Research On Fault Diagnosis And Fault Tolerance Control Of AUV Based On Neural Network

The economic value and military significance of the ocean have become increasingly prominent,so the ocean has been paid more and more attention by all countries of the world.As the core equipment for exploring,researching,utilizing and protecting the ocean,the autonomous underwater vehicles have been widely used in the civil and military fields,and it advances the process of developing oceans.However,AUV is unmanned and has no cable,besides the internal system of AUV is high autonomy and the navigation time is long.In the depths of the ocean where the work environment is extremely complex,the operational risk is difficult to predict,once the failure of its internal components happens,it will lead to inestimable losses.Therefore,in the process of developing or applying AUV,safety and reliability are the key issues that must be considered carefully.Fault diagnosis and fault tolerant control,as a new interdisciplinary subject,provide technical support and guarantee for security and reliability in the development and use of AUV.It has important research significance and practical value of engineering.The structure of AUV is complicated,and the fault mechanism,fault sign and fault mode which need to be considered is too more,that has exceeded the limitation of the study time.Therefore,this paper takes the neural network,fuzzy system and the AUV six degree of freedom equation as the theoretical basis,and only takes the key components such as the sensors and thrusters witch commonly used in the AUV control system as the research object.Then this paper makes the actual navigation data of my company's product as the training sample witch is under studying,and takes up research for fault diagnosis and fault-tolerant control of AUV's sensors and propellers.The main works are:(1)Basing on RBF and OS-ELM neural network established online fault diagnosis model of dynamic changing training sample,and on the basis of this model designed online fault diagnosis scheme of a product's depth sensor,then take the product's voyage data as training samples to carry out simulation test,the result shows that the scheme is effective,feasible and can accurate recognize sensor's varieties of fault patterns.(2)Researching neural network information fusion theory in depth,and according to the characteristic of redundant information between each sensor of a product control system,the paper designed the sensor signal reconstruction scheme of the control system,and taked the test data of the product as a sample to carry out simulation test,the result shows that the scheme is effective,feasible and can reconstruct the output signal of a fault the sensor in a certain precision range.(3)Combining fuzzy mathematics with neural network algorithm,the paper established the thruster fault diagnosis model based on fuzzy neural network,and the fault diagnosis scheme of the propeller is designed on the basis of the fault mode of the motor of a certain product propeller,finally let the real survey data of the product as sample to carry out simulation test,simulation result shows that the scheme is effective and feasible,and it can identify the fault modes of the propeller accurately.(4)Analyzing the mechanical properties of AUV in depth,the paper established in six degrees of freedom equation,and taked the equation of six degree of freedom as the foundation,designed fault-tolerant control scheme of dual thruster configuration,finally simulation tests were carried out using Matlab/Simulink software,the simulation result shows that the scheme is effective and feasible,and the fault tolerant control can be realized in a certain precision range.