Damage Identification Technology Of Submarine Pipeline Based On Deep Learning

A method of damage quantitative identification using cyclic neural network is established.Firstly,the finite element analysis of the pipeline is carried out by using ANSYS software,and the feasibility of the vibration modal parameters in the location and damage degree judgment is analyzed.Then,based on the principle of deep learning,the feasibility and high accuracy of damage identification using strain mode difference and cyclic neural network are obtained.Comparing the method with the traditional neural network,it shows the convenience of the method.The method solves the shortcomings of the traditional damage identification method,which needs to predict the number of injuries and train the neural network.It can widely identify various types of injuries,and has strong scalability.An integrated intelligent damage identification system is developed.The main program of intelligent algorithm based on pipeline parameters is generated by using random number generation program and finite element simulation software.The main program of intelligent algorithm is used to process the extracted modal shape data to form a regular array input,which can output visual pipeline damage identification results,including damage location and damage degree information.The model of API X52,API X65 steel pipe and the bimetallic pipe with 316 L austenitic stainless steel and 2205 duplex stainless steel as liner material under static and dynamic loading were constructed.Through the treatment and comparison of the data in the process of pipeline crushing under different load,a certain result was obtained.The results show that the overall performance of the pipeline is related to the specific mechanical properties of the material,and the overall strength of the bimetallic pipe is related to the material and thickness of base pipe and liner pipe.According to the above conclusions,the damage degree rating and corresponding pipeline management strategy recommendations are set for the output results of the intelligent damage identification system.