Zhou Feng-Bo, Li Chang-Geng, Zhu Hong-Qiu
School of Information Engineering, Shaoyang University, Shaoyang, China.
School of Physics and Electronics, Central South University, Changsha, China.
Front Chem. 2022 Feb 10;10:839633. doi: 10.3389/fchem.2022.839633. eCollection 2022.
Aiming at the problems of low accuracy and large prediction errors caused by the serious overlap of multi-metal spectral signals in zinc smelting industrial wastewater, a characteristic interval modeling method is proposed. First, according to the absorption spectra of mixed solution, the characteristic intervals of copper and nickel are preliminarily screened by using different partition lengths. Second, take the smallest root mean squares error of cross validation and the largest correlation coefficient as the evaluation indicators, compare the full-spectral model and each local model, and select the optimal feature sub-intervals of copper and nickel. Last, the partial least squares method is used to model the combined wavelengths of the optimal sub-intervals to realize the simultaneous detection of copper and nickel. The linear determination ranges are 0.3-3.0 mg/L for copper and nickel. the correlation coefficients of copper and nickel are 0.9974 and 0.9966, respectively. The results show that the method reduces the complexity of the wavelength variable screening process, improves the accuracy of the model, and lays the foundation for the accurate analysis of polymetallic ions in zinc smelting industrial wastewater.
针对锌冶炼工业废水中多金属光谱信号严重重叠导致的精度低、预测误差大等问题，提出了一种特征区间建模方法。首先，根据混合溶液的吸收光谱，利用不同的划分长度初步筛选出铜和镍的特征区间。其次，以交叉验证的最小均方根误差和最大相关系数作为评价指标，比较全光谱模型和各局部模型，选取铜和镍的最优特征子区间。最后，采用偏最小二乘法对最优子区间的组合波长进行建模，实现铜和镍的同时检测。铜和镍的线性测定范围均为0.3 - 3.0 mg/L，铜和镍的相关系数分别为0.9974和0.9966。结果表明，该方法降低了波长变量筛选过程的复杂度，提高了模型精度，为锌冶炼工业废水中多金属离子的准确分析奠定了基础。
