Computer aided feature selection for enhanced analogue system fault location

Abstract A digital computer based technique for the selection of optimum test frequencies for fault diagnosis of analogue systems is presented. Gain and phase deviations from the nominal system response for changes in component values are evaluated and the recognition matrix is established which in turn is normalized and made sparse. The heuristic optimization technique developed is applied to the sparse recognition matrix to select the best subset of features via a discriminatory index of measurement points by adding or discarding features until the optimum is found. The efficiency of the selected feature set is measured by the confidence level formally incorporated in the algorithm and which is found to correlate well with the actual diagnosability of faults for a simulation of varying fault levels and including varying production tolerances for the non-faulty components. Actual fault location is implemented using the nearest neighbour rule, which is complimentary to the feature selection technique. Only three measurements are found to be sufficient to achieve a high level of diagnosability. The simulation of 91,000 faulty 7 component passive circuits has been undertaken to verify the procedure, which may now be applied equally well to larger scale systems using input-output measurements only, or, alternatively on sub-systems of comparable size to the sample circuit which have been isolated by special electronic partitioning devices as recently proposed for the fault location in television receivers during the manufacturing process.