Evolutionary algorithms for global parametric fault diagnosis in analogue integrated circuits

An evolutionary method for analogue integrated circuits diagnosis is presented in this paper. The method allows for global parametric faults localization at the prototype stage of life of an analogue integrated circuit. The presented method is based on the circuit under test response base and the advanced features classification. A classifier is built with the use of evolutionary algorithms, such as differential evolution and gene expression programming. As the proposed diagnosis method might be applied at the production phase there is a method for shortening the diagnosis time suggested. An evolutionary approach has been verified with the use of several exemplary circuits - an oscillator, a band-pass filter and two operational amplifiers. A comparison of the presented algorithm and two classical methods - the linear classifier and the nearest neighborhood method - proves that the heuristic approach allows for acquiring significantly better results.