Research On Spectrum Detection And Recognition Technology In Laser Damage

Laser induced breakdown spectroscopy(LIBS)is a new technology of composition analysis.It has the characteristics of non-contact,real-time online,low requirements for samples and so on.It is widely used in military,aerospace,petroleum,metallurgy many other fields.The main research content of this paper is the target material identification module in the project of "Research on real-time laser damage effectiveness evaluation system",whose goal is to identify the target material quickly and accurately without contact,so as to facilitate the accurate analysis of the target damage state.In this paper,based on LIBS technology,the analysis and recognition of laser damaged targets are studied,and the following research contents are completed:1.In this paper,the theory of laser-induced breakdown spectroscopy is studied,including the generation process,radiation mechanism,plasma model and line broadening theory of laser-induced plasma.2.This paper studies the processing algorithm of LIBS spectral data,including the preprocessing algorithm to reduce the fluctuation of spectral data caused by environment and instrument,compares the denoising performance of different denoising methods for LIBS spectral data,and improves the threshold value calculation method and threshold function of wavelet transform threshold denoising,which has the best denoising effect,so that the SNR of processed spectral data is increased by 2 to 3 dB,RMSE has been reduced about 0.02.Finally,how to extract effective linear spectrum from spectral data is studied.3.This paper studies and compares different spectral data classification methods,including traditional spectral matching algorithm,principal component regression based on multiple linear regression fitting,soft independent modeling analysis and partial least squares discrimination,and support vector machine classification in machine learning field.According to the characteristics of high accuracy but time-consuming of SVM classification,this paper proposes a spectral vector construction method of segmented weighted spectral peak intensity,which can reduce the recognition time of single LIBS spectral data by an order of magnitude on the premise of ensuring the recognition accuracy,and balance the relationship between the recognition accuracy and time-consuming by adjusting the parameters4.Finally,based on the research content of this paper,a LIBS spectral data analysis and recognition software system has been developed by using C + + language.The system can run stably in Windows environment,which proves that the research content and the proposed method in this paper are effective and feasible.The software system can be effectively applied to the "real-time laser damage effectiveness evaluation system research" project Meet the requirements of the project for target recognition accuracy and high real-time.