人工神经网络分光光度法测定无机铬的形态

APPLICATION OF ARTIFICIAL NEURAL NETWORK (ANN) TO THE SPECIATION ANALYSIS OF INORGANIC CHROMIUM BY THE COLOR REACTION SYSTEM OF DBPF-CTMAB-EtOH

在二溴苯基荧光铜-溴化十六烷基三甲基-乙醇显色体系中,应用三层人工神经网络解析Cr(Ⅲ)和Cr(Ⅵ)的吸收光谱,分光光度法同时测定Cr(Ⅲ)和Cr(Ⅵ)。Cr(Ⅵ)和Cr(Ⅲ)的质量浓度分别在0～0．287 mg·L-1和0～0．131 mg·L-1范围内服从比耳定律,其表观摩尔吸光系数分别为ε580=1．13×105L·mol-1·cm-1和ε576=8．96×104L·mol-1·cm-1。使用改进的BP网络,避免了网络学习过程中可能产生的陷入局部极小点,并提高了其收敛速度和泛化能力。经合成样品和实际样品分析,考核了方法的回收率和精密度。

In an acetate huller of pH 4. 8, colored complexes were formed between DBPF and either of Cr（Ⅵ） or Cr（Ⅲ） in the presence of CTMAB and ethyl alcohol, with their absorption maxima at 580 nm and 576 nm respectively. The absorption spectra of Cr（Ⅵ） complex and Cr（ Ⅲ ） complex overlapped seriously, making the individual determination impossible. Application of the three layer ANN to the photometric speciation analysis of Cr（Ⅵ） and Cr（ Ⅲ ） by the color reaction mentioned above, the problem of spectra overlapping was solved, and the determination of Cr（Ⅵ） and Cr（Ⅲ） was carried out simultaneously. The apparent molar absorptivities for Cr（Ⅵ） and Cr（Ⅲ） were found to be （ε580） 1.13 × 10^5L · mol^-1·cm^-1 and （ε576） 8.96 × 10^4L·mol^-1·cm^-1, with their linearity ranges of 0-0. 287 mg·L^-1 and 0-0. 131 mg ·L^-1respectively. By using the improved back propagation algorithm, not only falling into part minimum in the training procedure was avoided, but also its convergence speed and spreading ability were raised. Recovery and precision of the method was tested by analyzing several simulated water samples and substantial water samples of tap water and water from East Lake.