RETRACTED: Fault Detection in Multi-input Analog Circuits using Regression Modelling

Abstract

The aim of this paper is to detect parametric faults in multi-input analog circuits by solving for the coefficients of a polynomial regression model using traditional linear least squares techniques, with some care applied in the solution. The multi-input circuit output is expressed in terms of more than one input variables using polynomial coefficients. In the proposed approach, the value of each component of the circuit under test (CUT) is varied within its tolerance limit using Monte-Carlo simulation to calculate its fault free polynomial coefficient bounds. The CUT is then declared fault free or faulty based on the result of the comparison of its estimated polynomial coefficients with the fault free coefficients. To the best of our knowledge, parametric fault detection in multi-input analog circuits using polynomial regression modelling is attempted for the first time in the literature. The effectiveness of the proposed method is demonstrated via two case studies, namely, a lead lag circuit and the PI compensator of a peak-current-mode-controlled buck-type switching converter.