Design Of The Water Quality Analysis System By Robotic Fish Based On Fuzzy BP Neural Network

Water quality analysis is not only the foundation of water resources planning andmanagement, but also a significant method of water resources protection, whichmonitors and evaluates the water quality in a quantitative way. Traditional water qualityanalysis system can’t follow the system’s desire of intelligence, flexibility andcompleteness. The development of Artificial intelligence algorithms and robot technologyprovides a new method in this field.This thesis states the current research situation of water quality analysis, presents thelimitation of single evaluation method, and proposes a new model which combines BPneural network and fuzzy Mathematics. Considering the particular characteristics of waterevaluation, and comparing the convergent speed and accuracy of BP neural network indifferent structure by simulating in Matlab, a3-layer Fuzzy-BP neural network model of4-8-1structure is designed. The network studies standard input output training sample databased on LM adaptive algorithm, takes the BP network’s output as the input of the fuzzysystem, selects the triangular membership function and Gaussian membership function tocalculate the degree. The final output evaluate results are given based on the biggestsubordinate principle, which was accurate and objective. Based on the principles ofbionics, a robotic fish platform is designed to monitor the water quality, includinghardware design based on LPC2368and software design using C and assembly language.The system applies LabWindows/CVI to design the GUI, which was used for remotecontrol and data exchanging. What’s more, a remote on-line upgrading technology base onIAP is presented. Applying the system to analysis the water quality of the lake in WuHan,the compare with the evaluation of other method proves that the new design is reasonableand superior.