摘要
利用激光诱导击穿光谱(Laser Induced Breakdown Spectroscopy, LIBS)技术采集得到磷精矿浆的光谱,基于无信息变量消除(Uniformative Variables Elimination, UVE)方法筛选出与磷元素相关的波长变量,将选择后的波长作为自变量建立偏最小二乘(Partial Least Squares, PLS)回归模型,并与传统全谱PLS和GA-PLS(Genetic Algorithm-Partial Least Squares, GA-PLS)定标模型进行比较。相比全谱PLS,UVE-PLS定标模型的性能更优,其预测均方根误差(Root-mean-square Error of Prediction, RMSEP)由0.38%下降到0.26%,决定系数(R2)从0.59提高到0.72。相比GA-PLS定标模型,UVE方法可以克服GA(Genetic Algorithm, GA)在参量选择上存在随机性的弊端,筛选出的变量仅为全谱的8.76%,而且计算速度更快,分析精度也优于GA-PLS模型。
The spectrum of phosphorus concentrate slurry was collected by laser-induced breakdown spectroscopy(LIBS)technology, and the wavelength variables related to P element were screened out based on the uninformative variable elimination(UVE)method, and then the selected wavelength was used as an independent variable to establish a partial least squares(PLS)regression model, which is compared with traditional full-spectrum PLS and GA-PLS calibration models.Compared with full-spectrum PLS,the UVE-PLS calibration model performs better, because its root mean square error of prediction(RMSEP) decreased from 0.38% to 0.26%,and the coefficient of determination(R~2)increased from 0.59 to 0.72.Compared with the GA-PLS calibration model, UVE method can overcome the lack of randomness in parameter selection of GA.The selected variables are only 8.76% of the full spectrum, and the calculation speed is faster, and the analysis accuracy is also better than the GA-PLS model.
作者
房胜楠
史烨弘
韩鹏程
赵振
李华昌
FANG Sheng-nan;SHI Ye-hong;HAN Peng-cheng;ZHAO Zhen;LI Hua-chang(BGRIMM MTC TECHNOLOGY CO.,Ltd.,Beijing100176,China)
出处
《矿冶》
CAS
2023年第1期109-114,共6页
Mining And Metallurgy
基金
国家重点研发计划项目(2021YFC2903103)。
作者简介
第一作者:房胜楠,硕士,主要从事在线设备研发。E-mail:fangsn123@163.com。
