Neural networks applied to automatic fault detection

The objective of this paper is an automatic fault detection scheme for applications in the automotive industry. In order to increase data reliability and for the purpose of monitoring and control of test equipment, a fault detection system based on multivariate data analysis is being developed. The detection scheme has to process up to several hundreds of different measurements at a time and check them for consistency. The main problem lies in the fact that besides the available data, no further information is provided. The chosen approach models the distribution function of available fault-free data using ellipsoidal basis function networks. An important requirement for the fault detection scheme is that it should be able to automatically adapt itself to new data. The present paper is focused on this feature. It is demonstrated how a gradient optimization with algebraic constraints can be applied to adapt a pre-existing network to new data points. Numerical examples with actual data show that the proposed method produces good results.