| Abstract [eng] |
Biotechnological processes are among the most complex control objects with all the characteristics that make the control difficult: non-linear relationships between process variables, time-varying dynamic properties and lack of sensors that can provide reliable process monitoring. The need for adaptive control algorithms is high, and they are needed to develop new and improve the already available biotechnological processes in both scientific laboratories and industry. The academic community has proposed various adaptive control methods; however, they tend to be complex, and they also require a lot of time and knowledge to develop and fine-tune. In this doctoral dissertation five easy-to-implement adaptive control algorithms that are based on fuzzy logic, gain scheduling, statistical and polynomial analysis and substrate feeding profile adaptation are presented. The main advantages of the developed adaptation techniques can be summarized as: a simple model structure which relies on the process operator’s level of knowledge and basic mathematical operations; usage of only controller input/output signals for controller tuning parameter adaptation; minimization of the required soft-sensor measurements for the realization of controller tuning parameter adaptation. To evaluate the performance of the developed systems, the proposed models were compared with standard PI controllers with fixed parameters or similar adaptive control techniques. |
