9521609 Chow In the previous project, the Principal Investigator has demonstrated the potential of artificial neural networks to learn the complicated motor fault detection mappings for a single-phase induction motor under constant load conditions. However, there are many uncertainty factors in the motor operations that can significantly affect the motor incipient fault detection process and which have not been addressed. In this proposed project, the Principal Investigators propose to extend the research to cover more realistic operating environments. The PIs will investigate different factors such as varying load conditions, saturation effects, temperature effects, noise effects, and their influences on the process of motor incipient fault detection for three-phase induction motors. In addition, the PIs will investigate and establish a general theory and principle for motor incipient fault detection using a set theoretic formulation. After setting up the problem in set theoretic formulation, the challenges will lie in the finding of the correct mappings from the appropriate measurements to the estimation of the actual motor faults and their severity. The PIs will investigate and demonstrate the advantages and feasibility of the use of neural network and fuzzy logic technologies to obtain the motor fault mapping from appropriate information to yield accurate motor incipient fault detection in an non-invasive, economical, and reliably manner, along with the ability to provide a qualitative and heuristic explanation of the fault detection process. ***
