应用光学, 2020, 41 (4): 761, 网络出版: 2020-08-20
基于PCA和BP神经网络的硝酸盐氮浓度检测方法 下载: 594次
Nitrate nitrogen concentration detection method based on principal component analysis and BP neural network
光谱分析 浓度 神经网络 硝酸盐氮 主成分分析 spectral analysis concentration neural network nitrate nitrogen principal component analysis
摘要
针对紫外分光光度法(UV法)检测混有干扰物质的硝酸盐氮溶液浓度精度不高的问题,提出一种基于主成分分析(principal component analysis,PCA)和BP神经网络的硝酸盐氮浓度检测方法。通过微型光谱仪物质成分检测系统测得硝酸盐氮试剂在196 nm~631 nm波段的吸光度数据,分为测试集和训练集。通过PCA计算训练集,得到主成分。根据BP算法搭建三层人工神经网络。将所得主成分除以8后输入网络展开训练。训练过程中采用留一法交叉验证。用该模型计算训练集和测试集,所得值与真实浓度的平均相对误差分别为2.411 5%和1.553%。实验结果表明,该方法能较好检测出混有干扰物质的硝酸盐氮溶液浓度。
Abstract
Aiming at the problem of inaccurate detection of the nitrate nitrogen solution concentration with interfering substances in ultraviolet spectrophotometry (UV method), a nitrate nitrogen concentration detection method based on principal component analysis (PCA) and BP neural network was proposed. First, the absorbance of the nitrate nitrogen reagent at 196 nm~631 nm was measured by the material composition detection system of the micro-spectrometer, which was divided into test set and training set. Then, the PCA was used to calculate the training set to obtain the principal components. Finally, a three-layer artificial neural network was built based on the BP algorithm. The obtained principal components were divided by 8 and input into the network for training. During the training, the leaving-one method was adopted for cross-validation. This model was used to calculate the training set and test set, the mean relative error between the obtained results and the true concentration is 2.411 5% and 1.553% respectively. The experimental results show that the method can better detect the concentration of the nitrate nitrogen reagent with interfering substances.
陈朋, 严宪泽, 韩洋洋, 吴晨阳, 昝昊. 基于PCA和BP神经网络的硝酸盐氮浓度检测方法[J]. 应用光学, 2020, 41(4): 761. Peng CHEN, Xianze YAN, Yangyang HAN, Chenyang WU, Hao ZAN. Nitrate nitrogen concentration detection method based on principal component analysis and BP neural network[J]. Journal of Applied Optics, 2020, 41(4): 761.