Optimization of Effluents Using Artificial Neural Network and Support Vector Regression in Detergent Industrial Wastewater Treatment

Abstract

The freshwater is a challenge as the world’s population grows. The largest sources of water in this planet are brackish water and sea water. So, water purification process is very important during this water crisis using desalination and various water treatment techniques. In this paper, we have developed some machine learning approaches for a detergent industry in India. The whole effluent and waste disposal in the detergent industry were treated by different treatment process like air flotation, chemical coagulation, sedimentation and biological treatment through completely mixed activated sludge process. The soft computing techniques (i) a five-layered feed forward ANN (ii) a five-layered cascade forward neural network and (iii) support vector regression have been applied to optimize the proposed models. Training function are considered as Feed-Forward BP(MLP), Cascade Forward BP and SVR where as Training algorithm Levenberg Marquardt and Sequential minimal optimization have been used. Graphical representation has been given for different types of pollutants, effluent treatment plant flow, and Change of color of wastewater after treatment and mathematical operations for Mechanism on Support Vector Regression has been established. To get the best number of neurons for the hidden layer, the network was trained for varied numbers of iterations (Nbest). The data was statistically examined as well. The Nbest value was found to be 10, with the lowest root mean square error (0.066), mean square error (0.0043), and greatest $R^2$ value (0.996); these values show that the predicted and experimental responses are similar, and plant performance was adequately predicted using the backpropagation ANN model thus ANN may be used to describe the process.