Frequency response adaptation of PI controllers based on recursive least-squares process identification

The frequency domain approach to adaptive control or autotuning of proportional plus integral (PI) controllers is considered in this paper. The frequency response of the process is computed from an overparametrized recursive least-squares (RLS) model and used at sample intervals to tune the PI. By simulations it is shown that the frequency response estimates converge more quickly and are more constant in the presence of disturbances than the RLS coefficients from which they are calculated. The frequency response estimates are therefore more reliable for controller tuning. The method can handle concurrent gain variation of 50% and time constant or delay variations of 100%. The method is implemented and experimentally verified on a pilot distillation column.