论文部分内容阅读
由于天然气管道阀门内漏声发射检测环境复杂、噪声干扰严重,极大地降低了阀门内漏流量的检测精度。为此,提出一种基于背景噪声的小波包软阈值降噪处理方法,并通过对降噪处理后的声发射信号采用基于支持向量回归(Support Vector Regression,SVR)方法进行输气管道阀门内漏流量的量化回归预测。结果表明:采用基于背景噪声的小波包软阈值降噪方法能够获取较为纯净的内漏源信号,降噪后获得内漏声发射信号信噪比为6.11。通过对小波包降噪处理后的特征参数进行输气管道阀门内漏流量回归预测,结果优于未进行降噪处理的预测结果,且软阈值降噪预测结果优于硬阈值预测结果,采用软阈值降噪的预测结果平均绝对比例误差为0.164,有效提高了阀门内漏流量量化回归预测的准确度。
Due to the complex environment of acoustic emission detection in natural gas pipeline valves and serious noise interference, the detection accuracy of the leakage flow in the valve is greatly reduced. Therefore, this paper proposes a wavelet packet soft threshold denoising method based on background noise, and uses the Support Vector Regression (SVR) method to detect the internal leakage of gas pipeline valves Quantitative regression prediction of traffic. The results show that the pure noise source signal can be obtained by using the wavelet packet soft threshold denoising method based on background noise, and the signal to noise ratio of the inner leakage acoustic emission signal after noise reduction is 6.11. The results of regression regression of the leakage flow within the gas pipeline valve by wavelet packet noise reduction are better than those without the noise reduction and the prediction results of the soft threshold noise reduction are better than the hard threshold prediction. The average absolute proportional error of the prediction of threshold noise reduction is 0.164, which effectively improves the accuracy of quantitative regression prediction of leakage flow within the valve.