| UNIVERSITY OF WATERLOOFaculty of Engineering | |

Steering Wheel Vibration Warranty Reduction
The Development of a Shimmy Quality Gate

Honda of Canada, Manufacturing
Alliston, Ontario

Prepared by
Timothy Chak
ID# 20132585
3B Mechanical Engineering
September 17, 2007

4700 Tottenham Rd
Alliston, ON
L9R 1A2

September 17, 2007

Prof. Roydon A. Fraser
Associate Chair Undergraduate Studies
Department of Mechanical and Mechatronics Engineering
University of Waterloo
Waterloo, Ontario
N2L 3G1

Dear Prof. Fraser:

This report entitled, “Steering Wheel Vibration Warranty Reduction,” was prepared as my 3B Work Report for Honda of Canada, Mfg. This is my fourth work term report and the first one that I have prepared for Honda. The purpose of this report is to determine the required equipment and parameters to increase detection for steering wheel vibration, with a focus on the Acura MDX.

Honda Motor Co. is the fourth largest automaker in the world, with cars in its line-up such as Canada’s best-selling Honda Civic and the Honda Accord. Including the vehicles in its luxury brand, Acura, Honda sells 15 different cars and trucks in Canada. The Honda Civic, Honda Ridgeline, and Acura MDX are all made on the same assembly line in HCM plant 2 in Alliston.

Ali Woodley manages the vehicle quality department, in which I was employed. For my project, however, I worked under Mike Kueneman, who is the group leader of the vehicle quality analysis group.

This report was written entirely by me and has not received any previous academic credit at this or any other institution. I would like to thank Mr. Roy Matsui and the rest of the Product Engineering Chassis Group team members for providing me with valuable advice and technical information and Mr. Raymond Foo for the technical support he provided for the MDT equipment.

Sincerely,

Timothy Chak
20132585
Table of Contents List of Figures iii List of Tables iv Summary v 1.0 Introduction 1 2.0 Shimmy Detection 3 2.1 Current Detection Method 3 2.2 Limitations of Current Detection Method 3 2.3 Proposed Detection Method 3 2.4 Proposed MDT Equipment Changes 6 2.4.1 Temporary Changes 6 2.4.2 Permanent Changes 7 3.0 Determination of Parameters 8 3.1 Setup of the Master Vehicle 8 3.2 Road Data – the Customer Experience 8 3.2.1 Worst Case Speed 9 3.2.2 One-Wheel Static Imbalance 9 3.2.3 Phasing – Even and Uneven Weight Distribution 10 3.3 Detection on the MDT 12 3.4 Substantiating Process Change 14 3.5 System Capability 15 3.5.1 Vehicle Studies 16 4.0 Conclusions and Recommendations 19 References 20 Appendix A: Road Data 21 Appendix B: MDT Resolution Trial 26 Appendix C: Online Non-Invasive Trial 29

List of Figures Figure 1: Warranty Claims for 2007 Acura MDX 1 Figure 2: Typical Vehicle Steering System 2 Figure 3: Shake vs. Shimmy 2 Figure 4: Vibration Measured in a Time-based Analyzer 4 Figure 5: Vibration Measured in an FFT Analyzer 4 Figure 6: Proposed Jig 5 Figure 7: Old Jig 5 Figure 8: Temporary and Permanent Changes to the MDT 6 Figure 9: Worst Case Speed on Highway 9 Figure 10: Road Imbalance Data 10 Figure 11: 20 grams in LS & RS FR Wheels - 40% of Max Expected Shimmy 11 Figure 12: 80 grams in LS & RS FR Wheels - 75% of Max Expected Shimmy 11 Figure 13: 40 grams in LS & 20 grams in RS FR Wheels - 60% of Max Expected Shimmy 12 Figure 14: 50 grams in LS & 10 grams in RS FR Wheels - 74% of Max Expected Shimmy 12 Figure 15: MDT Resolution Trial - Shimmy vs. Imbalance 13 Figure 16: Effect of Side Roller 14 Figure 17: Effects of Hands on Steering Wheel 15 Figure 18: Capability Analysis - Sample 1 17 Figure 19: Capability Analysis - Sample 2 18

List of Tables Table 1: Selection of the Shimmy Master Wheels 8 Table 2: Effects of Hands on Steering Wheel 14 Table 3: Non-Invasive Trial Procedure 16

Summary
During the 2007 model year, shimmy was a problem that plagued the newly redesigned Acura MDX. Measures were currently being taken to reduce outflow of vehicles with unacceptable levels of shimmy at both the supplier and at Honda – during the final inspection. However, a stack-up of the imbalances from various steering components in addition to minor process variations is to be expected. Also, the outside drive was being done at a low speed and only a small stretch of road was smooth enough to perform shimmy testing. Lastly, detection was subject to the individual’s judgement, which varies from person to person.
It was proposed that the subjectivity be removed by performing the detection with machines and instruments – an accelerometer was used in conjunction with an FFT analyzer as a way to objectively measure shimmy.
For testing purposes, a master vehicle was set up in which the wheels were replaced with the best four out of ten tire/wheel assemblies randomly chosen from the supplier’s safety stock. The entire front axle was then balanced to 0 grams imbalance.
In order to find the failure parameters, road data had to be measured first to understand the problem at the customer level. In a speed sweep, it was found that the worst case speed was approximately 120 kph. For that speed, a predictive formula was calculated for shimmy as a function of imbalance: shimmy=0.002248×imbalance+0.01229
Tests for phasing showed that the effects of phasing would be present (as expected). However, statistically speaking, there was a relatively low probability that it would have a significant impact on detection capabilities.
In the MDT trials, it was shown that the detection process only had a capability of 30% due to measurement variation and low resolution due to the detection environment. This low capability was apparent in the large overlap in measured values. For example, the highest value for a 0 gram imbalance could have been measured as the average shimmy for 20 grams imbalance. Interestingly enough, when the vehicle made contact with the side rollers, all the values averaged out to be approximately the same, regardless of the imbalance or manner in which contact was made.
A non-invasive online trial was performed in order to determine the system’s capability before implementation. The results for a failed unit showed good accuracy, with measured and predicted imbalances only being off by 4 grams. But precision was still poor because the possible range of values it could have measured was relatively high (as proven in the MDT trials). For a passable unit, both precision and accuracy was low. However, the unit still passed both the online analysis and the road test. In addition, the quality gate was not meant to determine exact shimmy values, but simply to reject no good units. Therefore, this system – with the parameters acquired from the various tests and trials – was deemed ready for implementation. 1.0 Introduction
The automotive industry is extremely competitive, and in recent years, quality has become an ever-increasing expectation of the average car buyer. It is said that the average consumer demands aerospace quality, but will only pay bargain-basement prices in order to obtain it. This is especially true of the automotive industry, where there is an almost permanent goal of reducing costs while improving quality.
One of the problems that plagued the newly redesigned 2007 Acura MDX was steering wheel vibration (also known as “shimmy”). The costs of the warranty claims up to July 17, 2007 were $70,560.94. Projected for the full year, that total warranty cost would be more than $80,000.

Figure [ 1 ]: Warranty Claims for 2007 Acura MDX

Shimmy is a rotational vibration at the steering wheel that occurs at highway speeds and is caused by vehicle sensitivity to tire and wheel force variation. It is the result of natural frequencies in the steering system due to force variations, run-outs, and imbalance of the tire/wheel assembly (TWA), and is a function of system characteristics such as mass and damping. Spatial interactions within the steering system provide a path for vibrations at the TWA to travel to the steering wheel and ultimately to the driver. Various components within the vehicle’s chassis affect the generation and provide a transfer path for shimmy, which originates in the road and is felt by the driver. The sensitivity of this system will greatly determine the perceived amount of vibration felt by the driver. A sensitive system is one in which shimmy occurs despite being within reasonable non-uniformity and imbalance tolerance at the wheel-ends. In the case of the MDX, the vehicle is highly sensitive to non-uniformity and imbalance.
The back and forth movement in the wheels are translated to rotational movement in the steering wheel through the rack and pinion gear [Figure 2].

Figure [ 2 ]: Typical Vehicle Steering System
It should be noted that shimmy is only the rotational vibration and the scope of this project did not include shake, or vibration in the vertical direction, which is caused by dynamic imbalance [Figure 3].

Shake
Shimmy Mode Figure [ 3 ]: Shake vs. Shimmy
In the customer’s viewpoint, shimmy can cause discomfort – especially when the vehicle is driven over long distances – and is an unacceptable manufacturer defect. In addition, shimmy, which is an “unexpected vibration” causes wear in the steering system and may also negatively affect gas consumption. The purpose of this project is to increase in-house shimmy detection in order to reduce steering wheel vibration market complaints and to improve initial quality. 2.0 Shimmy Detection
2.1 Current Detection Method
Measures are currently taken to reduce outflow of vehicles with unacceptable levels of shimmy at both the supplier and at Honda during the final inspection. The tier-2 supplier inspects all tires for non-uniformity, and the tier-1 supplier matches the high-point (the heaviest point) of the tire with the low-point (the lightest point) of the rim. After the tire and wheel are fitted together, the imbalance of the tire/wheel assembly (TWA) is measured and wheel weights are added in order to bring the TWA to an imbalance of ≤ 7 grams.
In the final inspection at Honda, every vehicle is driven on an outside track, a process called “outside drive.” During this time, the associates check for vibration, and vehicles found to be unacceptable would have its wheels replaced with a new set.
2.2 Limitations of Current Detection Method
In practice, no process can be guaranteed to work 100% effectively, 100% of the time. The TWA balancing operation at the supplier is relatively capable, but measurement error can result in no good units passing inspection. As previously mentioned, there are other components that also affect the total axle imbalance that result in shimmy. In addition, the TWA supplier does not check for tire uniformity and there is currently no inspection process specifically for this problem at Honda.
In the outside drive, the shorter track is always used for all vehicles produced in plant 2. The resonant frequency for shimmy occurs at a speed of 120 kph for the MDX. At this speed, the vibration must be ≤ 0.2g. However, very low speeds (~50 kph) are used on the outside drive. The amount of vibration that can be detected at these speeds is very minimal, and it is very difficult for the associate to be able to reliably detect any shimmy, and then project that vibration amount to that which would be measured at the critical speed of 120 kph. To make matters worse, vibration detection is subject to the environmental conditions and detection is done best on very long stretches of smooth road. Even if the outside drive is done at 120 kph and on a long stretch of smooth road, the fact remains that the current outside drive detection process is subject to the individual’s judgement, which may differ from person to person.
2.3 Proposed Detection Method
The purpose of this project is to increase the in-house detection of steering wheel vibration. A repeatable and accurate testing method must be developed that reduces or eliminates human judgement errors while minimizing the effect on the current process cycle time. It was proposed that steering wheel vibration be detected by means of instrumentation during the dynamic testing of the vehicle (on the MDT). Any equipment required must be robust (in order to withstand usage in an industrial production environment) and be lightweight (in order to minimize vibration damping).
In a previous trial, a system consisting of a jig embedded with an accelerometer and a time-based signal analyzer was used. This system produced many false failures; however, some proposed equipment changes can be made in order to reduce the number of false failures. The time-based analyzer recorded vibration over time. However, the system inherently cannot differentiate between vibration due to shimmy and other sources of vibration [Figure 4].

Figure [ 4 ]: Vibration Measured in a Time-based Analyzer
Vibration at other frequencies does not have an effect on shimmy detection.
In the proposed system, detection was done with an FFT analyzer. This is because shimmy vibration only occurs at ~15 Hz. An FFT analyzer is able to discriminate based on frequency, resulting in increased confidence in an accurate detection. With this system, a reduction in false failures was to be expected due to the assurance that the analytical comparison was only being done on the vibration due to the area of concern [Figure 5].
Analysis is done only in shimmy frequency

Figure [ 5 ]: Vibration Measured in an FFT Analyzer
Also in the previous trial, the jig that was used was heavy and dampened some of the vibration. This decreased the precision of the measurement because the resolution was reduced. In other words, the difference between a passed and a failed unit became smaller by a certain percentage due to damping; however, the measurement variation remained absolute. Resultantly, potential capability of the system was vastly reduced. The proposed jig [Figure 6] was a redesign of the old one [Figure 7]. The design was simplified and the “start inspection” button was removed, resulting in a reduction of cycle time by approximately 1 second compared with the old system. Figure [ 6 ]: Proposed Jig Figure [ 7 ]: Old Jig

2.4 Proposed MDT Equipment Changes Figure [ 8 ]: Temporary and Permanent Changes to the MDT
The detection method requires changes to the PLC and computer software on the MDT. Due to the short timeframe and the actual amount of time required to make these changes, a request was put in for temporarily implementation. This was because the software changes required to add an additional inspection item had to be done by EG-Japan.
2.4.1 Temporary Changes
The temporary changes required a Hands off Steering Wheel (SW) and Hands on SW message during the 30 mph speed check and the acceleration to 40 mph step, respectively. When the associate is able to hold 30±2 mph during the speed check, the PLC would then trigger the FFT analyzer to start collecting 1 second of data after a 1 second delay. The FFT will then analyze the vibration data and determine whether the car is good or no good and send a pass or fail signal back to the PLC, which would light up a green or red lamp. If the red lamp was light up, the associate would record a failure for shimmy on the final inspection card (FIC) and the vehicle would be repaired.
2.4.2 Permanent Changes
The proposed permanent changes are similar to the temporary changes. However, a new inspection item would be added to the MDT software. In other words, indication of a passed or failed unit would be on the computer screen with the other inspection items rather on the green and red lamps. In addition, the results of the inspection would be stored with the vehicle identification number (VIN).

3.0 Determination of Parameters
The equipment selection and setup was only a small portion of the project. A large portion of the project was the determination of the feasibility and parameters of the detection process. In addition, further testing was required in order to validate the effectiveness and capability of the process.
There are essentially four parts to the feasibility, parameter, and capability determination: * Feasibility of detection on the MDT * Correlation study between MDT and road data * Process verification * Capability determination 4.1 Setup of the Master Vehicle
A master vehicle was required in order to properly and accurately determine the detection parameters. Usually, special TWAs would be ordered with extremely low radial harmonics, and static and dynamic imbalances. It would usually take months for the TWAs to be verified by the manufacturer and shipped to the plant. However, due to the timeframe of the project, a different avenue was needed to acquire these “master wheel assemblies.” Ten TWAs would be measured and the four TWAs with the best radial harmonics would be selected, as shown in Table 1.
Table [ 1 ]: Selection of the Shimmy Master Wheels Tire | Dynamic | Static | H1 | H2 | H3 | 1 | 8.0,7.5 | 2.5 | 5 kg | 2 kg | 2 kg | 2 | 6.5,8.5 | 5.5 | 5 kg | 0 kg | 4 kg | 3 | 3.0,2.5 | 1.5 | 9 kg | 3 kg | 2 kg | 4 | 12.0,17.5 | 4.5 | 9 kg | 1 kg | 2 kg | 5 | 10.520.5 | 9.5 | 6 kg | 0 kg | 1 kg | 6 | 10.5,20.5 | 9.5 | 6 kg | 2 kg | 1 kg | | 10.0,17.0 | 6.5 | | | | 7 | 19.5,25.0 | 3.5 | 5 kg | 3 kg | 2 kg | 8 | 10.0,14.0 | 6.0 | 4 kg | 1 kg | 1 kg | 9 | 6.5,8.0 | 2.5 | 8 kg | 3 kg | 1 kg | | 5.0,8.5 | 3 | 7 kg | 3 kg | 1 kg | 10 | 12.5,20.0 | 9 | 5 kg | 4 kg | 1 kg |

These wheels would then be put on the wind-noise master vehicle and an on-car (or axle) balance would be performed on the front wheels (since shimmy only originates from the front axle). This essentially ensures that the entire front axle has a total imbalance of exactly 0 grams. 4.2 Road Data – the Customer Experience
It was of upmost importance to collect vibration data from driving on the road in order to understand the problem from the customer’s point of view. This data would then be used to find the critical speed and imbalance that generates vibration that is 0.167g, which effectively gives us the parameters that can be later used when testing on the MDT in order to correlate simulated (MDT) data with actual (road) data. All the road data is as shown in Appendix A.
3.2.1 Worst Case Speed
Steering wheel vibration is a function of the road speed. As the radial frequency of the wheels approach the resonance frequency of the vehicle’s chassis, the amount of shimmy will increase. To find the critical (or worst case) speed, a speed sweep was performed on the highway and three measurements were taken for each speed between 80 and 140 kph, in 10 kph increments. A static imbalance of 30 grams was added to the left-side front wheel at the RFV point.

Figure [ 9 ]: Worst Case Speed on Highway
The following formula was found to fit the curve with an R2 = 0.9962: y = 3.0000152·10-9·x5 - 1.6678114·10-6·x4 + 3.6508516·10-4·x3 - 3.9338871·10-2·x2 + 2.0882222·x - 43.7034103
Setting its first derivative to zero, and solving for x: y'(x) = 1.50001·10-8·x4 - 6.67124·10-6·x3 + 0.00110·x2 - 0.07868·x + 2.08822 = 0
We find that the speed with the worst case of shimmy was 122.99 kph. For our purposes, this was close enough to 120 kph.
3.2.2 One-Wheel Static Imbalance
For this portion of the trial, weights were added to the left-side front wheels in order to simulate the respective total axle static imbalance. The purpose of this test was to confirm that shimmy is approximately a linear function of mass imbalance (in order to determine its predictability). In addition, this data would help determine what mass imbalance induces the upper shimmy limit of 0.167g.

Figure [ 10 ]: Road Imbalance Data
As the data shows, data on the second peaks (the highest peak between 13 and 16 Hz), had an approximately linear correlation, with a correlation coefficient of 0.96421. Therefore, using the formula for the line of best fit, the total axle static imbalance required to induce a vibration of 0.167g is approximately 70 grams.
3.2.3 Phasing – Even and Uneven Weight Distribution
Static imbalance is a function of both tires and the phase difference between the two has an effect on the amount of shimmy that can be detected (based on the relative position of one imbalance to the other). Although the probability of both tires being imbalanced by exactly the same amount is low, it is still important to understand the effects of phasing to better understand the customer experience and consider these effects when setting up the detection parameters. Using the same stretches of the same highway, wheel weights were added to both front wheels this time. At least 15 measurements were randomly taken for each set of imbalances ranging from 20 to 60 grams (on each wheel). For each set, a maximum expected value was projected using the data from the one-wheel static imbalance trial, and the measured values were calculated as a percentage of this theoretical maximum. As expected, the measured shimmy varied with phase angle. The measurements tended to average between 40 and 75% of the maximum expected value, as seen in the charts in Figures 11 and 12:

Figure [ 11 ]: 20 grams in LS & RS FR Wheels - 40% of Max Expected Shimmy

Figure [ 12 ]: 80 grams in LS & RS FR Wheels - 75% of Max Expected Shimmy
This data is better understood when the weights were added unevenly. As previously mentioned, the probability of both wheels being imbalanced by exactly the same amount – and therefore having a lower expected limit of 0g – was very low. It is more realistic to have different imbalances, and therefore having a lower limit that is not 0. It was expected that with a larger imbalance difference, the variation would be smaller since the lowest amount of imbalance, which occurs when the imbalances are 180° out of phase, is the difference between the two wheel imbalances (i.e. if both wheels are 30 grams, the lowest expected imbalance is 30-30 = 0 grams; if the wheel imbalances are 10 and 50 grams, the lowest expected imbalance is 50-10 = 40 grams). This theory was confirmed in the road test, as shown in Figures 13 and 14:

Figure [ 13 ]: 40 grams in LS & 20 grams in RS FR Wheels - 60% of Max Expected Shimmy

Figure [ 14 ]: 50 grams in LS & 10 grams in RS FR Wheels - 74% of Max Expected Shimmy 4.3 Detection on the MDT
Because varying road and weather conditions made it difficult to get reliable and repeatable results from the outside drive, it was decided that it would be best to carry out the shimmy detection indoors. Indoor detection resulted in two choices: build an indoor track, or setup quality gate in MDT. Since an indoor track would have cost a lot of money, setting up the quality gate in the MDT was the obvious choice. However, even though the MDT sufficiently simulated driving conditions for the current tests being done, it presented a challenge with shimmy detection. Firstly, the MDT utilized rollers. The rollers kept the car in place while the dynamic tests were performed. However, this created two contact points instead of one and theoretically, it would prevent the back-and-forth motion of the wheel that creates shimmy. The other challenge was that the speed in which these dynamic tests were performed was very low. The only speed that is held for a sufficient period of time that is required for the shimmy data collection was 30 mph. As previously mentioned, the amount of detectable vibration is much less and the resolution would correspondingly be very low.
In the MDT resolution trial, the vehicle was driven at 30 mph to simulate the operating conditions in mass production. Mass imbalances of 0 to 50 grams, in 10 gram increments were placed in the middle of the front left wheel to simulate a total axle static imbalance. Because there was a slight slope to the right, all vehicles tended to drift towards the right side at higher speeds. Therefore, six measurements were taken with the wheels on random points of the rollers and five measurements were taken with different ways that the wheel can come in contact with the side rollers e.g. touching the side rollers throughout the data collection phase, wheels turned to hit and bounce off of the side rollers, the vehicle being allowed to drift slowly and touch the side rollers, and a sudden adjustment back to the left immediately after contact with the side rollers.
As expected, there was a linear relationship between imbalance and shimmy, and there was a large amount of measurement variation. (The raw data is shown in Appendix B.) Calculations showed that the system only had a capability of 30%. This can be seen in the large overlaps of data e.g. the highest value for 0 grams imbalance could have been measured as 20 grams imbalance.

Figure [ 15 ]: MDT Resolution Trial - Shimmy vs. Imbalance
An unexpected result of the trial was found in the side roller contact portion. It was found that when the vehicle made contact with the side rollers, it averaged out to almost the same number, regardless of the imbalance subjected to the vehicle and manner in which the vehicle came in contact with the rollers [Figure 16]. It is therefore extremely important that the vehicle remain within the centre portion of the rollers and not be allowed to drift far enough to make contact with the side rollers. Otherwise, the risk of a false pass is largely increased.

Figure [ 16 ]: Effect of Side Roller 4.4 Substantiating Process Change
In an earlier test to determine the damping effects of the shimmy jig on vibration detection, it was realized that even a relatively small inertial change had an effect on the measured amount of steering wheel vibration. After examining the inspection process, it became clear that allowing the associates to leave their hands on the steering wheel during the data collection process may introduce uncontrollable measurement variation, which reduces the system’s overall detection capability. A push/pull gage was used to ensure a controlled normal force was applied to the steering wheel. The purpose was to determine whether the amount of force that the associate uses to grip the steering wheel and/or the force exerted by their hands resting on the steering wheel produced a significant amount of variation in the readings. Forces of 0 to 8 kg·f, in 2 kg·f increments, were applied to the bottom of the steering wheel. The results are as shown below:
Table [ 2 ]: Effects of Hands on Steering Wheel holding force [kg·f] | shimmy [g] | | holding force [kg·f] | shimmy [g] | | holding force [kg·f] | shimmy [g] | 0 | 0.0133 | | 4 | 0.0091 | | 8 | 0.01 | | 0.0127 | | | 0.0091 | | | 0.0066 | | 0.0112 | | | 0.0122 | | | 0.0104 | | 0.013 | | | 0.0075 | | | 0.0074 | | 0.0089 | | average | 0.009475 | | | 0.0082 | average | 0.01182 | | | 80.2% | | average | 0.00852 | | | | 6 | 0.008 | | | 72.1% | 2 | 0.0108 | | | 0.0089 | | | | | 0.0107 | | | 0.0083 | | | | | 0.0096 | | | 0.0077 | | | | | 0.0084 | | average | 0.008225 | | | | average | 0.009875 | | | 69.6% | | | | | 83.5% | | | | | | |

Figure [ 17 ]: Effects of Hands on Steering Wheel
In Figure 17, the dashed, light blue line indicates the measured shimmy with no normal force on the steering wheel, but with an imbalance. The dotted, darker blue line is the measured shimmy with no normal force and no imbalance. (Signal noise is the reason why this is not 0.) It can be seen that even with a small grip force of 2 kg·f, the resolution has already halved. At a typically grip force of 4-6 kg·f, the resolution is now only a quarter of the maximum, hands-off steering wheel measurement. Therefore, it is concluded that for better accuracy, it is best to remove the hands from the steering wheel during the data collection process. 4.5 System Capability
Before implementation, it is always a good practice to hold a trial run. However, because of a backlog of vehicles that had to have its dynamics retested – on top of normal production – a different, non-invasive trial had to be designed and executed. The plan was to test the shimmy equipment, confirm its capabilities, and collect information that could be used for proper equipment setup, refinement of operating standards, and/or for future implementation purposes. An equipment study was to be done on 75 MDX’s. The study is designed to be as non-invasive as possible and did not affect cycle time or change any of the current processes.
The procedure is as follows:

Table [ 3 ]: Non-Invasive Trial Procedure Item | Responsibility | Vehicle enters the MDT | Associate | Fit shimmy jig onto steering wheel | Tim C | Rev engine | Associate | Go over 47 mph | Associate | Perform speed check at 30 mph, remove hands, and hold speed for at least 2 seconds trigger fft to start collecting data record VIN and amt of shimmy (in g’s) | AssociateTim C | Go over 40 mph and press VSA OFF | Associate | Decelerate and put transmission into Neutral | Associate | Remove jig | Tim C |

The results are in Appendix C.
Approximately one in every 20 vehicles as well as any failed units were held for further analysis on the long-track. These vehicles were marked so that the exact phasing could later be recreated on the long track. The tire pressure was adjusted to 32 psi to simulate the customer driving conditions. An estimate of the amt of total imbalance at the phasing that the tires are marked at was found by using the following formula: predicted imbalance
Due to measurement error (from results taken from a previous study), the actual shimmy would be ±0.0075g, 92.97% of the time.
At this point, data was collected from the long track. The amount of imbalance was then predicted using the following formula: measured imbalance
The measurement error for the road data from a previous study was ±0.05699g. The predicted range of numbers was then compared with the measured range of data to determine if the numbers have been correlated correctly and if the equipment is capable. 4.6.1 Vehicle Studies
As was briefly mentioned, during the trial, three MDXs were chosen randomly for verification of the correlation between the road data and MDT data. After the vehicles were analyzed for shimmy, the tires were marked (in order to recreate the same phase) and the vehicles were sent directly to XV hold. (They did not go through water leak check in order to keep the markings intact.) First, an imbalance was predicted and a range was estimated based on the online analysis. Then, the vehicles were tested on the long track and the vibration measurements were taken. The actual imbalance would be calculated using data from previous studies and a range would be estimated for this. Any overlap in the ranges would indicate that the data was correlated correctly.
It could be expected, however, that if the ranges do not overlap, and that the predicted imbalance is less than the actual imbalance. This is because there is a lag in the display of live data on the fft analyzer by Ono Sokki that was used in this trial. This did not affect the shimmy measurement in determining the pass/failure of a vehicle. However, the shimmy numbers collected would not be completely accurate. As a result, an adjusted online measurement was calculated by adding the difference between the smallest shimmy value that the FFT analyzer determined NG (0.0146g) and the failure criteria (0.0236g), which was determined by interpolating the shimmy amount corresponding to 35 grams of imbalance – since that is the total stack-up tolerance for the entire axle.
The two vehicles that were studied were: VIN#500562 and VIN#500845.
For 500562:

Figure [ 18 ]: Capability Analysis - Sample 1

For 500845:

Figure [ 19 ]: Capability Analysis - Sample 2
From the first vehicle, we can see that the adjusted predicted range completely engulfs the measured range. This is very good – showing that the correlation was done correctly and accurately. In addition, the nominals were very close and were only 4 grams apart from each other.
In the second vehicle, the range still had overlap, but it was not nearly as accurate as the first vehicle. This is to be expected, as a low resolution detection is expected to yield less accurate results in the lower end. Since the exact imbalance on each vehicle is not required, but only a simple pass/fail determination, it can be concluded that the low accuracy on the lower end is not very important.

4.0 Conclusions and Recommendations
Using machines and instrumentation in a quality gate eliminates the subjectivity that comes with human judgement. After performing numerous tests and trials, the exact failure parameters were found that could be used with the proposed equipment. It was shown in the capability study that within the upper passable limits, detection was fairly precise. In other words, the predicted vs. the measured imbalances were only off by 4 grams. However, the limitations imposed by the environment were apparent in other tests, and it became evident that the detection environment had a large impact on the system’s potential capability. This was shown in the lack of accuracy – the range of values that it could have measured for the same vehicle and the same conditions was very large. Because of a range of factors such as a lower speed, two contact points, and the jig (to a smaller degree), the range of values is wider because the resolution is much smaller. However, due to the excellent accuracy in the upper passable region, this system is deemed ready to be implemented and used right away with little or no required modifications.
In the future, in order to improve accuracy, the speed should be increased. Ideally, detection should occur at 120 kph – where the worst case shimmy occurs. This provides the largest resolution, and therefore, best detection capability. In addition, in future plans, a belt-type MDT system would be a much better replacement for the current roller-type MDT system for the purposes of shimmy detection. Lastly, although there were only 3 false failures (4%) in a sample size of 75, the failure criteria can be tweaked slightly in order to reduce that number.

References
Yu, Brickner, Nutwell, and Johnson (2007). Analysis of Vehicle Chassis Transmissibility of Steering
Shimmy and Brake Judder: Mechanism Study and Virtual Design of Experiment – SAE 2007-01-2342. Warrendale: SAE International.

Kim, Park, and Lee (2005). Tire Mass Imbalance, Rolling Phase Difference,Non-Uniformity Induced Force Difference, and Inflation Pressure Change Effects on Steering Wheel Vibration – SAE 2005-01-2317. Warrendale: SAE International.

Bagley, M.R. (2005). Tire Mass Imbalance, Rolling Phase Difference, Non-Uniformity Induced Force Difference, and Inflation Pressure Change Effects on Steering Wheel Vibration. Massachusetts: Massachusetts Institute of Technology.

Appendix A: Road Data

Worst Case Speed Data: Speed [kmph] | Frequency [Hz] | Shimmy [g] | Average [g] | 80 | 16 | 0.034 | 0.026 | | 16.14 | 0.024 | | | 16.065 | 0.02 | | 90 | 10.7 | 0.025 | 0.029 | | 10.7 | 0.038 | | | 10.7 | 0.024 | | 100 | 13.4 | 0.029 | 0.033 | | 13.4 | 0.028 | | | 13.4 | 0.042 | | 110 | 13.4 | 0.066 | 0.062 | | 13.4 | 0.057 | | 120 | 16.1 | 0.061 | 0.082 | | 13.4 | 0.12 | | | 13.4 | 0.064 | | 130 | 16.1 | 0.079 | 0.076 | | 16.1 | 0.073 | | 140 | 13.4 | 0.039 | 0.041 | | 13.4 | 0.043 | |

One-Wheel Static Imbalance Data: Imbalance [grams] | Frequency [Hz] | Shimmy [g] | Average Shimmy [g] | | Imbalance [grams] | Frequency [Hz] | Shimmy [g] | Average Shimmy [g] | 30 | 13.4 | 0.0379 | 0.0785 | | 50 | 13.4 | 0.1507 | 0.1025 | | 16.2 | 0.0611 | | | | 13.4 | 0.1150 | | | 13.4 | 0.1166 | 0.0779 | | | 13.4 | 0.0629 | 0.1144 | | 13.4 | 0.0640 | | | | 13.4 | 0.1507 | | | 13.4 | 0.0311 | 0.1114 | | | 13.4 | 0.0459 | 0.1487 | | 16.2 | 0.0562 | | | | 13.4 | 0.0951 | | | 10.8 | 0.0487 | | | | 13.4 | 0.0857 | | | 13.4 | 0.0375 | | | | 13.4 | 0.0720 | | | 13.5 | 0.1481 | | | | 13.4 | 0.1446 | | | 16.2 | 0.1439 | | | 60 | 13.5 | 0.1767 | 0.1570 | | 13.4 | 0.1559 | | | | 16.2 | 0.1410 | | | 16.2 | 0.0903 | | | | 13.4 | 0.2388 | 0.1514 | | 13.5 | 0.0289 | | | | 13.4 | 0.2144 | | 40 | 13.4 | 0.0898 | 0.1118 | | | 13.4 | 0.1364 | 0.1990 | | 13.4 | 0.1533 | | | | 13.4 | 0.2040 | | | 13.4 | 0.0954 | 0.1101 | | | 13.4 | 0.1465 | | | 13.4 | 0.0951 | | | | 13.4 | 0.0676 | | | 13.4 | 0.1641 | 0.1373 | | | 13.4 | 0.0679 | | | 13.4 | 0.1478 | | | | 13.4 | 0.1072 | | | 13.4 | 0.0563 | | | | 13.4 | 0.1699 | | | 13.4 | 0.0786 | | | | 13.4 | 0.1823 | | | 13.4 | 0.1257 | | | | 13.4 | 0.2066 | | | | | | | | 13.4 | 0.1387 | |
Phasing – Even Weight Distribution

Phasing – Uneven Weight Distribution

Appendix B: MDT Resolution Trial

Appendix C: Online Non-Invasive Trial