Impact of the Convolutional Neural Network Structure and Training Parameters on the Effectiveness of the Diagnostic Systems of Modern AC Motor Drives

Abstract

Currently, AC motors are a key element of industrial and commercial drive systems. During normal operation, the machines may become damaged, which may pose a threat to the users. Therefore, it is important to develop a fault detection method that allows for the detection of a fault at an early stage. Among the currently used diagnostic systems, applications based on deep neural structures are dynamically developed. Despite many examples of applications of deep learning methods, there are no formal rules for selecting the network structure and parameters of the training process. Such methods would make it possible to shorten the implementation process of deep networks in diagnostic systems of AC machines. The article presents a detailed analysis of the influence of deep convolutional network hyperparameters and training procedures on the precision of the interturn short-circuits detection system. The studies take into account the direct analysis of phase currents through the convolutional network for induction motors and permanent magnet synchronous motors. The research results presented in the article are an extension of the authors’ previous research.

Keywords

• convolutional neural network
• hyperparameters
• fault detection
• induction motor drive
• permanent magnet synchronous motor
• diagnostic system