Simplified ATPG and analog fault location via a clustering and separability technique

Automatic test program generation (ATPG) for analog system fault location is considered and a procedure for generating a set of tests for a linear network using gain and phase measurements from input-output measurements only is presented. The best subset of features for fault diagnosis is selected via a discriminatory index. Each feature subset selected during the optimization procedure is tested for enhanced separability of observation space for faults and the efficiency of the subset for diagnosis is indexed using a confidence level. The fault matrix formed using the selected feature subset is analized for cluster formation, the separability measure introduced is used to group the fault cases and a reduction of the cases to be considered on-line is obtained. In fault location, a fault is identified to a specific group which then is further located to the individual component and is carried out for varying production tolerances.