Fault Detection and Localisation of a Three-Phase Inverter with Permanent Magnet Synchronous Motor Load Using a Convolutional Neural Network

Fault-tolerant control of a three-phase inverter can be achieved by performing hardware reconfiguration the six-switch and (6S3P) topology to four-switch (4S3P) after detection localisation faulty phase. Together with reconfiguration, SVPWM algorithm must appropriately modified handle new 4S3P topology. The presented study focuses on diagnosing faults in two steps: fault localisation. Fault is needed recognise healthy or unhealthy state inverter. binary recognition problem solved preparing feature vector that calculated from phase currents (ia, ib, ic) time frequency domains. After diagnosis system recognises state, it investigates signals localise which faulty. multiclass classification was transformation into single RGB image training convolutional neural network. proposed methodology for inverters tested based simulation model representing laboratory test bench. learning process, possible 128-sample window (corresponding 0.64 ms) an accuracy 99 percent. In next step, selected individual performed basis 256-sample 1.28 100