Research On Loose Detection Of Clamps For Hydraulic Pipeline

Hydraulic pipelines are usually operated in complicated internal and external excitation environments such as high temperature,high pressure,strong vibration,and fluid-solid coupling.Long-term operation of the equipment or poor installation quality can cause looseness of clamps installed on the pipeline.Loosening fault is one of the common failures in various mechanical equipment including hydraulic pipelines.At present,the research on the looseness of clamps is mainly focused on dynamic system modeling.The method is mainly based on the finite element model or other numerical methods.The looseness faults are studied by analyzing the stiffness,natural frequency and strain mode of the structure.The system modeling method based on dynamics is of clear physical meaning and theoretical significance,but it has a big gap with the actual engineering application.This paper takes the hydraulic pipeline clamps as the research object,methods based on signal processing are used to detect the looseness of clamps installed on hydraulic pipeline.The main research contents are as follows:(1)Aiming at the looseness problem of clamps for hydraulic straight pipeline,a loosening detection method based on entropy features is proposed.Firstly,the acquired three-dimensional vibration signals are decomposed by Multivariate Empirical Mode Decomposition(MEMD).Based on the matrix-based correlation analysis,appropriate Intrinsic Mode Functions are selected to reconstruct and fuse the original signals.Then the Composite Multi-scale Fuzzy Entropy(CMFE)is calculated for the fused signal,and feature selection is performed according to the Laplace Score to obtain the feature vector reflecting the state of the clamps.Finally,Support Vector Machine is used as a classifier to detect and locate loose clamps.The experimental results show that the method can effectively use the information of spatial scale and time scale,and can improve the detection accuracy of loose clamps.(2)Aiming at traditional manual feature extraction methods difficult to extract effective features from hydraulic elbow with complex working conditions,a clamp loosening detection model for hydraulic pipeline based on one-dimensional convolution neural network(1D-CNN)is proposed.The method directly inputs the data level fused signals into the convolutional neural network,and take advantages of deep learning to perform adaptive feature extraction,feature fusion and feature selection according to the specific network structure,and the loose detecting is performed at the decision layer.Compared with the traditional manual feature extraction algorithms,this model has better flexibility and higher recognition accuracy in loose detection of clamps.(3)For a pipeline with multiple clamps distributed,a method based on distributed Fibre Bragg Grating(FBG)sensors and one-dimensional convolutional neural network is proposed for detecting and locating of clamp looseness.This method makes full use of the advantages of the distributed FBG and sticks FBG strain sensors near each clamps along the pipeline.Each clamp only selects the two adjacent FBGs for its looseness detection.The experimental results show that the method owns wide coverage and strong robustness in the loose positioning of clamps and keeps high recognition accuracy under different experimental conditions.It is especially suitable for pipelines with multiple clamps distributed.