1. School of Artificial Intelligence, Beijing Technology and Business University, Beijing 100048, China
2. China Light Industry Key Laboratory of Industrial Internet and Big Data, Beijing Technology and Business University, Beijing 100048, China
Abstract：Inductively coupled plasma atomic emission spectrometry (ICP-AES) has become a conventional elemental analysis method, but in the ICP-AES analysis process, most of the elemental analysis lines will be interfered by background or other spectral lines overlapping. Spectral interference seriously affects the accuracy of spectral line analysis. Therefore, in element analysis, an appropriate spectral interference correction method is needed to obtain a suitable element analysis line. In this paper, according to the characteristic that the spectral intensity is superimposed, the spectral line shape is expressed as a multi-peak spectral line superposition model summed by the Voigt linear function.Moreover, the root-mean-square error of the multi-peak spectral line superposition model and the target spectral line are used to construct a multivariate function as an evaluation function of the mathematical model. The adaptive particle swarm optimization (APSO) algorithm is designed to find the optimal solution of the separated spectral lines’ characteristic parameters. Based on the standard PSO algorithm, the APSO algorithm introduces a compression factor while making the population parameter inertia weight adaptively changed according to the individual fitness value of the particle and the linear change of the learning factor. Coordinate the global search capability and local development capability within the particle population during the algorithm iteration process to ensure that the algorithm effectively and quickly converges and achieve multi-peak spectral line separation. Reduce the influence of interference spectral lines to get more accurate elemental analysis lines. The two samples of light intensity AD sampling values of the two spectral lines at the 390.844 nm characteristic wavelength of the Pr element-containing solution and the 313.183 nm characteristic wavelength of the mercury lamp returned by the ICP-AES detector are used as two sets of measured data, and the two Voigt linear approximation functions. Three superimposed composite curves with different degrees of overlap are used as three sets of simulated data. On the data curve, 50 points that can contain all the characteristic parameter information of the curve are selected as the target data points. By performing the APSO algorithm on the above five sets of target data points, the results show that the relevant parameters of the multimodal spectrum superposition model obtained by the APSO algorithm can fit the corresponding target data curve more accurately. The error is low, and the algorithm shows that it can effectively deduct the interference of spectral line overlap. Under the same set of target data, select the characteristic parameter vector corresponding to the smallest optimal fitness value as the relevant parameter of the Voigt linear function, and the multi-peak spectral line superimposed model curve fitted by this method has higher curve accuracy and smaller relative error. This algorithm has good convergence and adaptability. The algorithm can be applied to the ICP-AES in the element qualitative and quantitative analysis of the study.
廉小亲，刘 钰，陈彦铭，黄 静，龚永罡，霍亮生. 基于自适应粒子群算法的多峰谱线分离方法研究[J]. 光谱学与光谱分析, 2021, 41(05): 1452-1457.
LIAN Xiao-qin，LIU Yu，CHEN Yan-ming，HUANG Jing，GONG Yong-gang，HUO Liang-sheng. Research on Multi-Peak Spectral Line Separation Method Based on Adaptive Particle Swarm Optimization. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2021, 41(05): 1452-1457.
