A neural network approach to hierarchical analog fault diagnosis

A novel technique involving a neural network for efficient hierarchical fault diagnosis of analog circuits and systems is presented. The fault clustering property of the neural network is utilized to conceptualized the fault equivalence and BC (behavioral condition) reduction at higher levels. It is also shown that fault diagnosis can be done at any desired level and the precision of the diagnosis can be controlled during the fault dictionary generation stage. This technique can be applied to diagnose systems irrespective of their domain of operation as long as they can simulated. The proposed methodology was verified using an OTA-C low pass filter