Evaluation of backtracing based diagnosis algorithms

Abstract

With the growing size and complexity of modern circuits, more algorithms are being developed nowadays for efficient fault diagnosis. Backtracing based diagnosis algorithms are effect-cause approaches that start from the failing outputs of the circuit and try to diagnose fault locations by backtracing lines toward the circuit inputs. In this thesis, general functionality was extracted between backtracing based diagnosis algorithms and implemented as an extension to an existing diagnosis framework. Furthermore, a simple graphical user interface was developed for the extended framework. The extended framework aims at facilitating the implementation and evaluation of different backtracing based diagnosis algorithms. In order to demonstrate its powerfulness, two modern backtracing based diagnosis algorithms were implemented on top of the extended framework. A number of diagnosis experiments on benchmark circuits was carried out in order to evaluate the two implemented algorithms. The experimental tools used and the results obtained are presented.