A partial least squares based spectrum normalization method for uncertainty reduction for laser-induced breakdown spectroscopy measurements |
| |
| Affiliation: | 1. State Key Lab of Power Systems, Department of Thermal Engineering, Tsinghua-BP Clean Energy Center, Tsinghua University, Beijing, China;2. China Guodian Science and Technology Research Institute, Nanjing 100034, China;1. Key Laboratory of Metallurgical Equipment and Control Technology, Wuhan University of Science and Technology, Ministry of Education, Wuhan 430081, China;2. Hubei Key Laboratory of Mechanical Transmission and Manufacturing Engineering, Wuhan University of Science and Technology, Wuhan 430081, China;3. The State Key Laboratory of Refractories and Metallurgy, Wuhan University of Science and Technology, Wuhan 430081, China;4. Department of Mechanical Engineering, New York Institute of Technology, Old Westbury, NY 11568, USA;1. Renewable Energy Resources Lab (RERL), Department of Mechanical and Aerospace Engineering, The University of California, Irvine, CA 92697-3975, United States;2. State Key Lab of Power System, Department of Thermal Engineering, Tsinghua-BP Clean Energy Center, Tsinghua University, Beijing, 100084, PR China |
| |
| Abstract: | A bottleneck of the wide commercial application of laser-induced breakdown spectroscopy (LIBS) technology is its relatively high measurement uncertainty. A partial least squares (PLS) based normalization method was proposed to improve pulse-to-pulse measurement precision for LIBS based on our previous spectrum standardization method. The proposed model utilized multi-line spectral information of the measured element and characterized the signal fluctuations due to the variation of plasma characteristic parameters (plasma temperature, electron number density, and total number density) for signal uncertainty reduction. The model was validated by the application of copper concentration prediction in 29 brass alloy samples. The results demonstrated an improvement on both measurement precision and accuracy over the generally applied normalization as well as our previously proposed simplified spectrum standardization method. The average relative standard deviation (RSD), average of the standard error (error bar), the coefficient of determination (R2), the root-mean-square error of prediction (RMSEP), and average value of the maximum relative error (MRE) were 1.80%, 0.23%, 0.992, 1.30%, and 5.23%, respectively, while those for the generally applied spectral area normalization were 3.72%, 0.71%, 0.973, 1.98%, and 14.92%, respectively. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
