Intelligent detection and analysis of motors based on signal extraction and improved machine learning

Yalatu Su,
Intelligent detection and analysis of motors based on signal extraction and improved machine learning,
Array,
2025,
100562,
ISSN 2590-0056,
https://doi.org/10.1016/j.array.2025.100562.
(https://www.sciencedirect.com/science/article/pii/S2590005625001894)
Abstract: The current intelligent detection and analysis methods for motors have limited ability to identify motor faults under complex working conditions, as well as weak model generalization ability. To address these issues, a motor intelligent de tection and analysis model based on signal extraction and improved machine learning is proposed. The study uses vibration signals as detection data and implements noise filtering of vibration signals using variational mode decomposition and whale optimization algorithms. Then, short-time Fourier transform is utilized to extract the time-frequency characteristics of the vibration signal. An improved machine learning architecture is developed that combines Transformer network design concepts, attention mechanisms, and convolutional networks, and used as a testing model. The outcomes denote that the detection accuracy of the raised model under different fault types reaches over 94 %, and the false positive rate and false negative rate are only 0.06 % and 2.44 %, respectively. The detection time is only 1.23 seconds. Compared to other models, the raised model has higher detection accuracy and efficiency. In addition, during practical application, research has proposed that the model continuously learns and optimizes, resulting in the greatest improvement in its detection performance. At the end of the practical experiment, the daily detection accuracy reaches 101. The raised model can address the task of motor fault identification under complex working conditions and achieve intelligent detection of motors.
Keywords: Signal; Improvement; Machine learning; Motor; Intelligent detection; Convolutional neural network