Application of NVH Techniques to Engine Production Line Test

Transcription

1 05NVC49 Application of NVH Techniques to Engine Production Line Test Copyright 2004 SAE International Alex Alvey Tecumseh Products Company Gabriella Cerrato Jay MTS Systems Corp. ABSTRACT A major manufacturer of small engines has recently implemented cold and hot start tests on the assembly lines of one of its engines. The cold start test refers to the test in which the engine is motored for a short time. Since this is a motored test, with no combustion occurring in the cylinder, the objective of this test is the detection of manufacturing and assembly defects. Next the engine undergoes a hot start test, in which fuel and air are supplied and the engine is started under a predefined load condition. This is a test performed to verify the performance of the engine under the application conditions. In both tests, typically multiple parameters are measured and extracted to assess the mechanical and thermal performance of the engine. Since noise level and sound quality of small engines have become major concerns for application manufacturers and end-users, an investigation was performed to assess the feasibility of including NVH parameters in the test criteria. Furthermore, some of the NVH parameters can be used to improve and refine the detection of mechanical defects in the cold start test and to detect abnormal combustion phenomena (knock, misfire) in the hot start test. This paper will describe the process and the tools used in the investigation along with all testing, signal processing and correlation methods. Conclusions related to the usefulness and feasibility of NVH parameters and to the efficiency of the test stands will also be presented. INTRODUCTION Tecumseh Products Company is a worldwide manufacturer of compressors for refrigeration, engines for lawnmower and generators, electrical motors, pumps, transmissions and transaxles for tractors etc. Manufacturing centers are located in North America, South America, Europe, and Asia providing customization to local markets with globally competitive costs and responsive delivery. Overseas production lines have to be populated with inspection stations in order to ensure that the manufacturing quality meets or exceeds its pre-existing standards (customer requirements). The inspection criteria and processes should be as consistent as possible throughout the entire product line and, for the same product, across different production facilities. This translates into the need for standardized quality control procedures and criteria which can then be implemented and deployed across the company using highly similar solutions (i.e. hardware and software). Several years ago, Tecumseh Products initiated the development of standardized tests for compressor kits. The compressor kit comprises the compressor moving parts, i.e. piston, connecting rod, cylinder, crankshaft and valves. The test criteria were developed at the Tecumseh Research and Development laboratory in Ann Arbor, MI, USA, and deployed in plants in Brazil and India. The compressor kit tester performs a functional test of the compressor kit before other more expensive parts are added or operations performed (i.e. the electrical motor and the welding of the housing and the lines). The main innovation of the compressor kit tester was that the compressor was tested with no oil in it. This of course could be done only by limiting the duration of the test to a few seconds, which posed some challenges from a signal processing and Pass/Fail criteria standpoint. The compressor kit tester is very similar to a cold start test for the long block of a small engine. The objective common to both tests is to identify manufacturing and assembly defects of the moving parts. Therefore, several of the lessons learned from the compressor kit tester development were applied in the development of the engine cold and hot start test. In addition, by adopting consistent procedures, similar hardware and same software platforms, the development cost of the engine cold start test was significantly reduced. In the following sections the authors will describe in detail the design and implementation of the cold and hot start test and how NVH parameters and signal processing tools are used as Pass/Fail criteria.

2 COLD AND HOT START TEST Traditionally, the engine industry, automotive and small engines, has relied on the hot start test for its 100% quality inspection process. In the hot start test, the engine is almost completely assembled, with oil and fuel, and it is run for a certain amount of time until conditions are stable, at which point performance and NVH data are acquired. By contrast, the so-called cold start test is conducted on an engine sub-system typically including all the reciprocating parts and the engine block. The engine in this case is motored, and no combustion occurs in the cylinders. The objective of the Cold Start test is essentially the inspection of the moving parts while the objective of the Hot Start test is the functional test of the complete engine. Therefore, the Cold Start test is run at an assembly station, not at the end-of-line, and allows the identification of defects earlier than the Hot Start test. With reference to Figure 1 in this page, the Hot Start test is done at the End-Of-Line while the Cold Start test is a Plant Diagnostic tool that can help to establish links between End-of-Line results and Front-of- Line engineering. There is an increasing manufacturing need for upstream in-plant diagnostics to reduce the amount of rejects at the end of the line. Another advantage of the cold start test is that it can be executed in a fraction (i.e. less that 1/10) of the time required for the hot start test. Therefore the Cold Start test can be implemented for 100% of the production with far less penalty on line throughput rate. DESIGN OEM Vehicle NVH Targets System NVH Targets DESIGN DESIGN VALIDATION YES NO Component NVH Targets COMPONENT SUPPLIER the hot start test is added at the end-of-line, to comply with customer quality and performance specifications. Based on the experience gathered with the compressor kit testers, the Tecumseh engineers were confident that the engine could sustain the motoring test with no oil for some time. Previous engine tests performed with and without oil had indicated that the same defects identified during tests with oil could be identified when no oil had been added to the engine. The Cold Start has been recently developed at Tecumseh Products Company based on the following requirements: 1. To catch defective engine parts in general (i.e. outliers that should not proceed further in the assembly line) 2. To identify the specific defect(s) responsible for the rejection in order to speed up the re-work of the rejected engine blocks. 3. The inspection tester should be developed using commercial, off-the-shelf, hardware and software products that can be easily ported to multiple facilities and tailored for product-specific needs 4. The inspection systems have to be supported and maintained locally, with no need for engine or NVH specialists on site. IMPLEMENTATION OF THE COLD START TEST ENGINE The engine under test is a single cylinder, 4-stroke engine used in lawnmowers, snowblowers, and generators. The engine long block, which is tested in the Cold Start test, is shown in Figure 2. The engine has a bore of 71 mm, a stroke of 49 mm, giving an overall displacement of 195 cc.. Depending on the application, it can have a vertical or horizontal crankshaft and the test stand has to support both types of mounting. FOL Engineering PROTOTYPING Qualified Parts DEVELOPMENT Investigate BENCHMARK NO PLANT DIAGNOSIS MANUFACTURING YES Quick re-work MANUFACTURING Fail EOL Quality Control Pass Figure 1. NVH Targets and Engineering Process The approach taken by Tecumseh Products Company was to develop and implement first the engine cold start test as a 100% quality check. The cold start improves the first pass yield of the hot test and does so before significant value has been added to the product. Then TEST SETUP Figure cc engine long block

3 The engine is manually loaded and clamped onto the test fixture. The operator also mounts the transducers. The crankshaft of the engine (vertical) is connected to the crankshaft of the servo motor with a self-aligning mechanism. The engine is instrumented with a multi-pulse encoder, a torque transducer on the driving shaft, an accelerometer on the cylinder head and a pressure transducer in the cylinder. The engine is driven at a steady and low RPM for a few seconds. The four signals (pressure, acceleration, torque and encoder) are acquired in the time and crank angle domain and stored to disk. The test automatically shuts down and the operator measures the leakage of intake and exhaust valves. DEFECT DETECTION The raw time and crank angle domain data are used to compute the parameters which identify the defects. National Instruments LabView was used for data acquisition, post-processing and application of pass/fail criteria (6). An example of the raw data captured during a test is given in Figure 5. The first prototype of the Cold Start test system for a single-cylinder, 4-stroke engine is shown in Figure 3 below. The transducers setup is shown in Figure 4. Engine long block Figure 5. Example of Cold Start Test display during test Torque transducer Figure 3. Prototype of cold start test system for small engine Pressure sensor and accelerometer Figure 4. Transducers setup in Cold Start test The defects are identified by applying typical NVH signal processing techniques to the signals acquired during the test (1,7). The parameters extracted from the recorded signals which are used for defect detection are: Statistics/features of the in-cylinder pressure Statistics/features of band-passed acceleration Statistics of the RPM function versus time. Frame Kurtosis of band-passed acceleration Start and running torque function All functions and derived parameters used in the Cold Start test are time and/or crank angle domain-based. No frequency analysis is performed. The analysis of the frequency spectra of cylinder head vibration of several good and bad engines revealed early on that, while the frequency spectrum of the measured signals could be used to set a threshold for Pass/Fail, it did not contain enough information to identify specific defects. This can be seen by looking at the frequency spectra shown in Figure 6. The solid line is the average acceleration spectrum from several good engines. The dashed spectrum is the average acceleration from several engines with defect #3 and the dash-dotted spectrum the result of averaging the vibration of several engines with defect #2. While both engines with defect exhibit much higher level of vibration than the good

4 engines for f > 1500 Hz, it is not possible to identify specific defects from the spectral content alone. Good engine Engine w/ defect #2 Engine w/ defect #3 Figure 6. Frequency spectra of acceleration of good engine (solid), of engine with defect 3 (dashed) and engine with defect 2 (dash-dotted). See Table 1. The defects are identified by a look up table, where abnormalities in specific parameters (pressure, torque etc.) are correlated to specific defects. These defects were first identified by the manufacturing engineers as the most frequent ones. Then a large DOE matrix of tests of engines with artificially introduced defects and of good engines was conducted to identify relationships between test data and defects. The defects that the system is currently capable of detecting are listed in Table 1. Most defects can be detected by extracting features from the in-cylinder pressure signal. An example of the effect of a defect on the in-cylinder pressure is shown in Figure 7, where the in-cylinder pressure for 3 consecutive combustion events (over 6 crankshaft revolutions) is displayed. The figure shows the pressure of a good sample and the pressure from a sample with a crack in the cylinder head gasket. Other defects can be better detected by using the acceleration signal. An example is shown in Figure 8, where the dashed signal is the acceleration measured on an engine with the gasket missing (higher peaks) and the solid line is the acceleration measured on a good engine. For other defects, the torque curve bears more information, as shown in Figure 9, where the higher torque is measured on an engine with connecting rod bearing installed backwards. In several cases, there is not a unique correspondence between parameter and defect and the current effort is aimed at identifying, for each defect, the dominant parameter. This correlation can be investigated with traditional statistical approaches, with neural networks or with auto-adaptive algorithms. Some of these approaches are currently under investigation at Tecumseh Products. In addition, more work is being done to broaden the range of defects that can be detected. ID Description of defect 1 First ring missing 2 Second ring missing 3 Compression rings bent/broken 4 Secondary ring inverted 5 MCR not functional 6 Head gasket missing 7 2 head gaskets (instead of 1) 8 cylinder head bolts not torques enough 9 rod bearing cap installed backwards 10 connecting rod bore oversize 11 main journal machined over/undersized 12 camshaft timing 13 piston installed backwards 14 rod bearing installed backwards 15 oil seal on crankshaft missing 16 connecting rod bore undersize Table 1. List of defects currently identified by Cold Start Test Pressure (psig) Time (s) Figure 7. Effect of crack in head gasket: dashed is good sample, solid is defective sample Vibration (m/s^2) Good engine Defective engine Crankangle 400 (Degree) Figure 8. Acceleration of engine with crack in head gasket (dashed) and good engine (solid)

5 -80 Friction Torque (PSIG) Crankangle (Degree) Figure 9. Effect of rod bearing installed backwards: solid is good sample, dashed is defective sample. IMPLEMENTATION OF THE HOT START TEST The transition from cold to hot start test is important especially because of its NVH implications. In the hot start, the engine is fully operational and its noise, vibration and sound quality can be measured and compared to target values. Even when an engine passes the Cold Start station, manufacturing and assembly variability may still affect its overall NVH performance, therefore this is checked at the Hot Start station. At this station, the inspection system measures engine vibration, noise, torque, RPM, exhaust and intake temperature, and emission characteristics. The system is very similar to its Cold Start counterpart, in its overall design and in the software and hardware tools. Figures 10a and 10b below show the wavelet spectrograms of the noise measured with two different engines at the Hot Start station (4). The differences especially in the high frequency range are evident and are mainly attributable to differences in the combustion process. Figure 10b. Wavelet spectrogram of engine during Hot Start test New Pass/Fail criteria are defined for the Hot Start test. The NVH criteria are set for the vibration signal and Sound Quality metrics are computed for the microphone signal. This analysis is performed over a relatively short time (5 seconds) and it does not impact the overall duration of the Hot Start test. The implementation on the same line of Cold and Hot Start test also offers considerable potential for engine source separation and target refinement Several references have been published on the subject of separating mechanical from combustion noise in different types of engine (single and multiple cylinders, 2- and 4-strokes, gasoline and diesel) (2,3). The objective of the current investigation at Tecumseh is to evaluate the applicability of these laboratory techniques to the manufacturing environment. The correlation between Front-of-Line and End-of-Line testers relies on the interrogation of the databases and is still work in progress due to the recent implementation of the Hot Start test. CONCLUSION A Cold Start test has been successfully implemented at an engine manufacturing plant. This inspection system screens for assembly and manufacturing defects and provides the line operator with a lists of suggested actions. A Hot Start test has also been implemented as 100% quality inspection at the end-of-line. In both functional testers, NVH parameters and signal processing techniques have been added to help in the screening and to assess engine NVH performance. ACKNOWLEDGMENTS Figure 10a. Wavelet spectrogram of engine during Hot Start test The authors would like to thank the following Tecumseh Products engineers: Mr. Fernando Windlin, Test Lab Manager, TMT, for providing data from the engine testers at the TMT plant in Curitiba, Brazil; Mr. David Hammer, Corporate Manager of Manufacturing, for providing invaluable insight on manufacturing processes and requirements; Mr. Jonathan Johnston, Research Engineer, for providing excellent coding support and LabView expertise; Mr. Todd Carpenter, Senior

6 Research Engineer, for sharing his knowledge of engine dynamics and design. REFERENCES 1. G. debotton et al., On-line detection of cylinder-tocylinder variations by a vibration analysis system, SAE F. Corcione et al., Analysis and Control of Noise Emissions of a small single-cylinder D.I. diesel engine, SAE Y. Miura, N. Kojima, Noise generating mechanism at idling for a four-cylinder in-line diesel engine, D. Scholl at al., Spectrogram analysis of accelerometer-based spark knock detection waveforms, SAE L. Gavric, M. Darpas, Sound Power of Hermetic Compressors using acceleration measurements, Sixteenth International Compressor Conference at Purdue, July National Instruments LabView 7.1 Users Manual 7. P. Tunestal et al., In-cylinder measurements for engine cold-start control, IEEE International Conference on Control Applications, Hawaii, USA, CONTACT Alex Alvey, Tecumseh Products Research Laboratory, 3869 Research Park Drive, Ann Arbor MI Tel: (734) x. 343, aalvey@tpresearch.com Gabriella Cerrato Jay, MTS Systems Corp., Software and Consulting, 800 E. Whitcomb Ave., Madison Heights, MI Tel: (248) , gabriella.cerratojay@mts.com