Soft Computing and Computer Aided Reliability Engineering – A Synergism for Online Monitoring of Electronic Systems

Computer aided reliability design techniques has proven to be an efficient tool for detailed analysis of the performance of electronic systems. The past few years have witnessed a growing recognition of Soft Computing (SC) technologies that underlie the conception, design and utilization of intelligent systems. This paper evaluates the use of SC methods for on-line monitoring and failure prediction of electronic systems. A novel technique that uses the stressor – susceptibility interaction is used for modeling the failure. According to this a circuit or a system is deemed to be failed once the stressor has exceeded the susceptibility limits. The soft computing methods considered are i) an Adaptive Network Based Fuzzy Inference System (ANFIS), ii) an Evolving Fuzzy Neural Network (EFuNN) iii) a Radial Basis Function (RBF) network and iv) an Artificial Neural Network (ANN) that uses the Levenberg-Marquardt (LM) algorithm. The SC models use preprocessed validated stressor vectors as their inputs. The SC models are evaluated based on their prediction of the leakage current and junction temperature in a electronic power supply circuit on which the analysis was carried out. Results indicating the strengths and weakness of each model are presented.