PREDICTION OF WATER QUALITY DEPENDENT PARAMETERS USING ARTIFICIAL NEURAL NETWORKT.V. MALLESH, S.M. PRAKASH, L. PRASANNA KUMAR AND N. JAYARAMAPPA
The present paper deals with the application of Artificial neural network for the prediction of water quality dependent parameters such as distance, temperature, conductivity, dissolved oxygen, total dissolved solids, depth of water, chlorides, phosphates, nitrates, biochemical oxygen demand, total kheldhal nitrogen, fecal coliform, total coliform, fecal Steptococci, before and after the domestic waste mixing zone river Kabini tributary to Cuavery at Nanjanagud, Mandya district (Karnataka, India). The ANN predicted values of water qualities are closure to the actual measured values and laboratory tested values. In this paper we have taken about 150 actual measured and laboratory tested values. For predictions of values using ANN, input and outputs parameters, learning rate parameters, error tolerance, number of cycles (iterations) to reduce the randomly assigned weights are required, to give input to these values, the actual measured and laboratory tested values are used, the learning rate parameter is 0.85, error tolerance is 0.001 and 6000 number of cycles have been chosen. The ANN pattern chosen is 11-12-12-3 (eleven neuron in input layer, two hidden layers of twelve neuron each and three neuron in output layer). Back propagation algorithm has been used to train the network, and delta rule is used to adjust the weights and to reduce the errors. The network predicted values, measured and laboratory tested values as shown in tables and graphs.
Enter your contact information below to receive full paper.