Identification of Clamps Looseness based on Multi-Scale Convolutional Neural Network for Hydraulic Pipelines

With deep learning developing rapidly, intelligent clamps looseness identification methods based on CNN are becoming more popular. Considering the distribution characteristics of the FBG sensors on the hydraulic pipeline, a distributed data reconstruction method is proposed to obtain a suitable sample dataset. And the proposed MSCNN model can broaden the neural networks to reach a good identification performance owing to multi-scale convolution layer. Moreover, looseness identification experiments have been undertaken to indicate the feasibility and advantage of the method. Compared with 1DCNN and BP neural network, MSCNN can not only achieve higher accuracy in the testing set, but also take less time in the training process.