| Since the time self-tuning control theory was developed, many practical applications have been presented. Although the several industrial applications have demonstrated the power of adaptive control, adaptive controllers have not been commonly accepted because of the complexity and expertise required for implementation.;The accomplishment of this work is that a non-specialist can implement an appropriate adaptive control scheme using only the input/output data collected from a process and the software tool presented in this thesis.;A simulated steam boiler, which is highly nonlinear, interactive and nonminimum phase, was developed as a test ground for the adaptive control design method proposed. This boiler model fully describes the dynamic responses between the inputs (water flow, fuel flow and air flow) and the outputs (steam temperature, steam pressure and flue gas emission).;A Model Order Identifier is developed in this project to extract model order information from the input/output data pairs. The identifier helps the user to identify the correct model structure, process gain and other model parameters. Process characteristics, such as open-loop stability, phase behavior, are also provided.;An Adaptive Control Design Consultant in the Expert System Environment (ESE) on IBM 4381 was developed to help a non-expert user find the correct controller to use. Based on the information provided by the Model Order Identifier and other a-priori knowledge provided by the user, the Adaptive Control Design Consultant will recommend the most suitable controller. If the user does not supply the minimum information required, recommendations will be made to direct the user to search for the information needed.;Two adaptive algorithms, Self-Tuning PID and Adaptive minimum variance control, were modified to improve the performance. A Direct Adaptive Pole Placement Controller was developed to control deterministic, stable or unstable, minimum or nonminimum phase processes. This algorithm simplifies the structure of the control strategy by directly identifying the control parameters, saving as much as 90% of computation time required by other indirect strategies. The performance of these algorithms was demonstrated on simulated processes, the proposed boiler model and a pilot scale shell-and-tube heat exchanger. |
