Comparative Analysis Of The Classification Characteristics Of Particle Collision Noise Detection Signals

Sealed electronic components are important basic components in aerospace and military equipment.Excess is the introduction of metallic or non-metallic particles due to the immaturity of existing production processes or human negligence.May seriously affect the reliability of components.Particle collision noise detection is the most widely used residue detection method.Signal characteristics play an important role in component recognition and material recognition of sealed electronic components.Previous researches mainly focus on the improvement of classification algorithm or pulse extraction method.The pulse features used are mainly based on the previous detection experience,and there is no systematic research on the features of the detection pulses,and no comparative analysis on the advantages and disadvantages of the classification features.In this paper,based on the particle collision noise detection method,the classification characteristics of component signal recognition and material recognition of residues are studied.This paper first refers to feature extraction methods in the field of voiceprint recognition to introduce or establish new features to increase the number of features.Then the characteristic data is cleaned and the abnormal data and missing data in the characteristic data are processed.After that,high quality features are selected for feature construction,and a new residue detection feature set is established by using the original features and post-construction features.After the new feature set is established,the top features are selected with different methods,and on this basis,feature construction is carried out to form a large number of new features.Finally,different feature selection methods are adopted to select the best feature combination.In this paper,the random forest algorithm is used to compare and verify theselected new classification features.The verification results show that the classification accuracy of the new classification feature is improved compared with the original feature.And the new classification features play an important role in signal recognition.The validity of the research results is proved.