Research On Weak Thruster Fault Diagnosis For AUV Based On Improved Stochastic Resonance System And Gray Relational Theory

With the reduction of non-renewable resources on land,the marine development process is accelerating.At present,underwater robots are the only equipment that can work in the deep sea.They play an indispensable role in the development and exploration of the oceans.For AUVs working in a complex marine environment,safety is an important indicator.Accurate and effective AUV status monitoring and fault diagnosis is the key technology to ensure its safety.As the heaviest and most frequently used component in AUVs,thrusters accurately diagnose possible faults in thrusters and greatly improve AUV safety.At present,in the research of AUV propeller fault diagnosis technology,most of the research focuses on the fault diagnosis technology which has a greater degree of thruster output loss,and the output loss of the propeller is less(Propeller output?10%)weak fault related research is less.The weakness of thrusters is usually predominant faults.If these early weaknesses of thrusters can be effectively detected as soon as possible,AUV can make corresponding judgments and treatments in a timely manner to avoid more accidents.In this paper,aiming at the weak fault diagnosis of AUV thrusters,this paper focuses on the characteristics enhancement and interference suppression of thrusters with weak fault signals.It studies how to effectively extract the fault characteristics from weak fault signals and studies how to realize the fault characteristics based on the extracted fault signals Unknown method to identify the degree of failure,the main work is as follows:AUV thruster weak fault signal characteristic enhancement and interference suppression method.In order to realize the signal feature enhancement and interference suppression,this paper deals with the fault signal of AUV based on the stochastic resonance method.After finding that the traditional stochastic resonance system has no obvious effect on signal enhancement,this paper presents a method based on the improved artificial fish swarm algorithm to optimize the stochastic resonance system to improve the stochastic resonance system to enhance the periodic components of the signal.Aiming at the problem that the non-periodic component of signal is not obviously enhanced by the principle of stochastic resonance,this paper also presents a new idea from the point of view of adjusting the signal: Firstly,adjust the fault signal,After the wavelet decomposition and reconstruction of the pre-processing,and its modulation and demodulation filter processing,to further enhance the stochastic resonance system for fault signals in the non-periodic components to enhance the characteristics and interference suppression.The experimental results show that the method proposed in this paper has better feature enhancement and interference suppression than the traditional stochastic resonance system.How to extract the fault feature of the enhanced signal.Aiming at the problem of how to accurately extract the fault feature from the weak fault signal after feature enhancement and interference suppression processing,in order to avoid the fault feature extraction from being insufficient,a signal time domain combining S transform and FFT transform is proposed.Frequency domain and time-frequency domain.In order to improve the capability of extracted fault feature set to characterize fault features,according to the statistical characteristics of signal and the feature quantity of this paper,we should emphasize the requirements of fault size.In this paper,The calculation formula was modified.Through experimental comparison and verification,the multi-dimensional feature extraction method used in this paper can effectively characterize the magnitude of the fault features in the time domain,frequency domain and time-frequency domain.The improved fault features proposed in this paper can be more effectively compared with the pre-modification features Characterize the degree of fault characteristics.AUV method to study the unknown degree of fault of AUV thruster.Because of the large span of the fault feature signal in this paper,the traditional method of feature extraction in time domain or frequency domain will be limited in the processing of such multi-dimensional information.Aiming at the problem of how to identify the fault degree of multidimensional fault features,a method of judging the fault degree of unknown signal based on gray relational theory is proposed in this paper.Due to the small difference between the different weak fault signals in this paper,the traditional method of calculating the correlation degree is difficult to accurately characterize the magnitude of each fault.This paper finds that there is a small correlation between different types of signals,and then proposes the calculation of the correlation degree of classification association by signal type.This paper also found that due to the change of some feature values is relatively small lead to the inadequate use of this part of the feature,and then put forward a relative feature to replace the absolute feature amount calculation method.Considering that the signal values in the period of stable operation of AUV are in accordance with the normal distribution theory of statistics,this paper presents a method to establish the normal distribution model of correlation degree of each reference fault signal.The experimental results show that the classification correlation method proposed in this paper is more accurate than the traditional correlation calculation method.In this paper,the relative characteristic quantities proposed in this paper are more fully utilized than the traditional absolute characteristic quantities.In this paper,the gray correlation theory is used to determine the unknown signal failure Compared with the traditional method of calculating the degree of correlation,the method of degree of accuracy is more accurate for identifying the fault degree of AUV weak fault signal.