Study Of Quantitative Analysis Approach Of Laser-induced Breakdown Spectroscopy For Steel

Steels are important industrial materials,and the concentrations of elements in steels determine their performance.Therefore,elemental analysis during steel production is necessary and important for quality control.Conventional analytical techniques cannot satisfy the current requirements of steel production due to numerous drawbacks(i.e.,complicated sample preparation,time-consuming measurements and the inability to perform rapid,on-site analysis).Laser-induced breakdown spectroscopy(LIBS)offers unique advantages for steel analysis,including fast,real-time,non-destructive analysis/monitoring,minimal sample preparation and multi-element analysis.However,LIBS is not accurate enough for steel analysis.To address this limitation,the effects of crater development on the LIBS plasma and spectra during LIBS signal collection were studied,and three quantitative analysis approaches based on statistical algorithms were developed.Analytical results were obtained using the quantitative analysis approaches developed in this thesis to analyze metallic elements in thick-wall steel tubes and elemental carbon(C)in the bulk tube samples.The main conclusions and innovations of this thesis are as follows:(1)The effects of craters on plasma and spectral lines during the collection of LIBS signals were studied.The results showed that the mouth and depth of the crater increased with an increasing number of laser shots to ablate the crater.Because of the plasma-crater interaction,with the crater developing,the initial emission intensity of the plasma increased,and the plasma cooled faster.Moreover,as the craters developed,the intensities of the ion lines decreased,and those of the atomic lines increased.(2)A multi-spectral-line calibration(MSLC)approach based on an artificial neural network(ANN)was developed,and the accuracy and precision of LIBS analysis for steels was improved.The root-mean-square errors of cross-validation(RMSECV)for Cr and Ni analysis decreased from 0.018 and 0.067 wt.%(using the internal calibration approach),respectively,to 0.010 and 0.023 wt.%(using the MSLC approach).The average relative standard deviations for Cr and Ni analysis decreased from 11.3 and 19.5%(using the internal calibration approach),respectively,to 6.4 and 12.9%(using the MSLC approach).(3)Based on the MSLC approach,a genetic algorithm(GA)assisted ANN(GA-ANN)approach was constructed to further improve the analytical accuracy of LIBS for steels.The GA-ANN approach used a GA to optimize the intensity ratios of spectral lines,and the optimized intensity ratios were used as inputs to construct an analytical model based on ANN.The GA-ANN approach was used to determine the concentrations of Cu,Cr,Ni and V elemnets in steels;the root-mean-square errors of prediction(RMSEP)decreased by 85.0,67.7,60.0 and 80.0%,respectively,compared with the internal calibration approach.In addition,the respective RMSEP values decreased by 78.6,42.9,27.3 and 33.3%compared with the MSLC approach.(4)A quantitative analysis approach based on the random forest(RF)algorithm was proposed to improve the analytical accuracy of quantifying carbon(C)element in steels under an open-air environment.The quantitative analysis approach was developed by optimizing spectral variables as inputs to train the RF model with the optimal parameters.This analysis model was used to analyze C element in steels,and the RMSEP was reduced by 32.4%,from 0.034 to 0.023 wt.%,compared with the conventional internal calibration approach.(5)The GA-ANN approach developed in this thesis was used to analyze the Cr,Mn,Mo,Ni,Cu and V elements in thick-wall steel tubes and the bulk tube samples.The concentrations of these elements were measured by spark discharge optical emission spectrometry as a reference.The average relative errors for Cr,Mn,Mo,Ni,Cu and V were 8.9,8.7,24.2,5.1,6.1 and 7.1%for the bulk tube samples,respectively.Moreover,the concentrations at different locations along axial direction of the steel tubes were also measured using the GA-ANN approach.The results demonstrated that the concentration distributions along the steel tubes can be obtained using this approach.The analysis approach based on RF presented in this thesis was used to determine the concentration of C element in the bulk tube samples in open air,and the RMSEP was 0.064 wt.%.Thus,the quantitative analysis approaches developed in this thesis can be used to monitor the production of thick-wall steel tubes.In summary,this thesis studied the effects of craters on plasma and spectral lines during the collection of LIBS signals.The MSLC,GA-ANN and RF approaches were developed based on statistical algorithms,and all three models improved the analytical accuracy of LIBS for steel analysis.In addition,these quantitative analysis approaches can be used for quality control of steel tubes.