论文部分内容阅读
目的建立一种基于列文伯格-马夸尔特-反向传播人工神经网络(Levenberg-Marquardt back-propagation artificial neural networks,LM-BP-ANN)的X射线荧光光谱(XRF)的定量检测分析方法。方法采集84个土壤样品光谱数据,预处理后应用主成分分析(PCA)提取特征参数,随机选取训练集、校正集、预测集样品个数分别为42、21、21。以均方差(MSE)、校正决定系数(R~2)、校正标准差(SEC)、验证决定系数(r~2)、预测标准差(SEP)和相对预测误差(RPD)为评价指标,同时分析比较LM-BP-ANN、BP-ANN、PLS三种算法的建模结果,并利用模型预测土壤重金属含量。结果实验确定隐含层神经元数目、学习率和迭代次数值依次为:6、0.1和8,3种建模方法中LM-BP-ANN效果最优,模型的相关系数高于0.98,表明模型有效。结论模型分析快速,可用于实际土壤样品中重金属含量的检测,对于改进X射线荧光光谱仪的检测准确度有着重要的意义。
Aim To establish a quantitative detection and analysis of X-ray fluorescence (XRF) based on Levenberg-Marquardt back-propagation artificial neural networks (LM-BP-ANN) method. Methods The spectral data of 84 soil samples were collected. After pretreatment, principal component analysis (PCA) was used to extract the characteristic parameters, and the training set and calibration set were selected randomly. The number of samples was 42, 21 and 21 respectively. The mean square error (MSE), the correction coefficient (R ~ 2), the standard deviation of correction (SEC), the validation coefficient (r ~ 2), the prediction standard deviation (SEP) and the relative prediction error (RPD) The modeling results of LM-BP-ANN, BP-ANN and PLS are analyzed and compared, and the model is used to predict the heavy metal content in soil. Results The number of neurons in hidden layer, the learning rate and the number of iterations were determined experimentally by 6, 0.1 and 8 respectively. Among the three modeling methods, LM-BP-ANN had the best effect and the correlation coefficient of the model was higher than 0.98, indicating that the model effective. Conclusion The rapid analysis of the model can be used to detect the content of heavy metals in real soil samples, which is of great significance for improving the detection accuracy of X-ray fluorescence spectrometer.