A Neural Network Classifier for Fault Correlation and Root Cause Determination in an Electronic Warfare System

摘要

An offline Artificial Neural Network (ANN) classifier is developed for fault correlation and root cause determination for an aircraft's Electronic Warfare (EW) system. Feature extraction is performed from individual Health Indication Codes (HICs) and the ANN is trained from individual and artificially manufactured HICs via Supervised Learning with the Scaled Conjugate Gradient (SCG) algorithm. This concept is unique in that feature correlation from multiple HICs provides a higher probability of fault localization, and introduces solutions not highlighted when looking at individual HICs features. K-Fold Cross Validation is used to find the optimal number of neurons in the hidden layer of the neural network. A simulation is created for a System Under Test (SUT) with 605 HICs, each one having a feature vector of length 48 and 40 output classes corresponding to the root cause of the failure. The ANN is shown to train to a high classification accuracy based on the labeled data and perform to a high level against un-trained test data.