On designing algorithms for controlling pH

This thesis describes an investigation of how techniques of modelling, estimation and control can be used to improve regulation of pH. A novel mathematical model of neutralization is used which explicitly accounts for two separate sources of non-linearities in pH systems: the inherent non-linearity of the pH measurement and the variable and uncertain non-linearity of the neutralization chemistry. A detailed interactive simulation package was developed. It has been used to study: 1. The application of a Bayes' non-linear estimator to a noisy pH measurement. The estimator is used in conjunction with an Extended Kalman Filter in order to generate estimates for use in combined feedforward/feedback controllers. A novel version of the Bayes' estimator is developed which successfully estimated jumping and drifting bias on the pH measurement.2. A novel cautious stochastic control law which uses the generated estimates and allows for their uncertainty. Reduced quality of estimation forces the controller to regulate the pH at an optimally tuned target value where control is easier.3. Comparisons of a variety of modern and conventional control algorithms. The modern algorithms include three self-tuners, the non-linear estimators and the cautious controller. The conventional algorithms include a PID controller, analytical linearisation of the pH measurement and the use of estimated or measured feedforward signals. Twenty-five pH control laws are simulated. The quality of regulation which can be achieved depends on what prior information is available about the neutralisation process. A strategy is proposed which matches controller structure to available information. This strategy provides guidelines to the control configurations that can be implemented as well as their relative limitations. It is shown that improvement in performance over a PID regulator is possible if one uses more complex controllers. This improvement is investigated by simulation because it varies from one process to another. It is also shown how the resulting simulation package has been used for computer aided design of pH control systems in the chemical industry.