Design and Development of Secondary Clarifier for Paper and Pulp Industry with a Case Study

In this paper mathematical models were created using Artificial Neural Network (ANN) for designing the thickener area of the clarifier by correlating the process control parameters, including the mean cell residence time (θc), initial suspended solids concentration (Co), underflow concentration (Cu), and recycling ratio (R). The test data were applied to the neural network for each value of θc and R. A feed-forward ANN model had been proposed to predict the performance of secondary clarifier. The training time was varied between 0.009 and 22 s. The epochs required for the trained feed-forward network varied between 100 and 500. Training specifications of the ANN model determined that the error was 1e−5 and the training data required 139 epochs. The simulation results obtained by the ANN coincided well with the experimental data. This narrow band of error measured throughout the groups for the modelled parameters was an indication of the robustness of the ANN. Models such as the one developed in this study allow plant operators to assess the expected plant effluent, given the quality of the waste stream at input locations.