| The ocean is rich in natural resources,and the exploration of the ocean has gradually become a new driving force for world economic development.Marine countries have shifted the focus of their economic development from the land to the ocean and are con-stantly intensifying their efforts to exploit the ocean.As a developing country,China has a large population base and limited land resources,so it is extremely urgent to develop marine economy.Deep-diving submersible vehicles with a certain depth are the prerequisite for the exploration and development of the ocean.If a worker wants to do his work well,he must sharpen his tools first,so the level of exploitation and utilization of a country’s deep-diving submersible vehicles directly affects the depth and breadth of ocean exploration.The structural support material of the deep-diving submersible vehicle is "ocean metal"-titanium alloy,and the stress state of the titanium alloy structure is related to the safety of the crew in the cabin and the deep-diving submersible vehicle during deep-ocean work.However,deep-ocean work environment is extremely complex,and the titanium alloy parts of the deep-diving submersible vehicle are vulnerable to damage,so it is very nec-essary to carry out strict fatigue detection on the titanium alloy parts.The traditional de-tection method is shore detection.Nondestructive testing of all components is carried out by ultrasonic testing,acoustic emission and other technical means,which has a certain time lag and cannot find the damage of the deep-diving submersible vehicle in the first time.In order to realize the real-time detection of the operating state of the deep-diving submersible vehicle,the acoustic emission signals can be identified and detected by tech-nical means in the process of deep-ocean work,so as to judge the fatigue damaged state of the titanium alloy parts in time,so as to make repair,floating and other decisions,and avoid more serious damage.The difficulties in the identification of acoustic emission signals for deep-ocean work are as follows:1.The acoustic emission signals are vulner-able to interference due to the complex deep-ocean ambient noise,resulting in a low iden-tification accuracy.Therefore,in order to identify the signals more accurately,noise re-duction processing should be carried out for the acquired acoustic emission signals.2.Abnormal acoustic emission signals,such as impact and fracture,last for a very short time.As a result,the abnormal acoustic emission signals can only be captured when the sampling window offset is low,and the low sampling window offset will increase the amount of identification calculation,resulting in the calculation cost in the deep-ocean work.This paper studies the recognition of acoustic emission signals for specific deep-ocean work based on principal component analysis and deep learning.The principal com-ponent analysis method is used to suppress deep-ocean ambient noise,and then the acous-tic emission signals are identified and the sampling window offset is predicted by the neural network.The specific work of this paper includes:1.In this paper,by the method of principal component analysis,respectively,the deep-ocean ambient noise of gaussian white noise components and non-gaussian white noise components are suppressed,and the noise reduction result is input to the neural network after cascade,and the noise reduction,classification and recognition of acoustic emission signal fragments are realized.The experimental results show that this method can significantly improve recognition accuracy.2.To realize the acoustic emission signal detection for a long time at a relatively low amount of calculation,this paper,on the basis of noise reduction,take the recurrent neural network on recognition of acoustic emission signals,and forecast sample window offset using RNN,and increase the sampling window offset when the signal is fed to noise,and on the basis of guarantee the recognition accuracy reduce amount of calculation. |
PDF Full Text Request