Neural recognition of diagnostic test data transforms

Abstract

By extending the concept of fault signatures on the primary outputs of the UUT to include the multiple parameters required of mixed signal testing, a fault dictionary approach to mixed signal UUT diagnostics can be developed. Transforms of fault signature ensemble information, as opposed to transforms of the time varying test signals themselves, can then be used as inputs to a neural net, the outputs of which are available to enhance conventional, fault dictionary processing of the original fault signature information.<>