论文部分内容阅读
提出用果蝇算法(FOA)优化最小二乘支持向量机(LSSVM)的混合模型校正能量色散X荧光(EDXRF)分析中铁和钛的基体效应。使用国产CIT-3000SMEDXRF分析仪、电制冷半导体探测器测得五类矿样共80组光谱数据,每类矿样16组光谱数据。运用FOA优化LSSVM的参数并建立最优模型预测30个样本的钛铁含量,对比化学分析值,FOA-LSSVM预测的钛、铁元素含量与化学分析值的相对误差在1%以内的共26个样本,占总量的86.67%;其余4个样本的钛铁含量预测值与化学分析值一致,占总量的13.33%。另外,运用粒子群算法(PSO)、遗传算法(GA)优化的LSSVM和MATLAB默认的LSSVM模型预测钛铁含量,将其与FOA-LSSVM模型预测的结果进行了对比。综合研究表明,FOA-LSSVM能够实现钛铁元素间基体效应的校正,是一种优选方法。
The matrix effect of iron and titanium in energy dispersive X-ray fluorescence (EDXRF) analysis is proposed using a hybrid model of the fruit fly algorithm (FOA) optimized least square support vector machine (LSSVM). Using domestic CIT-3000SMEDXRF analyzer, electric cooling semiconductor detector measured five types of ore samples of a total of 80 sets of spectral data, each type of mineral sample 16 sets of spectral data. FOA was used to optimize the parameters of LSSVM and the optimal model was established to predict the content of ilmenite in 30 samples. Comparing the chemical analysis values, the predicted error of FOA-LSSVM for titanium, iron content and chemical analysis value was within 1% Samples, accounting for 86.67% of the total; the other four samples of titanium iron content predicted value consistent with the chemical analysis, accounting for 13.33% of the total. In addition, the particle size (PSO), genetic algorithm (GA) optimized LSSVM and MATLAB default LSSVM model were used to predict the content of ferrotitanium, which was compared with that predicted by FOA-LSSVM model. Comprehensive studies have shown that FOA-LSSVM can be used to correct the matrix effect between the elements of ferrotitanium, which is a preferred method.