FPGA-Based Online Detection of Multiple-Combined Faults through Information Entropy and Neural Networks

For industry, a faulty induction motor signifies production reduction and cost increase, besides, it is a hazard for people and nearby machinery. Real-world induction motors can have one or more faults at the same time, so that one faulty condition could interfere with the detection of another one, and mislead to a wrong decision about the operational condition of the motor. The detection of multiple-combined faults is still a demanding task, difficult to accomplish even with computing intensive techniques. This work introduces a low-cost, real-time FPGA-based hardware processing unit for multiple-combined fault detection utilizing information entropy and artificial neural networks as tools for analyzing the information contents of the 3-axis vibration signals from the rotary machine during the startup transient. Results show great performance of the entropy neural system on accurately identifying in an automatic way a healthy condition, half-, one-, and two-broken rotor bars, outer-race bearing defect, unbalance, and their combinations.