Transporting NV Standardized Testing from the Lab to the Production Environment

Transcription

1 Transporting NV Standardized Testing from the Lab to the Production Environment G. Pietila and P. Goodes Sound Answers Copyright 2009 SAE International ABSTRACT NVH labs at Tier 1 and Tier 2 suppliers have in the past 15 years implemented testing techniques to validate the NV performance of their products. Validation tests are most often conducted to check compliance of the product to vehicle OEM specifications, less often to truly assess NV performance. Vehicle OEM specifications are sometimes outdated, narrow in scope and require labtype of conditions, i.e. anechoic or hemi-anechoic environment, fixed microphone position and specific operating conditions. Tier 1 and 2 suppliers often find themselves in a situation in which they have to transport lab tests to the plant floor, as they are requested by their customer to ensure 100% shipped quality. In order to do so, they are faced with several technical and commercial challenges as vehicle OEM specifications often have no provisions for plant-floor boundary conditions and cycle time requirements. For cost reasons, the component manufacturer typically aims at adding NV performance checks to an existing test station, which poses additional challenges over using a completely redesigned test station. This paper will discuss the challenges encountered by Tier 1 and Tier 2 suppliers having to implement end-of-line NV testing for quality control. The paper will discuss these challenges for a few different automotive components, namely steering wheel clock springs, sunroofs, axles and shock absorbers. The authors will describe some of the challenges faced and the solution implemented at the end of the line as it relates to boundary conditions, absolute levels at feedback sensor and pass/fail criteria INTRODUCTION As Tier 1 and Tier 2 suppliers are pushed by their customers to ensure 100% NVH quality they are forced to deploy lab based test specifications from the laboratory environment to the production floor. Unfortunately a large number of the published specifications are developed for a controlled lab environment and often do not include provisions to meet cycle times nor match boundary conditions in the production environment. The suppliers are faced with the need to develop a test standard that meets the requirements and intent of the OEM specification with the appropriate provisions to be implemented on the production floor. The following pages will highlight some of the challenges that are encountered in deploying laboratory specifications to a production environment for a variety of components. LABORATORY BASED SPECIFICATIONS In the majority of laboratory based specifications some reference is made to the boundary conditions or mounting configuration. These requirements often include compliance to ISO standards for acoustic measurements, such as ISO 3745 for sound power, or the reference to stringent drive systems, such as a quiet drive belt system to drive an automotive AC Compressor. For this reason the specifications, while challenging to meet in the laboratory, are simply not feasible for the production environment. In one situation an OEM specification requires the sound power of an AC compressor be reported according to The Engineering Meetings Board has approved this paper for publication. It has successfully completed SAE s peer review process under the supervision of the session organizer. This process requires a minimum of three (3) reviews by industry experts. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE. ISSN Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely responsible for the content of the paper. SAE Customer Service: Tel: (inside USA and Canada) Tel: (outside USA) Fax: CustomerService@sae.org SAE Web Address: Printed in USA

2 ISO 3745 (preferred) or ISO 3744 (accepted). When the supplier s 3.5 x 3.5 meter hemi-anechoic room was evaluated for compliance to ISO 3745, it was found that due to the size of the room and thickness of the wall panels (0.26 m) the room would only comply at frequencies over 500 Hz. When the room was requalified for ISO 3744, including the use of a three reflective plane test set-up, the size of the room still limited the lower frequency to 200 Hz (1). In addition to the room size requirements the background sound pressure levels inside of the chamber were evaluated, to ensure adequate separation between the Device Under Test (DUT) and background. In this case the hemi-anechoic chamber was located in a relatively quiet area in the warehouse section of the plant, with an ambient below 70 dba, shown in figure 1. Also shown in the figure is the difference in between the DUT and the interior background level. This difference indicates that the room complies with the background noise requirement at frequencies over 160 Hz with a correction factor, or over 400 Hz with no correction factor. Clearly if the same room was moved to the production line, the ambient surrounding the room could be expected to increase by 10 or even 20 db. This would obviously increase the interior background level to a point at which adequate separation between background and DUT could not be reached without significant room improvements. For reference the Noise Reduction of the room (calculated as difference between exterior and interior ) is also included in Figure 2. Figure 1: Background surrounding the anechoic chamber compared to the difference between DUT and interior background. Considering size of the room, the requirement for automated doors that seal properly and the need to automatically transfer the DUT into the chamber, it is not feasible to implement this test into an established production line. For this reason a supplier will typically develop a new test specification for the production environment that correlates to the Pass/Fail judgment of the laboratory test rather than to the absolute values of its results (such as sound power). To develop this production line specification it is important to understand the source of the failure modes. This will help in developing fixturing, measurement locations, transducers and test criteria that do not affect the performance of the DUT. PRODUCTION TEST CHALLENGES Generally the production line test specification is developed based on constraints from three general areas: Product, Production, and Customer, each of which having unique requirements. PRODUCT CONSTRAINTS In most cases a product to be tested inherently has a number of failure modes, which are excited during various modes of operation. In this case it is obviously necessary to cycle the DUT through each of the modes of operation during the production test. Once the failure modes have been identified, it is necessary to optimize the location and number of transducers. In the following example a ratchet sound was identified as a failure mode for an automotive sunroof and a study was done to optimize the transducer placement for this test. It was determined that the ratchet sound was caused by a cable rubbing on a metal corner, which could be identified by using an accelerometer or microphone. In this situation an accelerometer was used to eliminate the need for an acoustically isolated enclosure around the test stand. A waterfall plot of the ratchet event compared to no-ratchet event is included in figure 3. This figure clearly shows that the ratchet event is identified as an increased energy in the Hz frequency range. Accelerometers were also placed at various locations around the sunroof to optimize location and number of sensors needed to identify the ratchet event. Figure 4 compares the measured acceleration at four different locations that were tested. In this example a single accelerometer located at one of the locations shown as blue or green would be sufficient to identify the failure mode. Additional measurement locations would possibly be needed to identify additional failure modes. Figure 2 : Noise Reduction of an anechoic chamber.

3 Time Axis Ratchet The ability to detect unacceptable components can also be improved through the selection of an appropriate analysis technique, such as order analysis vs spectral analysis or envelope analysis. Time Axis Figure 3 : Waterfall plot of sunroof ratchet failure mode. No Ratchet 4 db ref 1 Impact type failure modes, including bearing faults, nicked gears or fan ticking, can often be difficult to identify in rotating components especially if the fault occurs at the same frequency that the device is spinning. Figures 6 and 7 compare the sound pressure signatures measured on 5 fans, two of which fail for a ticking noise. Figure 6 shows the frequency spectra for the five measurements using a standard FFT. This plot demonstrates the difficulty in separating the fan blade pass noise from the ticking noise, both of which occur at 38 Hz. In contrast to the simple FFT, if the same five measurements are re-analyzed using an amplitude demodulation method (2), envelope analysis in this example, the parts that fail for ticking can be clearly differentiated from the acceptable parts (Figure 7). Figure 4 : Comparison of four accelerometer locations for ratchet failure mode capture In addition to location, the type of transducer used can have a significant effect on the integrity and usefulness of the data. A common failure mode for an axle is the gear mesh order, which can be measured using a torsional accelerometer located on the axle pinion or with a linear accelerometer contacting the part. Figure 5 compares the 11 th (mesh) order measured using a linear accelerometer (black) to a torsional accelerometer (red) for the same axle tested three times. In this case one can see that the mesh order response has less variation if it is tracked using a torsional accelerometer, therefore the acceptance criterion using this sensor would be more robust. An additional example is shown in Figure 8. In this case there were three axles with various levels of nicks in the pinion gear teeth, and five axles with no nick. When the axles were tested at 400 rpm the two axles with a marginal nick (black and orange) could not be clearly separated from the good axles using an FFT or order analysis. As shown in figure 9, the first order response masks the gear nick response in all of the axles. Envelope analysis is able to identify the nicked axles by focusing on the higher frequency response that is excited by these impact-type events. Three measurements using Linear Accelerometer Three measurements using Torsional Accelerometer.2 Figure 6: Frequency Spectra of two bad (orange and blue) and three good (red, yellow and green) fans. Ticking Failure Figure 5 : Order Spectra of a torsional accelerometer (red) compared to a linear accelerometer (black) Figure 7 : Envelope Spectra of two bad (orange and blue) and three good (red, yellow and green) fans.

4 3 Nicked gears long as the boundary conditions that are used do not eliminate the DUT response due to the failure mode, and the difference between acceptable and non-acceptable components is measureable and repeatable. Figure 8 : Envelope Spectra of three axles with various levels of nicked pinion gears compared to five good axles. Figure 9: Frequency Spectra of three axles with various levels of nicked pinion gears compared to five good axles. One of the most important considerations in developing a test specification to be used in the production environment is the selection or design of the fixturing that will hold the component during the test. Figure 10, compares the gear mesh order response measured on the same sample on two test stands with different boundary conditions. As this plot shows the response measured on the DUT depends on the boundary conditions imparted by the test fixture. Both approaches shown in Figure 10 are valid as there was clear and repeatable separation between an axle with and without the gear mesh failure mode. 20 db ref 1u 2 Nicked gears 4 good gears 20 db (re 1u ) PRODUCTION CONSTRAINTS In addition to developing a test that can consistently excite the expected failure modes of a product, it is extremely important to consider the constraints that are inherent to testing in the production environment. These constraints include cycle time, ambient sound pressure or acceleration levels, component build status and existing station configuration. The cycle time constraint is relatively easy to understand in that the time to test the component cannot introduce a bottleneck into the production process. This time must include the time to test as well as the time required to load and clamp the component prior to testing. In addition to meeting this cycle time the fixture must be developed in such a way that a high level of maintenance is not required. Figure 11, compares the gear mesh order for an axle tested on the same test stand, before and after maintenance. The bottom plot compares the axle tested 6 times prior to maintenance and the top plot compares the axle tested 6 times after maintenance was performed, which includes tightening all of the joints and connections. In this case one can see that if the tolerances were set with tight connections when the system was initially configured, the criteria would no longer be valid once the joints and connections loosened up. RPM 1000 rpm 1000 rpm 0.2 Time Figure 11 : Comparison of two axle test stands after (top) and before (bottom) maintenance. Figure 10 : Comparison of the gear mesh order on two different axle test systems. In some cases the production test stand is designed to model the in-vehicle constraints, but at times this approach is not feasible. Either approach is valid as A second common constraint is the high ambient sound pressure or vibration levels that are inherent to the production environment. For this reason it is often necessary to isolate the test stand from the surrounding environment using isolation pads or soft mount suspension systems for vibration, or a sound chamber for sound pressure. In many cases the implementation of an adequate sound chamber is cost or space prohibitive, so it becomes

5 necessary to develop a tactile measurement, acceleration, velocity, etc, which correlates to the sound pressure based acceptance criteria. An example of this was experienced with automotive shocks, where a common failure mode is a squeaking sound generated by debris that gets lodged in the damper valves. This noise is easily heard inside of a vehicle and is generally a high pitched sound that occurs at high compression velocities. However it is much more difficult to audibly identify squeaky shocks in the production environment due to the high levels of noise generated by stamping presses, test machines and assembly equipment. For this reason a much more robust method to identify shock squeaks is through the use of an accelerometer mounted on the reserve tube. This type of measurements is much less influenced by the high ambient sound pressure levels, and is still able to clearly identify the various levels of shock squeaks. Figure 12 compares the band-pass filtered acceleration levels measured on the reserve tube of a production Force Test Stand. The black trace represents a good shock and the red, green and blues traces were shocks that failed for squeaks. A small sound chamber was designed and built around the test stand to isolate the microphone used for the acceptance criteria from the high ambient levels in the manufacturing environment. The sound chamber was not designed to qualify as an anechoic chamber, rather was intended to simply reduce the influence of the plant noise on the sound pressure measurements. Figure 13 compares the measured on the production line adjacent to the test stand to the measured in the sound chamber. The difference between the measurements indicates the Noise Reduction of the sound chamber. Figure 14 compares the background measured in the sound chamber to the measured level during a clock spring test. Previous testing in a quiet laboratory environment had shown that the scraping sound occurred at frequencies over 1000 Hz. Therefore the separation, between DUT and background, greater than at frequencies over 1000 Hz, is sufficient to identify acceptable and nonacceptable components. 12 db ref 1 Figure 13: Comparison of production line ambient vs production line sound chamber Figure 12: Band-pass filtered acceleration vs time for 1 good shock (black) and 3 squeaky shocks. Range used to evaluate clocksprings In some cases the component failure mode cannot be identified using tactile measurements and it is required that sound pressure measurements be used. An example of this is an automotive steering wheel clockspring. The clock-spring is located just under the steering wheel in vehicles so any scrape or click that it generates will be identified as an unacceptable noise by the OEM and end customer. The clock-spring scrape is a failure mode that occurs when the steering wheel is turned in one direction to the stop then turned all the way back past the neutral point. This motion causes the ribbon cable to flip from being wrapped around the inside to the outside of the ribbon guide. As the ribbon cable flips it can scrape against the clock-spring cover. This scraping motion is audible in the vehicle, and generates measureable acceleration levels on the cover. Unfortunately the test stand in which the production test was to be implemented had high ambient vibration levels that masked the scraping event, so the acceptance criteria was defined using sound pressure measurements. Figure 14: Comparison of clock-spring 1/3 rd octave spectrum vs sound chamber The OEM specification also mandated the use of loudness as an acceptance criterion for component testing in the lab environment. With respect to this specification, a loudness metric was used in the acceptance criteria in the production line. During the implementation of the system it was determined that the loudness measured on the production line (in the chamber) did not correlate to the absolute value of the loudness measured in lab environment, as seen in Table 1. This is due to the high level of low-mid frequency (<1kHz) background noise that is present in the production line sound chamber. To account for the low

6 frequency background level, a high pass filter was applied to the time history recordings taken in the production line sound chamber. This method was then validated initially using a sample of 11 clock-springs, 6 of which are shown in Table 1, then again with a sample set of ~45 parts. The validation procedure included subjective ratings of the clock-springs, Zwicker Loudness (3), Zwicker Loudness on filtered time history and Transient Loudness (3) on filtered time histories. The study shows that the filtered loudness metrics correlate well with the loudness that was measured in the laboratory environment. The study also shows that the Transient Loudness metric (based on DIN (4) ), which better accounts for non-stationary sounds, correlates much better to the subjective ratings of the components. By using filtered signals to compute loudness, or other standard metrics, correlation to rankings from subjective or objective lab measurements was achieved. Table 1: Comparison of the subjective rating and the measured loudness for automotive clock-springs A final concern when implementing a test specification into the production environment involves the performance of the test stand, especially when using an existing station. Production line NVH tests are often added to an existing performance test station, which cycles the DUT through a series of tests designed to insure the proper operation of the component when it is near build completion. Although these stations may operate a test cycle similar to that which would be implemented for an NVH test, the control parameters may be different. For example a load applied to a seat track during motion may affect the NVH performance significantly, but may not be important for an operational test. In one example, an axle test stand was generating data that had a large amount of variation, when testing the same part. After evaluating the test sequence and procedure it was determined that the test stand was allowing the axle wheel ends to differentiate during the test. Figure 15, compares the rpm profiles of the pinion speed compared to the wheel-end speeds, demonstrating the differentiation issue that was experienced. Although the functional test of the axle may include a differential test it is important to understand and control the amount of differentiation during an axle NVH test. RPM 1000 rpm Figure 15: RPM profiles for the pinion (red) and wheelends (blue and green) during an axle speed sweep test. CUSTOMER CONSTRAINTS In addition to the constraints that are inherent due the products design or the manufacturing environment in which it will be implemented, there are often a few additional constraints that are incurred due to the OEM or customer expectations. Most of these constraints are tied to the expected product performance for vibration or sound quality. There are times in which a particular test procedure can be conducted in the current production environment with a very high success rate in identifying particular failure modes, but the OEM or customer has an expectation that forces the use of a different method. In this situation it may become necessary to design and build an additional test station to test for NVH performance according to the customer expectations. Most often this type of situation is addressed by implementing two test stations. One in the production line that correlates to the customer s expectations and addresses the production constraints of cycle time, ambient sound pressure/vibration levels, etc, and a second off-line station that is used to test a subset of parts according to the customer specification. Occasionally it becomes necessary to develop a test stand around the customer s expectations and implement it into the production environment. This situation often leads to a perception based acceptance criteria, developed though the implementation of a Jury analysis. An example of this type of project is described in detail with respect to automotive seat track testing by Bernard T. et. al. (5) and Cerrato G. et. al. (6). In this case the customers expectation was defined as a sound pressure level as well as no objectionable noises. The criteria of objectionable noises was formally defined by implementing a Jury analysis during which a large sample set of seat tracks with various levels of acceptance were evaluated by a group, or Jury, of people. The Jury s preference ratings were then correlated to objective measurements to develop a preference equation. This preference equation was then implemented in a production station to predict the customer s subjective preference prior to shipping the product.

7 CONCLUSION As Tier 1 and Tier 2 suppliers are forced to transport lab based NV performance tests to the production environment there is a series of questions that should be addressed prior to developing the production based test station. These questions are centered on the challenges that are often faced in implementing the NV tests in a production environment. General concerns: Can the OEM (customer) test specification be directly implemented, or is it necessary to develop a new test method that correlates to the OEM specification? Product based concerns: What modes of operation excite all of the potential failure modes? Where are the optimal transducer locations to identify all failure modes? Is it necessary to identify specific failure modes or is a general DUT failure acceptable? What types of measurements give the most repeatable results? What analysis methods capture each particular failure mode? What boundary conditions allow for repeatable measurements, without adversely affecting the ability to identify the failure modes? Production based concerns: What test sequence meets the production line cycle time and excites all failure modes? Is it necessary to implement vibration or acoustic isolation material? Does the component build status raise concerns about failures due to lack of lubrication? Does the build status hide specific failure modes? What measurement types allow for the most significant separation between measurement and ambient sound pressure/vibration levels? Does an existing test station control the parameters that most affect NV performance? Customer based concerns: Do the OEM or customer expectations drive the acceptance criteria, or is it acceptable to use acceptance criteria that correlate to the OEM specifications? Is it necessary to develop an acceptance criteria based on customer subjective preference? As these questions are addressed prior to transporting NV testing out to the production floor, there will be significantly fewer interruptions to the production work flow during implementation of the production test system, and more meaningful NV data. REFERENCES 1. Pietila, G., Cerrato, G., Davis, K., Sumerton, S. Practical Comparison of ISO-3744 and ISO Sound Power Standards for Automotive Compressor Testing INCE NoiseCon Dearborn, MI, July Hans Konstantin-Hansen, Envelope Analysis for Diagnostics of Local Faults in Rolling Element Bearings, Bruel & Kjaer, Application Note 3. Maya Sound Quality Manual, version DIN 45631, Calculation of loudness level and loudness from the sound spectrum Zwicker method Amendment 1: Calculation of the loudness of time-variant sounds 5. Bernard, T., Braner, L., Cerrato-Jay, G., Davidson, R., Dong, J., Jay, M., Pickering, D., "The Development of a Sound Quality-Based End-of-Line Inspection System for Powered Seat Adjusters", SAE 2001 World Congress, Detroit, MI, March SAE# Cerrato, G., Crewe, A., Terech, J., "Sound Quality Assessment of Powered Seat Adjusters", SAE N&V Conference, Grand Traverse, MI, May SAE# CONTACT Glenn Pietila Sound Answers, Inc. (248) glenn.pietila@soundanswers.net