Feature Selection Methods for Fault Diagnosis of Axial-Flow Fans
A fault diagnosis model for axial flow fans based on multi-signal feature selection has been developed to ensure the safety of coal mine workers and enhance production efficiency.This model primarily monitors the vibration acceleration signals of fans under various operating conditions,process-ing these signals to obtain velocity and displacement data.The diagnostic accuracy is significantly en-hanced by extracting these signals'time-frequency domain statistical features and using Extreme Gradi-ent Boosting(XGBoost)for feature selection.Experimental results indicate that the optimized multi-signal dataset,after feature selection,achieves an average recognition accuracy of 98.33％on the test set,with a log loss of only 0.0534.Compared to datasets using single signals or without feature selec-tion,the model demonstrates higher efficiency and accuracy,significantly improving the reliability and speed of fault diagnosis,thereby effectively reducing the safety hazards potentially caused by fan fail-ures.
万方数据
维普
