Research On Weak Thruster Fault Diagnosis For AUV Based On Improved ISOMAP And Artificial Immune Method

With the gradual diminishing of non-renewable resources on land,countries all over the world are also accelerating their process of ocean development.Underwater Vehicle as the only carrier that can work in the deep-sea environment,plays an irreplaceable role in the development of marine resources.AUV(Autonomous Underwater Vehicle)works independently in the complex and volatile marine environment without carrying any cable nor human,so the security of AUV plays an important prerequisite for its practical.The AUV thrusters sustains the heaviest workload of the itself so thrusters have become the most common and serious.Meanwhile the weak thruster faults(thruster output loss less than 10%)are mostly early failures.Currently,there is no mature theory and agree-upon solution.Effectively detect and diagnose early failures to avoid AUV catastrophic failures in early stage has significance importance.With respect to the weak propeller fault feature diagnosis of AUV,the research is mainly conduct from following three aspects,which are the extraction of propeller fault feature,the detection of propeller fault and the identification of the unknown degree fault.Research on the propulsion system of the "Beaver-II" AUV experimental platform.To accurately and effectively carry out the condition monitoring and intelligent diagnosis of the AUV propeller system and to verify the feasibility of the fault diagnosis method,this paper studied and improved the propulsion system of the "Beaver-II" AUV experimental platform.Through the underwater environment,the improved propeller dynamics models have been established.Finally,the experimental method is used to simulate the experimental data.Research on extraction method of the AUV’s fault feature.The working environment of AUV with strong noise interference,is complex and volatile.Because of the complex structure and strong non-linearity of itself,it contains faw information in a single state or control signal of AUV.Aiming at this problem,this paper presents a fault feature extraction method basing on the improved D-S evidence theory and ISOMAP(Isometric Mapping)algorithm.In this method,the phase space is reconstructed after the merge of AUV state signals and control signals.After generating a high dimensional feature matrix,we apply nonlinear dimensionality reduction and then the fault feature in AUV thruster is obtained.Experimental results of the experimental prototype in pool experiments verify the effectiveness of the proposed methods.Research on detecton method of the AUV’s propeller fault.As the extracted propeller features are in the form of two-dimensional discrete points,it’s hard to observe whether the feature points fall in the fault area.Aiming at this problem,this paper presents a fault detection method basing on the improved Artificial Immune algorithm.This method is used to train and study the feature data extracted in the normal operation to generate the detector set,and the fault feature data is detected by the mature detector set.Meanwhile,to obtain the distribution of the fault points at different fault levels and quantify the distribution of the fault points,firstly this paper defines the fault radius based on the Euclidean distance and then uses this definition to measure the fault radius corresponding to the fault points.Experimental results of the experimental prototype in pool experiments verify the effectiveness of the proposed methods.Research on the identification method of the AUV’s unknown degree fault.The artificial immune algorithm is unable to identify the unknown fault when diagnosing the thrusters.And the fault radius defined by the Euclidean distance not only can not reflect the distribution of the fault point,but also can not be the minimum classification surface.Aiming at this problem,this paper proposes an identification method of the AUV’s unknown degree fault basing on the improved Support Vector Domain Description(SVDD).This method establishes the faulted hypersphere of the fault point.By analyzing the distribution law of the fault hypersphere radius,we can identify the unknown degree of failure.Experimental results of the experimental prototype in pool experiments verify the effectiveness of the proposed methods.