A built-in rule scan structure for analog fuzzy processors test and fault diagnosis

In this paper a novel methodology for testing dedicated analog fuzzy processors is presented. It relies on small additional hardware, and features independence from the actual input-output relationship of the processor and low time for the test scan. In fact validation time does not increase exponentially with the number of inputs, as it would with an exhaustive sampling of the input-output relationship. Some simulation results show that all the most likely faults are detected.