| Water distribution system is an important part of water supply system which assumes responsibility to carry water from waterworks to human beings. The finished water generally can meet drinking water standards, so the merits of water distribution system impacts on the health of people directly. Turbidity is one of the most conventional in water quality monitoring indicators which reflects the comprehensive status of drinking water in water distribution systems. Meanwhile, drinking water turbidity index is also one of the most common sensory properties index. Increased turbidity and discoloration of drinking water will cause a large number of user complaints, which brings difficulty to the water supply enterprise. Therefore, reducing the limit-exceeding problems of turbidity in WDS is always one of the hot issues and major challenges for water suppliers worldwide.Based on the statistical analysis of the research status of water quality of WDS at home and abroad, the water turbidity properties and evolution rule was discussed and analyzed. The general regularity of turbidity index in WDS is summarized; three aspects of great impact on the turbidity in WDS are filtered out to be focused research, i.e., hydraulic conditions(flow rate), water age, and total iron concentrations in WDS. Through the dynamic experiments by the water supply network simulation platform that simulates the state of water in WDS, the three selected sets of factors to the water turbidity changes in WDS are researched. Results show that the main processes to cause the increasing of water turbidity are pipeline corrosion, iron release and hydraulic flushing against the pipe wall. The pipeline corrosion and iron release in steel pipes, cast iron pipe and other metal pipes are serious particularly, so that the problem of increased turbidity in WDS is more common than other tubular products. The increased rate of turbidity will become slow when the stability of the pipe scale layer increases. The velocity also has a great effect on the increasing of water turbidity, which displays in two aspects: the velocity affects the processes such as diffusion, transfer of related factors in water, then indirectly affect the process of pipeline corrosion; the hydraulic flushing against the pipe wall makes loose particles on the surface of pipe scale layer enter the water and will damage the stability of pipe scale layer, which causes the increase of water turbidity. In addition, destruction of the stability of pipe scale layer at the same time may cause the rapid increase of water turbidity.On the basis of the above data, building BP neural network and RBF neural network model separately by using MATLAB. Selecting TRAINSCG as training function, TANSCG as the hidden and output layer functions in BP neural network, performance goal is 0.001, and spread constant is 1.0 in RBF neural network. Using velocity, water age, total iron content and the adjacent turbidity as input and corresponding turbidity as the output of the neural network. Selecting the appropriate sample to the built BP neural network for training, and applying the trained neural network to the simulation of test samples. The simulation results are quite good. However, the non-linear approximation and the prediction accuracy of the RBF are better than the BP neural networks. |
PDF Full Text Request