Electrical Fault Diagnosis Using Weighted Support Vector Machine

Nowadays, developing element bearings are the most lying components to faults in electrical machines. From both operational and economic perspectives, efficient monitoring and diagnosis of bearing electrical are critical to prevent unexpected downtime. Executing an effective bearing electrical fault diagnosis system enables early fault detection, allowing for timely maintenance and minimizing the risk of catastrophic failures. This paper introduces an effective Weighted Support Vector Machine (WSVM) model for accurate electrical fault diagnosis, to achieve outstanding performance. This process begins with the collection of electrical data from Electrical Motor Fault (EMF) dataset, as it contains thermographic images of electrical faults. Next, Short-Time Fourier Transform (STFT) is applied for processing and to produce modified signals. Then, features are extracted from modified signals with the help of Time-Domain Signal Analysis (TDSA) and sent for feature selection. Feature selection is accomplished through the Binary Particle Swarm Optimization (BPSO) algorithm for selecting optimal features. Finally, the proposed WSVM model effectively differentiates between faulty and healthy electrical machines, enabling early fault detection and diagnosis. From the results, the proposed WSVM outperformed when compared with existing model Fault Diagnosis in Electrical Machines using Particle Swarm Optimization (FD-EM-PSO) by achieving 98.9% accuracy.