Research On Fault Diagnosis Of Aviation Hydraulic Pipe Clamp Based On Improved 1D-CNN And Bidirectional GRU

The aero-engine piping system is like the "cardiovascular" of the engine,which mainly transmits power to the actuator of the engine.The aviation piping is connected to each other through the clamp.The clamp can enhance the stiffness and fixed position of the aviation piping.Due to the coupling effect of the foundation excitation of the casing at different positions and the pulsating excitation of the pipeline pump source body,it is easy to induce various faults such as loosening of the aviation clamp under the excitation of multiple source loads.Therefore,accurate identification of the aviation clamp fault is of great significance for the condition monitoring and reliable operation of the engine.The vibration fault of aviation pipe clamp is studied.Because the signal of clamp fault is nonlinear and unstable,and it is interfered by strong noise,it is difficult to extract the fault characteristics of clamp fault by vibration processing method to achieve accurate identification.Therefore,an improved One-Dimensional Convolutional Neural Network and a Bi-directional Gated Recurrent Unit,are designed method for fault diagnosis of aviation pipeline clamp.Firstly,since the aviation pipe clamp fault signal is one-dimensional time series data,the classic two-dimensional CNN model will have the problem of feature loss when it deals with the aviation clamp fault signal.Therefore,the CNN model is improved,a 1D-CNN spatial feature extraction model is designed,and the measured clamp fault data set is directly input into the 1D-CNN model for training.The fine-grained features of the clamp fault are extracted adaptively,and the local fault features are extracted from the aviation clamp fault data.Secondly,because the clamp fault data has the characteristics of time point,while the current fault diagnosis model seldom considers the time flow or the comprehensive influence of the before and after moments,the GRU module is introduced into the designed 1D-CNN spatial feature extraction model for optimization.In order to comprehensively extract the coarse-grained features of the aviation clamp fault data,A Bi-GRU time feature extraction model was designed,and coarse-grained features were further extracted from the fine-grained features extracted from the 1D-CNN model,so as to obtain the global features of the fault data,so that feature fusion could be realized in space and time of the aviation clamp fault data.The experimental data show that: the designed 1D-CNN-Bi-GRU air-time model of aviation clamp can realize accurate identification of four kinds of faults,such as clamp loosening,crack,wear and pit,which provides a new way for the early intelligent fault diagnosis of aviation hydraulic pipe clamp.Finally,based on the same data set,The proposed 1D-CNN-Bi-GRU aviation pipeline clamp space-time model fault diagnosis method is combined with the current advanced deep convolutional neural network(DCNN),gated cycle unit(GRU),Recurrent neural network,RNN),Support Vector Machine(SVM),Back Propagation Neural Network(BPNN)and other 5 fault diagnosis methods.The performance of the designed diagnosis method of 1D-CNN-BiGRU air time model of aviation pipe clamp is obviously better than that of DCNN and other five intelligent identification methods.In order to carry out the research of vibration failure mechanism of aviation hydraulic pipeline clamp,it has certain application value and practical significance.