Automatic control of pH using an artificial intelligence approach

The control of pH has been widely studied over the last several years due to the nonlinear behavior of the pH system and because it is difficult to control. In this work conventional control and Artificial Neural Networks were used in an experimental study in order to control the pH of an acidic waste water neutralization process. The neural networks were used as a feedforward controller while a PI controller was used as a feedback controller. The neural network feedforward controller consisted of two backpropagation neural networks. One network was trained to model the pH process and to predict the reactor pH based on the disturbances. The other network was an inverse model of the pH process, which used the pH prediction to calculate the base flow rate necessary to keep the system at steady state. This control system was compared to traditional feedforward control. Also, the effect of a buffer solution on the pH process and the control system was studied. The addition of the feedforward controller to the loop improved significantly the performance of the control system; however, the traditional feedforward controller performed better than the neural network controller. Despite this, it can be said that there is much room to improve the performance of the control system by either retraining the networks or using other types of networks.