Soft sensor for adaptive pH control, an industrial application

Summary form only given. A non linear soft sensor (EKF) has been coupled with a linear adaptive controller to solve pH control problems in actual plants. Direct linearisation of the titration curve (non linear S-shape) to perform classical linear control is not a general solution because this curve, characteristic of the process, is constantly changing as the properties of the inlet flow vary. Our approach enables us to consider the highly non-linear pH control problem as a linear one. It consists of two steps: a soft sensor (EKF) based on a simple dynamic model (derived from mass balances) is designed to provide estimates of the unmeasured acid/base concentrations; a linear controller is then applied for the control of the composition of the outlet flow (instead of the pH) This pH control scheme is based on a simple and classical model of acid-base titration. It ignores the precipitation of any salts. The pH controller can be applied to a wide variety of neutralisation processes such as a continuous stirred tank or tubular reactor, in which an acid flow is titrated with a base (or vice versa), with or without precipitation of salts, pre-titration and re-circulation. Prior to on-line implementation, a feasibility study and the necessary tuning of the controller (and soft sensor) were done by dynamic simulations of the process using SpeedUp^®. The simulations highlighted the good performance of the controller for extreme throughput change (under variable recycle time delay), inlet pH variations and set point changes. Even in the case of significant process/model mismatches, the pH estimate and measurement agreed perfectly. Implementation results are displayed for an effluent treatment unit of a Shell chemical plant. They clearly show the improvement in pH control since this tool was switched on. pH variations have decreased significantly, and operators are now freed from pH control problems.