Research On Instrumentation And Data Analysis Methods Of Portable Laser-induced Breakdown Spectroscopy

Laser-induced breakdown spectroscopy(LIBS)is a rapid chemical analytical technology that uses short laser pulses to generate plasma on the sample surface and analyze its atomic spectroscopy.Portable LIBS has become more and more popular with researchers due to its advantages of the small size,lightweight,convenient operation,and better applicability to complex industrial fields and outdoor harsh environments.However,the current portable LIBS technology is not mature enough yet,and its analysis performance is still limited in some fields,which hinders the progress of this technology in scientific research,industrial application and technology transformation.On basis of this,the system integration technology,data preprocessing,analytical line selection,and quantitative analysis of portable LIBS were studied,and the miniaturized and integrated portable LIBS instrument with improved performance was produced.Detail contents and results of this thesis are as follows:(1)The system integration technology of portable LIBS instrument was studied.The problems of spectral acquisition sequence,dust prevention under harsh environment,weak spectral intensity and poor stability in compact LIBS system were solved.The independent design and integration of the optical system,hardware system and software system of portable LIBS were achieved,and finally,a new generation of portable LIBS elemental analyzer JGTZ-0503 was developed.The total weight of the equipment is 12.05 kg,and it can work for 8 hours under normal working conditions.With the initial application on the fields of geological exploration and metal analysis,the rapid identification of rock and ore and the composition analysis of steel samples were realized successfully.Taking microalloy steel samples as an example,the limits of detection of Cr,Ni,Si,Cu,Mn,V and Ti elements reached 355 ppm,587 ppm,767 ppm,120 ppm,1210 ppm,82 ppm and 85 ppm,respectively.(2)Aiming at the problem of spectral fluctuations in portable LIBS,data preprocessing methods based on the clustering characteristics and statistical distribution of spectral intensities were studied.3 methods namely minimum standard deviation method,minimum distance method,and Weibull distribution method were applied to the spectral data preprocessing of portable LIBS respectively.Based on the above preprocessing methods,the effective spectrum screening and analysis of rock samples were realized.The results show that,with above 3 methods,the relative standard deviations(RSDs)of spectral intensities were improved by 19.11%,24.72%and 16.77%,respectively.Combining this method with the LDA(Linear discriminant analysis,LDA)algorithm,the classification accuracies of rock samples were increased by 5.98%,3.97%and 5.93%,respectively.Among them,the time complexity and actual time consumption of minimum distance method are the least,it thus has better applicability than other methods in portable LIBS.(3)Aiming at the requirements of in-situ and real-time detection of portable LIBS,as well as the problems of manual dependence and inefficiency of traditional manual line selection method,an image features assisted line selection(IFALS)method which based on the spectral morphology and characteristics of Harris corners was proposed.Combining this method with the LDA(Linear discriminant analysis,LDA)algorithm,the time required for the whole classification process was decreased significantly,the classification accuracy of 24igneous rock samples was increased from 94.38%to 98.54%.Therefore,the efficiency and accuracy of the classification were greatly improved by the proposed method.(4)Aiming at the poor quantitative performance of portable LIBS due to the poor representation ability of conventional spectral features,the principles of spectral image generation and feature extraction were studied,and on this basis,a quantitative analysis method based on image HOG(Histogram of oriented gradient,HOG)features and machine learning algorithms is proposed.Taking the quantitative analysis of Ni elements in stainless steel as an example,its coefficient of determination(R~2)was increased from 0.9833 of a conventional spectrum quantitative analysis(SQA)method to 0.9996 of the IQA method by this method,and the average relative error of cross-validation(ARECV)and root mean squared error of cross-validation(RMSECV)were decreased from 56.80%and 1.08 wt.%to15.93%and 1.00 wt.%,respectively.The results show that the proposed method provides an effective approach to improving the quantitative performance of portable LIBS.In conclusion,this thesis starting from the basic structure of portable LIBS,the technology of system integration,data preprocessing methods,the automatic line selection method and quantitative analysis method based on image features were researched.The independent design and integration of portable LIBS have been achieved,and its quantitative performances have been improved,which laid the foundation for the development of the portable LIBS.In conclusion,the independent design and integration of portable LIBS were achieved in this research,and its anti-interference performance and qualitative and quantitative analysis performance were improved,which laid a foundation for the development of lightweight,miniaturization,precision and high efficiency of portable LIBS technology.