Study On Correction Of Laser-induced Breakdown Spectroscopy Based On Plasma Image-spectrum Fusion

Laser-induced breakdown spectroscopy(LIBS)has become a research focus at home and abroad because of its advantages,such as fast,in situ,remote,micro-damage,no sample preparation and synchronous analysis of all elements.However,due to the complexity and uncontrollability of the spatio-temporal evolution of laser-induced plasma,the measurement stability is affected by the changes of laser light source,sample properties,experimental environment,resulting in relatively poor quantitative accuracy.Therefore,how to improve the spectral stability and quantitative accuracy is the core problem of LIBS.Focusing on the improvement of spectral stability and quantitative accuracy,this work started from plasma,the source of spectrum,and proposed a new method of simultaneous collection of plasma homologous emission spectrum and image at the same frequency,so as to realize the fusion of image-spectrum multidimensional information.The main research achievements and innovations of this work are as follows:(1)A spectral standardization method based on plasma image-spectrum fusion(PISF-LIBS-1)was proposed to solve the problem of large fluctuation in the original spectra and poor quantitative accuracy.By extracting the features of the plasma image,such as the area and brightness as well as the full width at half maximum of non-self-absorption spectral lines,this method can achieve accurate inversion of these three key parameters of total number density,plasma temperature and electron number density,and carried out standardized correction of the spectral intensity.Further,by using three different matrix of aluminum alloy,alloy steel and ore pressed samples for verification,the experimental results show that compared with the calibration curves established by the original spectral intensity,the R~2 of the calibration curves corrected by PISF-LIBS-1 method increased by15.656%on average,and the RMSE and STD of the verification set decreased by 16.233%and 8.117%on average.(2)On the basis of these three plasma influence parameters,the plasma morphology,self-absorption effect and matrix effect were further analyzed based on the spectral radiation model,and these six key parameters affecting the spectral stability were systematically summarized.Furthermore,a spectral error correction method based on plasma image-spectrum fusion(PISF-LIBS-2)was proposed.This proposed method inverted the above six key influencing parameters through the area,brightness,morphological characteristics,inner and outer layer area ratio of plasma image and the full width at half maximum,interference element spectral line intensity of spectrum,realizing the estimation and containment of spectral error,so as to effectively improve the stability and quantitative accuracy of LIBS.Furthermore,the reliability and robustness of this proposed method were verified by artificially adjusting the laser energy and defocusing amount to simulate the complex detection conditions.The experimental results show that,under complex detection conditions,the R~2 of calibration curves corrected by PISF-LIBS-2 method were all increased to more than 0.974.The RMSE and STD of verification set were further reduced by 24.783%and 22.804%based on the results of PISF-LIBS-1.(3)In addition to these above six spectral stability influence parameters,there are still more and more complex high-dimensional influence parameters in plasma that need to be further explored.Therefore,this work proposed a more universal deep learning intelligent correction method based on plasma image-spectrum fusion(PISF-LIBS-3).This method fully excavated and extracted the implicit features in plasma image and spectrum by deep learning method.Based on this,accurate inversion of high-dimensional plasma influence parameters was carried out,and prediction models of spectral deviation and spectral error were constructed,so as to realize high-precision correction of spectral intensity.The experimental results show that,under complex detection conditions,the R~2of calibration curves corrected by PISF-LIBS-3 reached above 0.988,and the RMSE and STD of verification set were further reduced by 25.191%and 17.059%based on the results of PISF-LIBS-2 method.In summary,starting from the laser-induced plasma radiation model,this work systematically corrected the influential factors of spectral fluctuation by comprehensively utilizing the dominant and implicit features of plasma image and spectrum,and finally realized the high-precision quantitative analysis of LIBS.It lays a solid experimental and theoretical foundation for promoting the rapid development and wide application of LIBS.