A Novel Test Generation Methodology for Adaptive Diagnosis

Rajsekhar Adapa,  Edward Flanigan,  Spyros Tragoudas
Southern Illinois University


Abstract

This paper presents a automatic test pattern generation technique to improve the diagnostic resolution of a given test set. Each test pattern generated by existing techniques detects a large number of faults. Identifying the faulty candidate from a large set of possible fault candidates is extremely difficult and time consuming. A novel framework to adaptively generate additional patterns for diagnosing the faulty location is presented. The additional patterns prune a set of fault free candidates from the possible fault candidates. The proposed technique improves the diagnostic resolution where each new pattern detects only a small number of faults and each fault is detected by only few patterns. The proposed method is applicable to any fault model and distinguishes a large number of faults with a small number of patterns. For simplicity we demonstrate the effectiveness of the approach on the path delay fault model.