| This thesis proposes some of the basic components of an adaptive hierarchical multilayered multi-agent supervision and control framework for nonlinear and distributed processes. Various data reconciliation, process understanding and product grade control methods are implemented in a multi-agent framework, enabling flexible, modular and robust supervision of distributed processes. The ultimate objective is to create a dynamic process supervision environment that can adapt itself and its actions in response to variations in operating conditions, set point changes, faults, disturbances and configuration changes by using historical data collected from the process.;One novelty of the research is the development of self-organizing decentralized multi-agent structures for the supervision tasks by designing auction-based mechanisms to give adaptation and self-organization capabilities, analyzing the performance of these mechanisms on the process and ultimately creating inter-agent rules that drive the system towards desired behavior. Another novelty of the supervision system is the automation of knowledge extraction from the process data using different methods, enabling the use of alternative and/or complementary methods.;The framework and software suite developed performs (M)odeling, (A)nalysis, (D)iagnosis and (C)ontrol using (A)gent (B)ased (S)ystems (MADCABS). A number of data reconciliation and preprocessing agents operate locally and globally for outlier detection, noise reduction and estimation. Process understanding modules analyze process data collected from normal operation, identify correlations between different sections, detect process changes and faults, and identify abnormal process behavior locally, regionally and globally using multivariate statistical techniques. Control modules of the framework contain agents distributed among different sections of the system. The grade transition control modules use auctioning techniques to adjust their local objectives that are parallel with the global objective which could be to reject a disturbance or to change the different product grade.;The proposed framework is tested using case studies on two simulated processes and one industrial process. The first simulator represents an interconnected reactor network hosting multiple autocatalytic species. The second simulator represents a reversible addition-fragmentation chain transfer (RAFT) polymerization system in continuous stirred tank reactor (CSTR) networks for producing. The third case study uses data from is an industrial extrusion process for automotive industry products used by Corning Incorporated. |
