基于声发射的机械密封状态识别

State recognition of Mechanical seal based on AE

摘要:

为了适应声发射信号频率覆盖范围广且频率高的特点,本文使用了在S变换的基础优化而来的广义S变换,并通过对广义S变换的时频分辨率特性进行了分析,选取广义S变换的参数，使得其在20khz~1000khz 的频率范围内依然拥有1khz~2khz的频率分辨率。随后在广义S变换分析的基础上对机械密封端面的声发射信号进行滤波和特征提取,并使用GA优化的TWSVM对密封的磨损状况进行了识别,准确率达到了100%,高于GA-SVM的识别率。说明相比传统的SVM在TWSVM在模式识别上更具有优越性。

Abstract:

AE signal is wide and high range in frequency. In this case we used generalized S transform to analysis AE signal in frequency domain and choose the parameter made the generalized S transform has a frequency resolution ranged 1khz~2khz in the frequency ranged 20khz~1000khz .and then we filter the AE signal and extract it’s characteristics based on the generalized S transform . and at last we diagnosis the condition of Mechanical seal by the method of GA-TWSVM and get a 100% accuracy rate which is high than the accuracy rate of GA-SVM. the result shows TWSVM is better than SVM in pattern recognition.