Identification of pipe leak apertures based on dynamic depth-separable convolutional neural networks
Aiming at the complexity of the model structure,the number of parameters and the large amount of computation caused by the traditional model to improve the accuracy of pipeline leakage detection,a pipeline leakage aperture recognition method based on the dynamic depth-separable convolutional neural network was proposed.The dynamic convolutional layer takes the extracted leakage signal features through the calculation of channel-attention weights and the dynamic weights fu-sion,and then obtains a stronger feature through the dynamic depth-separable convolutional layer.The dynamic convolutional layer reduces the parameters of the network model by using the global average pooling layer,and the pipeline leakage aper-ture was identified by the fully connected layer.The results show that the new method has a high recognition accuracy,over-comes the problems of a large resource overhead and a high power consumption of the traditional model,reduces the training time of the model,improves the recognition speed of pipeline leakage aperture,and can be used to monitor the degree of pipeline leakage in industries.
identification of leakage aperturedynamic depth-separable convolutionlightweight networksdynamic convolu-tion
万方数据
维普
