Development Of Signal Recognition Technology For Sealed Electronic Components Based On SVM

In recent years,with the rapid development of China's space military and other fields,the problem of surplus has become an important issue limiting the performance and safety of sealed electronic components.Particle noise collision detection is a factory-tested test for the detection of excess matter.Since the component signal is similar to the output waveform of the reminder signal,even if the component signal covers the reminder signal at the output,the conventional particle noise collision detection system cannot accurately determine the presence or absence of the reminder signal,which directly leads to the error of the excess signal.Judging and missing judgments.It not only affects the safety of sealed electronic components,but also directly reduces the safety of aerospace.This paper is carried out under this current situation.The goal is to build an efficient support vector machine classification model in Python environment,so as to optimize the classification and identification of component signals and reminder signals,solve the misjudgment and miss judgment of the current surplus,and formulate it.Incoming component signal identification preparation rules.In this thesis,the PIND signal measured by the PIND hardware detection system is firstly used,and the three-threshold extraction pulse algorithm,zero-padding method and discrete Fourier transform method are used to process the regular pulse signal to obtain the research significance that can be recognized and calculated by the computer.sample.After that,the component signals and the reminder signals obtained by the experiment are analyzed,and the features that can effectively distinguish the two signals are selected and recorded.Sample data with classification features and label definitions is then imported into Python and SVM classification models for different kernel functions are trained.By comparing the classification performance,the RBF kernel function with the best classification effect is selected.Later,the optimized mesh search method is used to find the optimal hyper parameter combination(C,gamma)of the RBF kernel function,and it is substituted into the SVM classification model to obtain the most ideal classification recognition effect.Finally,the optimized SVM classification model is saved to the database,so that it can be directly called during subsequent data processing to realize the actual application.The classification accuracy of the SVM classifier model finally achieved in this paper is as high as 92.8%,which effectively improves the misjudgment and missed judgment of component signals and reminder signals,and It has important research significance for the safety detection of sealed electronic components.