摘要
以食用油中铜、铅、锌三组分重金属同时检测为目的,用差分脉冲溶出伏安法对三种金属混合溶液进行了电化学检测,获得了检测信号。利用平滑、平滑求导、卡尔曼滤波、小波包分析四种降噪方法对检测信号进行了降噪处理;运用主成分分析融合数据,以降低数据维数。同时,借助最小二乘支持向量机回归构建了四种不同模型。通过对预测集检验以及食用油样品的实际测试,得出基于平滑求导的回归模型预报结果较好,且满足食用油中铜、锌、铅检测精度要求。该研究为食用油中多组分重金属含量快速检测提供了一种新手段。
To detect the concentration of Cu, Pb and Zn in edible oils, differential pulse stripping voltammetry was used to detect their mixture solution samples, and their electrochemical signals were obtained. The stripping signals were preprocessed respectively with smoothing, smoothing-derivation, kalman filtering and two-layer wavelet packet analysis for improving the signal-to-noise ratio. Then principal component analysis was employed to merge the signal data so as to reduce data dimensions. At the same time, the regression prediction models of Cu, Zn and Pb corresponding to the four denoising methods were established by least squares support vector machine, respectively. According to the test results of forecasting samples and practical samples, the prediction result of regression model based on smoothing-derivation was better than other methods, and it could meet the detection precision requirements of Cu, Pb and Zn in edible oils. Therefore, a new means for quickly detecting the contents of Cu, Pb and Zn in edible oils is provided in this investigation.
出处
《核农学报》
CAS
CSCD
北大核心
2013年第5期641-646,共6页
Journal of Nuclear Agricultural Sciences
基金
河南省科技攻关资助项目(0324010008)
关键词
食用油
重金属检测
最小二乘支持向量机回归
溶出伏安法
Edible oils
Heavy metal detection
Least squares support vector machine regression
Differential pulsestripping voltammetry
