The Importance of Feature Processing in Deep-Learning-Based Condition Monitoring of Motors

Table 1

Application of MLP in condition monitoring of motors.

Year	Input data and features	DL model	Application	Remarks
2016	Current signature	MLP [47]	Induction motor fault classification	Operating frequency of the motor did not affect the accuracy of the classifier. However, increase in the accuracy was observed as severity of the fault increased
2017	Current signature and voltage signal amplitudes	MLP [48]	Induction motor fault classification	The method yielded better classification results and it was computationally efficient compared to the traditional methods
	Current signature + MI	MLP [49]	Induction motor fault classification	MI feature extraction method allowed MLP to produce promising results compared to RBF
2018	Current signature + FFT + WPT + statistical parameters	MLP [50]	Rotor broken bar fault classification	Superior feature extraction allowed to detect the fault severity level even with no-load condition
	Current signature + CCA	MLP [51]	Induction motor fault classification	The method allowed to correctly classify the faults with higher accuracy owing to reduced dimensions of the data by CCA
2019	Current signature + MI	MLP [52]	Bearing fault classification	MI feature extraction method allowed to classify the bearing faults more effectively compared to conventional methods