摘要
为了提高机械检测的效率,提出一种基于掩码自编码器的机械声音异常检测技术.在该技术中,可以利用掩码自动编码器来学习和提取数据信息中的特征,通过训练实现异常/正常的机械声音的检测.验证实验结果显示本次实验中提出的方法在风扇、水泵、滑轨和阀门声音检测中的AUC值分别为0.729 2、0.726 2、0.914 1和0.905 0.该方法的平均AUC值为0.819 1,相对于基线系统的AUC值提升11.8％.实验结果表明,本次实验中所提出的方法能够有效检测机械声音异常.
Abstract
In order to improve the efficiency of mechanical detection,a mechanical sound anomaly detection technique based on masked autoencoder is proposed.In this technique,the mask autoencoder can be used to learn and extract the features in the data information,and the detection of abnormal/normal mechanical sound can he achieved through training.The results of the validation experiments show that the AUC values of the proposed method in this experiment are 0.7292,0.7262,0.9141,and 0.9050 for fan,pump,slide,and valve sound detection,respectively.The average AUC value of the method is 0.8191,which is an improvement of 11.8％with respect to the baseline system's AUC value.The experimental results show that the method proposed in this experiment can effectively detect mechanical sound anomalies.
基金项目
哈尔滨职业技术学院校内课题(HZY2023JE001)
NETL
NSTL
万方数据