Oil pipeline operating condition identification based on deep convolution self-coding

To solve the problem of difficult identification of pipeline signals under different working conditions, a one-dimensional deep convolution self-coding method for pipeline leakage identification was proposed. One-dimensional deep convolution and adaptive encoder are used to conduct unsupervised learning training for signal features, extract data features, and learn data feature information through the convolution layer and pooling layer. Finally, the results of pipeline working condition discrimination are output. The experimental results show that the method can accurately judge the working conditions of different types of pipelines with an accuracy of 97.37%. The superiority of this method in leakage diagnosis is verified by comparison with other methods.