Fault Diagnosis And Emergency Decision Of UUV Under Generalized Environment

With the rapid development of artificial intelligence,materials and energy,the mission capabilities of unmanned underwater vehicles(UUVs)also improved,and the mission tasks have become more remote,dangerous,complex and need less human intervention,thus requiring to increase the intelligence level of UUVs gradually.Fault diagnosis and emergency decision making are important manifestations of the intelligence level of a UUV,which act as important support for a UUV in its autonomous capacity and viability under generalized environment.Based on the fact that the emergency decision behavior of a UUV can be affected by generalized environment such as nature marine environment,social environment of the battlefield situation,and the internal state of the environment individually,this paper aims at solving key problems in practical engineering when failures accure in a UUV,with researches on the problems of fault diagnosis of sensors and actuators,situation assessment of the generalized environment,as well as emergency decision making.The main results achieved in this dissertation are summarized as follows.Multi-agent architecture for fault diagnosis and emergency decision of UUV under generalized environment are constructed.According to the characteristics of the UUV mission without human interference under generalized environment,the mission process of a UUV is designed.And by combining UUV failure event with the complexicity,hierarchy,and knowledge distribution of a UUV,a UUV hierarchical multi-agent system architecture with fault diagnosis and emergency decision is designed.The fault diagnosis for actuators and sensors is proposed.Aiming at the nonlinearity and uncertainty of a UUV model,a fault detection method for actuators based on nonlinear sliding mode observer is proposed.For low speed when the course and speed of a UUV are easily interfered,an adaptive threshold selection method based on the current speed is proposed.For the case that the data of the sensors has high randomness,and uneasy to modeling,the optimized GM(2,1)fault diagnosis method based on historical experience group particle swarm(GHEPSO)is proposed.The generalized environment estimation method based on Bayesian network model is proposed.With the diversity and complexity of the generalized environment,the paper proposes a thread assessment method for generalized environment based on Bayesian network.This paper also analysed and described the attributes of the multi-source of generalized environment situation,and established the UUV mission posture assessment model.As the Bayesian network can be applied to multi-source data fusion,we proposed the multi-source reasoning method and sensitivity analysis method for generalized environment task situation based on Bayesian network.The study presents the generalized environmental emergency decision method based on the UUV influence diagram model.For the complexity and uncertainty of the environment when failures occur,and in order to guarantee the stability of the controller as well as the response of emergency decision,we proposed an interactive emergency decision method based influence diagram and utility theory with reasoning,prediction,and plan selection.During the process of plan selection,we used the influence diagram of the decision node and the utility node,as well as the estimation result of the situation to ratiocinate and predict the validity of decision plan on each decision node and choose the best plan according to the maximum utility principle.The UUV actuator sensor fault diagnosis method,generalized environmental situation assessment method,emergency decision-making method,as well as failure diagnosis and the multi-agent emergency decision making structure constructed in this thesis actually solved practical problems in the engineering applications,and have high practical value in UUV or other marine application projects.