基于多尺度一维卷积自编码器的输气管道泄漏判别

Gas Pipeline Leakage Identification Based on Multi-scale One-dimensional Convolution Adaptive Encoder

针对声发射输气管道不同泄漏信号难以判别的问题,提出了一种基于多尺度一维卷积自编码器的输气管道泄漏判别方法。利用多尺度一维卷积自编码器对信号特征进行无监督学习训练,提取数据特征信息,通过多尺度的卷积层和池化层对数据特征信息进行学习,最后,输出管道泄漏判别结果。实验结果表明:该自编码器能够精确判断不同类型的管道泄漏,准确率达到97.13%。通过与其他自编码器进行对比验证了该自编码器在泄漏判别方面的准确性。

Aiming at the problem that it is difficult to distinguish different leakage signals of acoustic emission gas transmission pipeline,a multi-scale one-dimensional convolution adaptive encoder for gas transmission pipeline leakage identification was proposed.Multi-scale one-dimensional convolution adaptive encoder were used to conduct unsupervised learning training on signal features,extract data feature information,and learn data feature information through multi-scale convolution layer and pooling layer.Finally,the results of pipeline leakage discrimination were output.Experimental results show that the adaptive encoder can accurately judge different types of pipeline leakage with an accuracy of 97.13%.The accuracy of this adaptive encoder in leakage distinguish is verified by comparing with other adaptive encoder.