Simultaneous Determination Of Cadmium, Lead And Mercury In Water By Feed-forward Networks-Spectrophotometry

In this paper, a data processing system of feed-forward neural networks and spectro- photometry were combined to simultaneously test cadmium, lead and mercury in water. The software Visual C++ 6.0 of the Microsoft developed a data processing system named BPChem ?that was based on the back propagation algorithm. The data processing system which has a Chinese interface can run based on the Windows9x Chinese operating system. It is simple and easy to operate the BPChem because this software has a concise and perspicuous interface, small volume and quick running speed. The best parameters were selected by training the neural networks with the different parameters, and finally: learning rate was 0.85, momentum was 0.35, the 2 ]~ T topology of networks was 21-14-3, and the transfer parameter was 0.5. The determination conditions of these three components were investigated, and the best conditions were proposed. Selecting 5-Br-PADAP as the color-developing agent, adding dodecyl sodium sulfate as the surface- active agent, adjusting pH as 9.0 with borax buffer, cadmium, lead and mercury were determined simultaneously. The limit of detection of Cd2~~ Pb- and Hg2~ respectively were ( ~g/mL) 0.0134, 0.286, 0.0796. The result of samples which included man-made water and waste water met the requirement of error. The design method of uniformity was successfully applied to design the training set, the inspecting set and running parameters.