A Method Of Recognition On Abnormal Behaviors For UUV Emergency While Man-not-in-the-loop

Unmanned Underwater Vehicle (UUV) has gradually become one of the effective marine explorate in marine environment and equipment, with the rapid development of ocean development in recent years. Due to the harsh conditions of work, the loose of UUV caused enormous porperty damage. The provable security of UUV has become a major research in this area. In order to improve the safety and reliability of the UUV, the system and model were analyzed in this paper. According to the problem that how the UUV identify its own abnormal behavior and conduct emergency measure when man out loop, the main contents in this paper are shown as following:Firstly, for the identification of abnormal of perceptual system, sensor error model was analyzed and the original GM (2,1) model improved. With the improved GM (2,1) model,identification of perceptual system abnormal realized. The four kinds of abnormal of UUV sensors were simulated with fiber-optic gyrocompass, the results show that improved GM (2,1)model recognize the sensor fault in time.Secondly, for the fault-tolerant control of UUV sensor, BP neural network improved by Genetic Algorithm (GA) is applied in the heading fault-tolerant. Train the optimized GA-BP neural network when UUV works well, and reconstruct the abnormal sensor's output with current data from other sensor when there is an sensor fault. The fault-tolerant control of UUV sensor is simulated with fiber-optic gyrocompass. The results show that the GA-BP neural network can achieve fault-tolerant of UUV sensor well.Thirdly, when comes to the identification of implement agency. In the case of perceptual system works well, the identification of implement agency commenced based on UUV behavior. Develop identification tasks and unusual point location rules according to the actuator feature of study UUV. Slide mode observer was used to indentify wether UUV abnormal. If UUV fault, execute identification task and lotating the fault point by comparing the unusual point location rules, using fuzzy rules determining abnormal levels. The simulation results show that the unusual point location rules are accuracy and completeness.The methods which used to identify the abnormal behavior based on behavioral were simulated with rigtht thruster.The results show that the methods above are effective and feasible.Finally, for method to deal with the fault, fault tree is used to analysis the anomalies which the study UUV may exist; eight kinds of emergency measures proposed according to the characteristics of UUV; Fuzzy expert system established, abnormal state and emergency strategy as the input and output of the emergency system were measured, and a knowledge base was established. Fuzzy expert system knowledge repository were established in MATLAB,UUV emergency decision were simulated by set left thruster force halved and rudder card respectively.