Research On Loose Particle Location Method For Sealed Electronic Devices

Sealed electronic device refers to important components within a system or system with specific functions such as satellites,missiles,engines and drones.The problem caused by loose particle has always been an important factor affecting the reliable operation of sealed electronic devices.As the equipment becomes larger,the cleaning of the loose particle is very difficult.The problem of the redundant location is attracting more and more researchers' attention.The positioning of the redundant can help the staff to clean up the loose particle and guide the researchers to prevent and control the loose particle.This paper has carried out in-depth research on the location method of the loose particle in sealed electronic device.Firstly,The redundant location experiment system is built,including the selection and layout of the acoustic emission sensor,a sealed electronic device model is designed,to simulate the structure of the actual electronic device,multi-channel sound collection device was developed,for real-time synchronous acquisition of multiple redundant detection signals,to ensure that the collected detection signal of the loose particle can fully reflect the location information of the redundancy.The traditional acoustic emission source localization method and the classification algorithm in the machine learning field are simply summarized.The advantages and disadvantages of the two methods and the feasibility for the location of the redundant object are analyzed.Secondly,the redundancy signal is classified,the propagation characteristics of the acoustic emission wave caused by the loose particle are studied,the time-frequency domain difference of the multi-channel detection signal is analyzed,and the feasibility of using the attenuation characteristic of the wave to determine the location of the loose particle is verified.The redundancy detection signal is preprocessed,two-level double threshold pulse extraction algorithm is designed for the characteristics of the redundancy signal in this experiment.Then performing time-frequency domain analysis on the redundancy pulse signal to obtain the feature quantity for characterizing the excess object.The location features are analyzed from the correlation,information gain and classification accuracy,and the feature with the greatest value is selected as the location feature.Finally,considering that traditional loose particle location method has low positioning accuracy and limited application,a method based on machine learning for loose particle location is proposed.From the location research of single classifier to the location research of multi-classifier integrated learning,the loose particle location model based on support vector machine and random forest is designed.After the basic model is determined,the location model is further optimized to improve the location performance of the model.The random forest-based redundancy location model has better generalization ability than the single classifier's redundancy location model.The experimental results show that compared with the previous research on the redundancy location,The redundancy location method based on the machine learning solves the problem of low accuracy and poor universality of traditional redundancy location method.