Research on Spark Spectrum Signal Processing Based on Ensemble Empirical Mode Decomposition
LI Ming1, 2, LI Yan-bing3, ZHANG Qiao-chu2, SHI Yu-tao2, CUI Fei-peng2, ZHAO Ying1, 2
1. Central Iron and Steel Research Institute, Beijing 100081，China
2. NCS Testing Technology Co. Ltd., Beijing 100094，China
3. Yiyang Vocational & Technical College, Yiyang 413055, China
Abstract：The spark spectrometer based on Charge Coupled Device (CCD) is a kind of spectrometer which is used for element composition analysis. Its output signal is a composite signal of high-frequency CCD effective signal and low-frequency background noise. The effective information of spark spectrum is mainly concentrated in the higher frequency band of signal, which is easy to be submerged and interfered by background noise. Therefore, in order to obtain complete and effective spectral information, it requires effective signal processing. Empirical Mode Decomposition (EMD) method can analyze signals adaptively without setting parameters, but there is the problem of mode mixing, and the components of different frequencies in the signal may be confused; Ensemble Empirical Mode Decomposition (EEMD) successfully solves the problem of mode mixing in EMD method, It can more clearly decompose the different frequency components in signal, so it is more suitable for processing spark spectrum signal with dispersive frequency components. In this paper, the spark spectrometer is used to excite and collect the stainless steel standard samples (Carbon C, Manganese Mn, Nickel Ni, Chromium Cr and Aluminum Al, which are representative elements in the short, medium and long band), and the original spark spectrum signals of the standard samples are obtained. Through the adaptive analysis and processing of EEMD method, each CCD signal is obtained 11 order Intrinsic Mode Function (IMF). According to the amplitude and frequency characteristics of the signal, IMF1-IMF2 is characterized as the high-frequency characteristic signal component, and the last IMF11 is the low-frequency background noise component. By reconstructing the processed signal and combining with the continuous wavelet transform-penalizedleast squares, the final processed signal is obtained. The processed signal is introduced into the instrument data processing software, and the content gradient curve of Carbon, Manganese, Nickel, Chromium and Aluminum elements is obtained. The results show that the signal processed by EEMD method is equivalent to the original instrument processing method, but the additional link of collecting blank noise section is omitted, and the analysis time is largely saved, so the operation efficiency of the instrument is improved.
李 明，李颜冰，张翘楚，史玉涛，崔飞鹏，赵 迎. 基于EEMD方法的火花光谱信号处理研究[J]. 光谱学与光谱分析, 2020, 40(06): 1923-1928.
LI Ming, LI Yan-bing, ZHANG Qiao-chu, SHI Yu-tao, CUI Fei-peng, ZHAO Ying. Research on Spark Spectrum Signal Processing Based on Ensemble Empirical Mode Decomposition. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2020, 40(06): 1923-1928.
[1] ZHANG Kai(张 恺). Electro-Optic Technology Application(光电技术应用), 2018, (6)：44.
[2] XIAO Qi-yang, LI Jian, WU Si-jin, et al(肖启阳，李 健，吴思进). Acta Photonicia Sinica(光子学报), 2018, (3)：1.
[3] LI Ming, ZHAO Ying, CUI Fei-peng, et al(李 明，赵 迎，崔飞鹏). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2020, 40(1)：54.
[4] Huang N E, Long S R, Wu M C, et al. Proc. R. Soc. Lond., 1998, 454: 903.
[5] Wu Zhaohua, Huang N E. Technical Report, 2005, 193: 51.
[6] LI Ming, ZHAO Ying, CUI Fei-peng, et al(李 明，赵 迎，崔飞鹏，等). Metallurgucal Analysis(冶金分析), 2019, (5)：57.
