Alternator Health Monitoring For Vehicle Applications David Siegel Masters Student University of Cincinnati
Report Documentation Page Form Approved OMB No. 0704-0188 Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. 1. REPORT DATE 30 APR 2009 2. REPORT TYPE Briefing Charts 3. DATES COVERED 05-05-2009 to 07-05-2009 4. TITLE AND SUBTITLE Alternator Health Monitoring For Vehicle Applications 5a. CONTRACT NUMBER W911NF-07-D-001 (#07023) 5b. GRANT NUMBER 5c. PROGRAM ELEMENT NUMBER 6. AUTHOR(S) David Siegel 5d. PROJECT NUMBER 5e. TASK NUMBER 5f. WORK UNIT NUMBER 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) U.S. Army TARDEC,6501 East Eleven Mile Rd,Warren,Mi,48397-5000 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) U.S. Army TARDEC, 6501 East Eleven Mile Rd, Warren, Mi, 48397-5000 8. PERFORMING ORGANIZATION REPORT NUMBER #19804 10. SPONSOR/MONITOR S ACRONYM(S) TARDEC 11. SPONSOR/MONITOR S REPORT NUMBER(S) #19804 12. DISTRIBUTION/AVAILABILITY STATEMENT Approved for public release; distribution unlimited 13. SUPPLEMENTARY NOTES 14. ABSTRACT To develop and evaluate methods and algorithms for monitoring and predicting the health of military vehicle components or systems. 15. SUBJECT TERMS 16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF ABSTRACT a. REPORT unclassified b. ABSTRACT unclassified c. THIS PAGE unclassified Public Release 18. NUMBER OF PAGES 16 19a. NAME OF RESPONSIBLE PERSON Standard Form 298 (Rev. 8-98) Prescribed by ANSI Std Z39-18
Vehicle Component Health Monitoring Overview Program Area: Prognostics and Health Monitoring Objectives: To develop and evaluate methods and algorithms for monitoring and predicting the health of military vehicle components or systems. Deliverables: (a) Develop a methodology to monitor and predict the health of a vehicle component or subsystem. Task # TASKS Task 1: Critical Component Selection Task 2: Study Failure Modes Task 3: Construct Test-bed to Validate Method Task 4: Evaluate Health Monitoring Algorithms Task 5: Technical report (b) Evaluate algorithms for health monitoring and prognostics for vehicle components. (c) Technical report which document works and results. (09/20/2007 12/31/2008) Task# 1 2 3 4 5 6 7 8 9 10 11 12 1 2 Task1 Task2 Task3 Task4 Task5 Industry Mentors: Joe Gothamy, Ken Fischer, James Bechtel of TARDEC University: Professor Jay Lee, Professor Teik Lim Budget: 1 GRA, PI support 2
Outline Project background Alternator failure modes Experimental test-bed Signal processing and feature extraction Health monitoring algorithms and results Summary and conclusions 3
Project Background The objectives of the project was to develop and evaluate prognostic and health monitoring algorithms for military vehicle components. In order to move forward with this objective, it was necessary for this 1-year study to focus on monitoring the health of a critical component on the vehicle. A list of potential components were formed where there would be a benefit for health monitoring and prognostics. This list was further narrowed down based on the feasibility, and the vehicle alternator component from the HMMWV vehicle was chosen as the case study for this project. Alternator was chosen The health monitoring methods developed will be validated using data collected from a test-bed from a degraded used alternator and a healthy new alternator. 4
Alternator Failure Modes Failure Mode Signals To Measure Expected Degradation Pattern Alternator Bearings Measure vibration signal near drive end and anti-drive end bearing Greater magnitude at bearing fault frequencies and RMS for degraded component Alternator Diodes 28V and 14V output voltage signals and current signals Higher level of ripple and differences in 36 th and 72 nd voltage harmonics Alternator Stator Windings AC current signals and vibration signals Differences in vibration and current higher order harmonics such as 30 th and 36 th. 5
Alternator Test-bed DAQ Hardware: NI Compact DAQ (cdaq-9172) 1. NI 9215 (for AC current measurement) 2. NI 9221 (Voltage Measurements) 3. NI 9233 (Vibration Measurements) Measured Signals: 1. AC Current on 28V outlet 2. Alternator Tachometer Signal 3. 28V Outlet Voltage Signal 4. 14V Outlet Voltage Signal 5. DC Current on 14V outlet 6. Vibration near DE Bearing 7. Vibration near ADE Bearing 6
Feature Extraction for Monitoring Alternator Diodes 28V Signal (V) 28V Ma agnitude (V) 30 29 28 27 28V Signal: Used and New Alternator 26 0.46 0.462 0.464 0.466 0.468 0.47 0.472 0.474 0.476 0.478 0.48 Time (s) 28V Order Spectrum Zommed in View of 12th Order 0.2 0.02 0.15 0.1 0.05 0 0 50 100 Order 28V Ma agnitude (V) 0.015 Used Alternator New Alternator 0.01 11.96 11.98 12 12.02 12.04 Order Voltage Ripple Standard Deviation (V) 0.17 0.16 0.15 0.14 0.13 0.12 28V Ripple Standard Deviation Feature Used Alternator New Alternator Alternator Running at 4900 rpm 150A Load on 28V Outlet 0.11 0 5 10 15 20 25 30 35 40 Sample The 28 voltage time signal and order spectrum are shown, each signal is sampled at a rate of 50KHz due to the 36 th and higher harmonics in the signal. Larger amount of variation in the 28V ripple signal for the degraded used alternator compared to the healthy new alternator. 7
Feature Extraction for Monitoring Alternator Bearings Drive End Vibration (g) Drive End Vibration (g) 2.5 2 1.5 1 0.5 Drive End Vibration Order Spectrum: Used and New Alternator 0 0 10 20 30 40 50 60 70 80 90 100 Order Zommed in View of 3rd order (BPFO) Zommed in View of 36th Order 0.06 0.04 0.02 0 2.95 3 3.05 Order Drive End Vibration (g) 0.4 0.3 0.2 0.1 Used Alternator New Alternator 0 35.9 35.95 36 36.05 36.1 Order BP PFO Vibration Magnitude (g 2 ) 6 x 10-3 5 4 3 2 1 Drive End Bearing BPFO Vibration Magnitude Feature Used Alternator New Alternator Alternator Running at 4900 rpm 150A Load on 28V Outlet 0 0 5 10 15 20 25 30 35 40 Sample The degraded used alternator has clear differences in the magnitude in the bearing fault frequencies, such as the amplitude at the BPFO frequency which is associated with a bearing with an outer race damage. 8
Feature Extraction for Monitoring Alternator Stator Winding 2 AC Current Order Spectrum: Used and New Alternator 5.5 42 Order ADE Vibration Feature (Stator Health) AC Current (A) ) AC Current (A) 1.5 1 0.5 0 0 10 20 30 40 50 60 70 80 90 100 Order Zommed in View of 12th Order Zommed in View of 18th Order 0.2 0.15 0.1 11.98 12 12.02 Order ) AC Current (A) 0.2 0.1 0 17.9 17.95 18 18.05 18.1 Order r Anti-Drive End Vibration Magnitude (g 2 ) 42 Order 5 4.5 4 3.5 3 2.5 2 1.5 1 Used Alternator New Alternator Alternator Running at 4900 rpm 150A Load on 28V Outlet 0.5 0 5 10 15 20 25 30 35 40 Sample The degraded used alternator has clear differences in the magnitudes in the high order current and vibration harmonics, such as the 42 nd order vibration harmonic. 9
Multi-Regime Aspect The used and new alternator were tested under different electrical loads and speeds, 24 different load and speed combinations. A self-organizing map (SOM) was used to see whether the data formed clusters based on electrical load or speed. No Load 50A Load 100A Load U-matrix 4.22 2.36 Labels 0 0 0 0 0 0 50 0 50 50 0 50 50 50 50 50 50 50 100 100 100 100 100 100 100 150 150 As shown for the data when rotational speed was held constant at 4800 rpm, the vibration and electrical data collected at different loads formed clear clusters. Main Points: 150A Load 0.497 100 150 150 150 150 150 150 150 150 Similar results were also seen for rotational speed, when electrical load was held constant. 10 1. Necessary to segment and train for each operating regime. 2. Analyze health monitoring results from data collected in each regime.
Logistic Regression Health Assessment Results Take Extracted and Selected Features Results Sample Features Label Fit Regression Parameters Use Trained Model to Calculate Health Value Pr ob( event) 1 = P( x) = g ( x) 1+ e = e g ( x) 1+ e g ( x) Overall results using logistic regression were quite good, 5% missed detection. 11
Statistical Pattern Recognition Health Assessment Results Concept of Statistical Pattern Recognition Results Key assumption that features follow normal distribution Distribution of 28V Ripple Standard Deviation Feature 140 120 Used Alternator New Alternator 100 80 60 Alternator Running at 4900 rpm 150A Load on 28V Outlet Method can distinguish the healthy new alternator from used alternator but is too sensitive. 40 20 0 0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 28V Ripple Standard Deviation (V) Many features do not follow normal distribution so this algorithm was not considered any further. 12
Self-Organizing Map-MQE MQE Method Health Assessment Results Concept of SOM-MQE Method Train self-organizing map with data from healthy component Results New data comes in and best matching unit on map is found Distance between new data and best matching unit is defined as minimum quantization error (MQE) value MQE= D w bmu MQE value indicates how far from normal the component is, indicates amount of degradation Accuracy of health results very close to logistic regression method. 13
Summary and Conclusions Understanding the failure modes and expected failure patterns are the key to the proper signal processing and feature extraction and selection (the features are the inputs to the health monitoring and prognostic algorithms). One has to consider the effect of different operating regimes and settings, it is typically necessary to segment the data and train the algorithm for each operating regime. In a few operating regimes it might be more difficult to monitor the health, for example at idle speed, monitoring the alternator bearing health was more difficult and it was harder to classify the degraded alternator from the new alternator bearings at idle speed. Statistical pattern recognition is based on the assumption of a normal distribution for the features, and if this is not met this algorithm should not be used. Both the logistic regression method and self-organizing map method provided similar level of accuracy, however logistic regression is less computationally demanding, so the simpler algorithm is the best option (logistic regression). 14
Benefits and Impact to Members Algorithm Development and Evaluation Evaluation of Health Monitoring Algorithms Algorithms tested in application with many regimes and variety of signals Performance accuracy results indicate which algorithm is best. Impact: Know which algorithm is most suitable. PHM Method for Key Components Developed Method Applicable to Motors and Generators Machine tools, power plants, vehicles, most any application has a type of motor or generator Impact Adaptable methodology for monitoring the health for vehicle alternator is developed and can be used in many applications 15
Thank You Questions? 16