shown in Table 11. Each row represents one lapping condition. The forward-side and reverse-side transmission error values were measured after lapping. Lower values are not always better, but for process optimization I aimed to bring the hypoid bevel gear pair into the suitable range of 8 to 12 μrad forward and 4 to 8 μrad reverse.
| Test number | Combination | Clearance (mm) | Spindle speed (rpm) | Cycles | Forward-side TE (μrad) | Reverse-side TE (μrad) |
|---|---|---|---|---|---|---|
| 1 | A1B1C1 | 0.05 | 1200 | 2 | 17 | 14 |
| 2 | A1B2C2 | 0.05 | 1800 | 5 | 16 | 12 |
| 3 | A1B3C3 | 0.05 | 2400 | 8 | 15 | 8 |
| 4 | A2B1C2 | 0.06 | 1200 | 5 | 15 | 16 |
| 5 | A2B2C3 | 0.06 | 1800 | 8 | 10 | 7 |
| 6 | A2B3C1 | 0.06 | 2400 | 2 | 18 | 13 |
| 7 | A3B1C3 | 0.07 | 1200 | 8 | 15 | 14 |
| 8 | A3B2C1 | 0.07 | 1800 | 2 | 10 | 5 |
| 9 | A3B3C2 | 0.07 | 2400 | 5 | 21 | 11 |
I used range analysis to determine the influence order of the three factors. For each factor and each level, I calculated the sum and mean of the measured transmission error. The range is
$$K_i=\sum_{j=1}^{m}y_{ij}$$
$$\bar{K}_i=\frac{K_i}{m}$$
$$R=\max\left(\bar{K}_1,\bar{K}_2,\bar{K}_3\right)-\min\left(\bar{K}_1,\bar{K}_2,\bar{K}_3\right)$$
The factor with the largest range has the strongest influence. The range analysis results are shown in Table 12. For both forward-side and reverse-side transmission error, the influence order was B, C, A. That is, spindle speed had the largest effect, followed by lapping cycles, and then tooth side clearance. The preferred combination was B2C3A2: spindle speed at level 2, lapping cycles at level 3, and tooth side clearance at level 2.
| Evaluation index | Factor A range | Factor B range | Factor C range | Influence order | Preferred combination |
|---|---|---|---|---|---|
| Forward-side TE | 1.67 | 6.00 | 4.00 | B > C > A | B2C3A2 |
| Reverse-side TE | 1.33 | 5.00 | 3.33 | B > C > A | B2C3A3 |
Because spindle speed had the strongest influence, I performed a further optimization of the spindle speed. I fixed the tooth side clearance at 0.06 mm and the lapping cycle number at 5, and varied the spindle speed from 1500 to 2400 rpm in steps of 100 rpm. For each speed, I lapped five hypoid bevel gear pairs and calculated the average forward-side and reverse-side transmission error. The results are shown in Table 13.
| Spindle speed (rpm) | Average forward-side TE (μrad) | Average reverse-side TE (μrad) | Observation |
|---|---|---|---|
| 1500 | 19.0 | 14.4 | Too high |
| 1600 | 17.8 | 12.4 | Too high |
| 1700 | 17.0 | 11.8 | Too high |
| 1800 | 15.4 | 10.0 | Still high |
| 1900 | 12.6 | 9.4 | Close to range |
| 2000 | 9.6 | 8.2 | Within target range |
| 2100 | 11.8 | 9.0 | Slightly above range |
| 2200 | 14.6 | 9.0 | Too high |
| 2300 | 15.0 | 10.8 | Too high |
| 2400 | 17.6 | 11.0 | Too high |
The best spindle speed was 2000 rpm. At this speed, the average forward-side transmission error was 9.6 μrad and the average reverse-side transmission error was 8.2 μrad. Both values were close to the suitable ranges. Therefore, the final optimized lapping parameters were: tooth side clearance of 0.06 mm, spindle speed of 2000 rpm, and lapping cycle number of 5. These parameters produced a hypoid bevel gear pair with forward-side transmission error mainly between 8 and 12 μrad and reverse-side transmission error mainly between 4 and 8 μrad.
I validated the optimized lapping parameters by producing a new batch of hypoid bevel gear pairs and assembling them into main reducers for road testing. Five gear pairs were randomly selected from the batch. Their forward-side and reverse-side transmission error values and the corresponding main reducer peak noise values are listed in Table 14.
| Gear pair | Forward-side TE (μrad) | Fifth-gear acceleration peak noise (dB(A)) | Reverse-side TE (μrad) | Fifth-gear coast peak noise (dB(A)) |
|---|---|---|---|---|
| 1 | 10 | 59.63 | 5 | 63.03 |
| 2 | 12 | 60.20 | 6 | 66.30 |
| 3 | 11 | 61.38 | 6 | 64.75 |
| 4 | 9 | 62.31 | 8 | 65.53 |
| 5 | 10 | 60.01 | 7 | 63.43 |
The validation results show that the optimized lapping process kept the hypoid bevel gear pair within the suitable transmission error range. The fifth-gear acceleration peak noise was about 60 dB(A) or lower, and the fifth-gear coast peak noise was below 65 dB(A), except for a small experimental variation of about 2 to 3 dB(A). This confirms that controlling the transmission error of the hypoid bevel gear pair within the identified range can improve the NVH performance of the main reducer. The combination of process control, contact pattern consistency, housing tolerance control, and lapping parameter optimization produced a stable and repeatable improvement.
In my study, the most important conclusion is that the transmission error of a hypoid bevel gear pair should not be treated as a value that is always better when smaller. Instead, it should be controlled within a suitable range. For the rear-axle main reducer I studied, the suitable forward-side transmission error range was 8 to 12 μrad, and the suitable reverse-side transmission error range was 4 to 8 μrad. When the hypoid bevel gear pair was within these ranges, the main reducer peak noise under fifth-gear acceleration was generally below 60 dB(A), and the peak noise under fifth-gear coast was generally below 65 dB(A). The optimized lapping parameters were tooth side clearance of 0.06 mm, spindle speed of 2000 rpm, and lapping cycle number of 5. These parameters reduced the transmission error of the hypoid bevel gear pair and brought it into the suitable range. The results provide a practical reference for controlling hypoid bevel gear transmission error and improving main reducer NVH performance in mass production.
From a theoretical point of view, the vibration model I established shows that the hypoid bevel gear pair generates internal excitation through time-varying mesh stiffness and manufacturing error. The transmission error is an important part of the manufacturing-error excitation. The dynamic response of the main reducer depends on the interaction among the gear pair, the bearings, the housing, and the axle structure. A very small transmission error may reduce the excitation amplitude in an ideal system, but it may also make the contact pattern too sensitive to installation error and load deflection. A very large transmission error increases the meshing impact and noise. Therefore, an optimal range exists. My road tests and bench tests both support this conclusion. The road test data showed that some hypoid bevel gear pairs with small transmission error had poor NVH performance, while some with moderate transmission error had good NVH performance. The bench test data, which covered thirty gear pairs, provided a clearer statistical trend. The smooth spline fitting identified the valleys where the main reducer NVH performance was best.
From a manufacturing point of view, the final lapping process is the most direct way to adjust the transmission error of a hypoid bevel gear pair. The orthogonal experiment showed that spindle speed is the most influential parameter. This is because spindle speed controls the relative sliding speed and the cutting behavior of the abrasive particles. A moderate increase in spindle speed improves the finishing effect and reduces the transmission error. However, an excessively high spindle speed causes over-lapping and may damage the tooth surface, which increases the transmission error again. The lapping cycle number is the second most influential parameter. Too few cycles cannot correct the tooth surface, while too many cycles remove too much material. The tooth side clearance is also important, but its influence is smaller than the other two parameters. The best combination I found was 0.06 mm clearance, 2000 rpm spindle speed, and 5 lapping cycles.
The NVH bench test also proved to be a valuable supplement to road testing. Road testing is the most realistic method, but it is expensive and difficult to repeat many times. The bench test can simulate the fifth-gear acceleration and coast conditions, and it can measure the main reducer noise at the fixed 12.52 order. The order tracking method removes interference from other noise sources and makes the comparison among different hypoid bevel gear pairs more reliable. The good agreement between the road test and bench test for the same gear pairs gave me confidence in the bench results. Therefore, I could collect enough data to fit the relationship between transmission error and main reducer peak noise. This would have been difficult with road testing alone.
In future work, I would like to extend the method to more operating conditions, such as different loads, temperatures, and vehicle speeds. The current study focused on fifth-gear acceleration and coast because they are the most representative conditions for the main reducer NVH problem. However, real driving conditions are more complex. I would also like to include more lapping parameters, such as lapping path and braking torque, in the optimization. The current orthogonal experiment considered tooth side clearance, spindle speed, and lapping cycles. The lapping path and the applied torque also affect the cutting depth and the final tooth surface. In addition, the relationship between transmission error and NVH may be influenced by the housing stiffness, bearing preload, and lubricant. A more comprehensive model could include these factors and provide a more complete understanding of the hypoid bevel gear main reducer system.
Overall, my research demonstrates that the transmission error of a hypoid bevel gear pair is a key factor for main reducer NVH. By establishing a vibration model, measuring tooth profile error before and after lapping, conducting road tests, building an NVH bench, applying order tracking, fitting the experimental data, and optimizing the lapping process, I identified a suitable transmission error range and a practical way to achieve it. The suitable range was 8 to 12 μrad on the forward side and 4 to 8 μrad on the reverse side. The optimized lapping parameters were 0.06 mm tooth side clearance, 2000 rpm spindle speed, and 5 lapping cycles. These results can help manufacturers of hypoid bevel gear pairs improve the consistency of transmission error, reduce main reducer noise, and enhance the NVH performance of rear-drive vehicles.