反弹传播神经网络用于痕量铬的示波计时电位法测定

首次将反弹传播算法神经网络用于铜箔钝化液中痕量铬的示波计时电位法测定。探讨了网络层数、层结点数和结点转移函数等网络参数对预测结果的影响。实验结果表明 :Cr 浓度在 4.0× 10 -7～ 1.3× 10 -6mol/L范围内与示波计时电位曲线上的切口深度呈线性关系 ,检测下限可达 8× 10 -8mol/L ;与标准BP神经网络的训练和预测结果相比较 ,反弹传播神经网络用于示波测定时不仅具有较高的预测精度 ,而且大大提高了网络训练的收敛速度

Application of Resilient Propagation Neural Network in Determination of Trace Chromium with Oscillographic Chronopotentiometry

A resilient propagation neural network (RPNN) was first appliedin the determination of trace chromium in the passivation solution of copper foil with oscillographic chronopotentiometry. The effect of parameters about RPNN on the prediction results was discussed. Experimental results showed that the incision depth of Cr(III) on dE/dt-E curve was rec tilinearly related to the concentration of Cr(III) in the range of 4.0×10-7 to 1.3×10-6 mol/L, and the detection limit for determination of Cr(III) was 8 ×10-8 mol/L. Compared with the application of standard BP neural network technique to the oscillographic chronopotentiometric determination, RPNN has the advantage of higher predication accuracy and faster convergent rate.