AbstractA new self‐tuning control algorithm appropriate for industrial systems which exhibit non‐linear characteristics and fast dynamic behaviour is proposed. The resulting self‐tuning controller makes use of a weighted sum of the control input generated from a standard self‐tuning control algorithm combined with a set of control input suggestions evaluated retrospectively. Use is made of a past sequence of steady state gains evaluated over a prespecified retrospective horizon together with a filtered prediction of the current system output which is obtained from a steady state Kalman filter. Adopting such an approach provides an element of corrective action based on information which would otherwise be disregarded by a standard self‐tuning scheme.Simulation studies involving a number of identified linear models of an industrial hydraulic servo system are presented. In terms of reduced control input variance and increased accuracy of set‐point following, the effectiveness of the new retrospective self‐tuning controller is compared to that of a standard self‐tuning pole placement controller and results indicate that significant improvements in overall performan
展开▼