To engender confidence in the application of adaptive control, it is necessary that the properties of various proposed algorithms be examined. For design of discrete-time adaptive controller, an upper bound is required on the plant order and also time delays are to be known. In addition we may have a plant having nonzero d.c.-value or measurable disturbance. Effects of errors in the order of a plant and time delays on adaptive controller were investigated using computer simulation and experiments with a packed distillation column separating a mixture of methanol and water.
The results show that precise knowledge of time delays is of the most while the order of the plant is not serious on the performance of adaptive controller. There were no problems experienced with the plant having nonzero d.c.-value, and performance of the controller can be improved if additional parameter is used to take into account the nonzero d.c.-value. Feedforward adaptive controller was adopted to compensate for measurable disturbance. It's simulation results show a good performance, but experimental performances were not as satisfactory as expected. Nevertheless, a possible improvement can be obtained with feedforward adaptive controller when time delay between disturbance and output is not shorter than time delay between input and output.
It can be concluded from both simulated and experimental results that adaptive control scheme is robust and more effective than PI-controller for single-input single-output systems and multi-input multi-output systems.
