Infrared Thermal Image Fault Detection of Electric Motor Based on Improved VGG16 Network Model

• Published in: International Conference on Communications, Information System and Computer Engineering (Online) pp. 856 - 862
• Main Authors: Zhu, Xianghua, Hu, Shan, Bing, Zhigang, Zhao, Di
• Format: Conference Proceeding
• Language: English
• Published: IEEE 10.05.2024
• Subjects: Accuracy; AdamW; Fault detection; Focus loss; Induction motors; motor infrared thermal image; Performance evaluation; Stator windings; Thermal engineering; Training; Transfer learning; VGG16
• ISSN: 2833-2423
• DOI: 10.1109/CISCE62493.2024.10653406

As a core component of modern industry, electric motors are widely used in important fields such as electric power, machinery and national defense, and the fault detection of their thermal images faces problems such as diverse fault types, insufficient complex data processing capability and long training time. To address the above problems, this paper proposes a method for fault detection of infrared thermal images of electric motors based on an improved VGG16 network model. In the model construction and implementation, based on the VGG16 network model, the significant optimization of training time and accuracy is achieved by improving the structure of the latter layers of the model and introducing the migration learning strategy, as well as adding the callback function, the AdamW optimizer and the focal point loss function to the model. The experimental results show that the improved VGG16 model increases the average accuracy by 12.96%, reduces the average loss rate by about 23 times, and shortens the training time by half in infrared thermal image fault detection of induction motors. In the model performance evaluation, whether adjusting the test conditions or conventional model evaluation methods, the results reflect the superiority of the improved model in the realm of infrared thermal image anomaly identification of induction motors, which has practical application value.