A real-time neural fuzzy (NF) power control system is developedand compared with a backpropagation neural network (BNN) system. Theobjective is to develop computation hardware and software in order toimplement the fault classification of a three-phase motor in real-timeresponse. With online training capability, the NF system can be adaptiveto the particular characteristics of a particular motor and can beeasily modified for the customer's needs in the future. Thepreprocessing of a BNN-based fault classifier normalizes the magnitudebetween [-1,1] and transforms the number of samples to 32 for a cycle ofwaveform. The trained BNN is used to classify faults from the inputwaveforms. Real-time response is achieved through the use of a parallelprocessing system and the partition of the computation into parallelprocessing tasks. Compared with a four-processor BNN system, the NFsystem requires smaller cost (three processors) and recognizes waveformsfaster. Moreover, with the appropriate feature extraction, the NF systemcan recognize temporally variant spike and chop occurring within a sinwaveform
展开▼