Methods of Interpreting Diagnostic Data for Locating Faults in Digital Machines

Several techniques for translating the results of diagnostic tests into specific fault identities are described. This translation can be difficult in large and complex machines. The amount of test data required to isolate faults, and the obscure symptoms some faults generate, preclude efficient manual test-by-test interpretations. The additional observed fact that a significant number of faults yield inconsistent test results from diagnosis to diagnosis demands a flexible interpretation of data. Techniques are described for producing fault dictionaries which can be used by the maintenance craftsman to identify machine faults in a relatively short time. These techniques utilize multidimensional geometric representations of diagnostic results, methods for identifying and ignoring inconsistent tests, pseudo-random mappings, and other procedures for condensing and organizing the information contained in diagnostic test data. The results of applying these techniques to data obtained from the Bell System's No. 1 Electronic Switching System are also discussed.