Semi-automatic development of test program sets (TPS)

Abstract

This paper describes the research and development of methods and processes to automate the generation of adaptable test programs, from functional specifications, that will reduce the cost and time required to establish an organic depot test and repair capability. With increasing emphasis on performance based acquisition (PBA) throughout the DoD and USAF, this technology is essential. The prototype being described in this paper starts with a functional model of the circuit. Possible faults are automatically generated, and then the modified circuits are simulated to create training data for a set of neural networks. The new TPS consists of a series of trained neural networks that automatically identify failed component groups or determine no-fault in the actual units under test (UUT). The technology being developed in this effort compliments ongoing PBA activity to improve reliability and reduce the cost of sustaining avionics systems by reducing test results of no fault found. This paper shows the integration of neural network technologies with software and hardware modeling to develop innovative, adaptive and effective diagnostic test programs.