Dissolved oxygen control system in polishing unit using logic solver

The research consists of two parts, the first one is to design the dynamic plant model of polishing unit using artificial neural network (ANN) type backpropagation, and the second one is to design a simulation of a close loop control system on Simulink consisting of logic solver, control valve and ANN polishing unit. The ANN polishing unit was trained and obtained the best model structure 4-24-3 with four inputs chemical oxygen demand (COD) influent, oil in water (OIW) influent, urea, and triple superphosphate (TSP), twenty-four hidden layer nodes, and three outputs (OIW effluent, COD effluent and dissolved oxygen (DO)). The mean square error (MSE) and root mean square error (RMSE) from ANN trained were 0.00485 and 0.06964, obtained by the second iteration. From the simulation results on Simulink by giving several scenarios in the logic solver condition table, the action is brought in the form of urea and TSP nutrition issued by the control valve. The values are used to achieve the DO setpoint (2 mg/L), among others: when COD and OIW influent exceed the quality standard, COD exceeds the quality standard, and OIW does not exceed the quality standard, and the DO error is below zero.