Research On Data Processing And Fault Diagnosis Of Ship Structural Stress Monitoring

In order to adapt to the continuous growth of the world’s shipping,ships are becoming larger and larger.In the complex and variable water environment,the hull structure may be damaged due to wave excitation and foreign object attack and personnel misoperation.If the structure of the ship is damaged,it will be an irreparable disaster.Therefore,it is necessary to establish a real-time health monitoring system to ensure the safety of the whole ship.As ships become larger and larger,more sensors will be deployed to monitor the hull structure.Under these circumstances,the possibility of sensor failure will be higher,and at high sampling times,a large amount of data will be transmitted at every moment,which requires high real-time performance of the stress monitoring system.In the specified time,the algorithm must have the corresponding operation speed.The research on the algorithm of stress monitoring system is in the developing stage.Any excellent system design needs to be based on reliable theoretical support.By further studying the optimization algorithm of intelligent data processing and fault diagnosis program design,the false alarm rate and false alarm rate of the monitoring system can be reduced,and the reliability of the monitoring system can be guaranteed.The main contents of this paper are as follows:1.Several key factors affecting the real-time filtering effect of stress monitoring data of ship structures have been studied.This paper presents a symmetric additive floating data window method for orthogonal wavelet transform of discrete time series based on Mallat algorithm.It is used to solve the problem that wavelet transform can not realize time series recursion and has edge effect.Integrated these studies,An improved threshold estimation algorithm and a new threshold function have been proposed to improve the filtering effect of the stress monitoring system.2.The real-time prediction method based on BP neural network has been studied.In order to meet the requirements of long-term monitoring of hull structural health of stress monitoring system,a simple offline training method and a combination of offline training and online training have been proposed.The applicable environment can choose different methods according to different application environments.Based on the data monitoring data of a large-scale ship structure,the setting of important parameters in BP neural network has been studied.3.The particle swarm optimization algorithm has been studied,and the variation factor has been introduced according to the variation theory in genetic algorithm.The particle swarm optimization algorithm of nonlinear inertia weight reduction strategy has been proposed to ensure that the neural network can converge to the global minimum.The feasibility of the new algorithm has been verified by four evaluation functions.Finally,an improved BP neural network method based on particle swarm optimization algorithm has been proposed according to the problems of slow convergence speed and easy to fall into local minimum of BP neural network,and the effectiveness of the new algorithm has been verified by simulation experiments.4.IPSO-BP neural network has been proposed to find the initial training weights and thresholds of the network by using the non-linear inertial weight particle swarm optimization(IPSO)algorithm,enhance the convergence speed and performance of the network,and replace the original signal with the scale coefficient and wavelet coefficient after the large overlap wavelet transform for fault diagnosis.The characteristics and causes of sensor failure in ship structure stress monitoring system have been studied,and the condition of fault identification and diagnosis based on large-scale ship structure stress monitoring data has been analyzed.According to the difficulty of low drift detection rate,an adaptive identification method has been proposed,which can effectively identify faults with any drift rate.