Using neural networks to improve built-in test

摘要

This paper describes an investigation and demonstration of the use of neural network technology to improve Built-in Test (BIT) by filtering out false alarms from the BIT output and by identifying intermittent faults. The Neural Network False Alarm Filter (NNFAF) program was a research contract for Rome Laboratory at Griffiss Air Force Base, New York, The contract work was completed in September, 1994. The purpose of the project was to identify and develop a set of approaches for applying neural network learning techniques to improve the performance of BIT. The approaches focussed on the need to filter out false alarms and to identify intermittent failures. The NNFAF methodology involved a state-of-the-art assessment of neural networks and BIT techniques, selecting candidate BIT technique and neural network conibinations for analysis, simulating the selected BIT techniques, training and testing the selected neural networks, and analyzing the results.