ABSTRACT
This paper presents the application of an adaptive neural network architecture in the implementation of a fuzzy diagnostics system. The fuzzy diagnostic system attempts to diagnose automotive failures for the End-of-Line testing procedures. The neural network architecture consists of three neural network subsystems, an Action Selection Network (ASN) which is the implementation of the fuzzy diagnosis, an Action Evaluation Network (AEN) which plays the role of critic in evaluating the results of ASN, and a Stochastic Action Modifier which combines the results of ASN and AEN to finalize the output. The fuzzy diagnostic system implemented on the neural network architecture has been tested on a large set of data directly downloaded from automobile assembly plants.
