SCIAD - Self-Learning Industrial Control Systems Through Process Data

The objectives and scope of the project are in R&D on data-based control methodologies based auto-tuning and auto-adaptive approaches for PID and other controlers for linear and non-linear systems; R&D control methodologies based on process data for auto-design and auto-adaptation of controllers for linear and nonlinear systems; R&D on nonlinear control systems applying methods based on neural networks using processa data, and other methods; R&D on linear and non-linear multi-variable control methodologies; R&D of model predictive control (MPC) methodologies, considering knowledge driven and data driven approaches; R&D of MPC methods that are able to auto-tune, auto-adjust, auto-adapt, and learn the system model from observation of system variables; Research on the identification of process models by cognitive algorithms (neural networks, support vector regression (SVR), fuzzy systems, etc); Research of mathematical prediction methods and/for application on the optimisation of the MPC methods; R&D of mathematical optimisation, stability, and robustness of the researched control methodologies; R&D of learning methodologies for the determination of linguistic values of linguistic fuzzy variables, as well as for the determination of the control rules: R&D on evolutionary algorithms and/or genetic algorithms and/or hybrid and/or optimization methodologies and/or unsupervised learning methodologies for learning the linguistic variables and the rules for the knowledge base of a fuzzy controller; R&D on application to simulated processes, real prototype processes, and real industrial processes.