Clamps looseness detection of hydraulic pipelines based on convolutional neural network

摘要

Clamps are commonly used to fix hydraulic pipelines and are generally damaged by vibration fatigue, which further induce the failure and safety accidents in hydraulic system. Therefore, it is significant for the looseness detection of incipient clamps. Commonly feature extraction is the principle step of state detection, however, it is difficult to effectively extract structural looseness features through the conventional fault detection methods due to the complexity of the fluid motion and fluid-structure coupling. Convolutional neural networks (CNN), a model of deep learning, can fully extract the characteristics from complex signals themselves and commonly used in 2D signal recognition and classification. Thus, we present a looseness detection method for clamps in hydraulic system based on the CNN and distributed Fiber Bragg Grating (FBG) sensing. A simple hydraulic pipeline testing platform is designed and realized, and the experimental data and detection results show that this method based on CNN can detect the clamps looseness well and effectively. Furthermore, the effects of different parameters of CNN are also analyzed for the detection results.