Automatic Overload Detection System Application in Induction Motors with Deep Convolution Neural Networks

In induction motors (IM) fed by a sinusoidal PWM (SPWM) inverter, the sudden load current increases at different inverter switching frequencies, and different stator coil pitches differ characteristically. Being able to predict these differences is very important for rotary electrical machine designers. In this study, for an automatic overload detection system, the detection of different load levels of three-phase cage IM fed by PWM inverter, depending on the operating conditions of the IM, using the deep convolutional neural network (DCNN) method was performed. The designed DCNN automatic overload prediction model has been trained with the data set obtained by experimental applications and its reliability has been increased by testing with the five-fold cross-validation method. In the study, it was concluded that in addition to the 96% accuracy achieved with test data, DCNN overload prediction models can be used very effectively in the design of rotary electrical machines.