Méthodes de diagnostic et de maximalisation de la couverture des fautes singulières fréquentes dans les circuits analogiques

Abstract

This thesis focuses on the application of new techniques to diagnosis, detection and classification of faults in analog integrated circuits, which can break down and degrade during the operating period. These circuits have become more necessary in the world of electronics. This need is due to its importance in the areas of telecommunication, avionics, biomedical... This requires a diagnosis whose the main purpose is to provide an automatic tool that allows early detection of faults to protect the circuits, and to classify this faults two. Test techniques have been applied to these analogue circuits, namely the functional test, the improvement of the detection, the simulation of faults. For the development of the algorithms of classification of faults in the analog circuits, at the beginning, a database is collected using a simulation mounted Carlo in frequency and transient domain, for the fault free and faulty circuit. Artificial intelligence-based classifiers have been constructed, such as neural and neuro-blur classifiers. Finally, the diagnosis was improved by the introduction of a new classifier (Multi-class neuro-blur classifier). This classifier was compared to other types of classifiers and over classed to them by its efficiency and accuracy