Application of Autoassociative Neural Networks to Health Monitoring of the CAT 7 Diesel Engine

Report No. ARL-TN-0472
Authors: Andrew J. Bayba, David N. Siegel, and Kwok Tom
Date/Pages: February 2012; 22 pages
Abstract: An autoassociative neural network (AANN) algorithm was applied to fault detection and classification for seeded fault testing on a Caterpillar C7 diesel engine. Data used for this work is a subset from the seeded fault testing performed at the U.S. Army Tank and Automotive Research, Development and Engineering Center (TARDEC) test cell facilities. This report extends previous work performed on fault detection and classification performed by the U.S. Army Research Laboratory (ARL) on the C7 engine by including analysis using AANN [1]. We believed that AANN would be particularly useful in the diagnosis of faults in these tests because the correlation of several sensors appeared to be nonlinear. Although AANN performed quite well, the results were similar to the previous work using linear Principal Component Analysis (PCA) Statistics. We believe that the potential benefit in using AANN was not achieved due to the nature of the tests analyzed—in particular, data collection at discrete set-points in engine operation—and that within these set-point regimes, the sensor readings tend to be linearly correlated.
Distribution: Approved for public release
  Download Report ( 0.736 MBytes )
If you are visually impaired or need a physical copy of this report, please visit and contact DTIC.

Last Update / Reviewed: February 1, 2012