The task of Control Configuration Selection (CCS) in multivariable processes addresses the grouping of each controlled variable with a set of actuators which will be used for calculating its control action. The CCS problem is related to model reduction, since it is resolved by choosing a reduced model which reflects the most important interconnections between actuators and controlled variables. This selection can be performed by considering controllability and observability aspects of the process.
Previous results at LTU
Whilst the traditional CCS methods require the availability of complete models of the plant, we have developed at Luleå University of Technology new data-driven CCS methods which can reveal the significance of the process interconnections from tailored process experiments with the use of non-parametric modelling techniques. Being a current limitation of the data-driven CCS methods is that they require the availability of the process to perform open-loop experiments.
Master's Thesis Work
The challenge addressed in this Master’s Thesis project is to extend the current theory on data-driven CCS to be applicable on systems under closed-loop control. This is important for practical applications, since plant experiments have to be performed under the presence of controllers which at the same time ensure the stability of the process during experiments and avoid the degradation of production goals which would otherwise have a large economic impact. The results of this Master’s Thesis project will be tested on simulations using available models from real-life processes belonging to industry partners in the pulp & paper and/or mining industries. It is targeted that the results in this thesis should be disseminated in a conference or journal publication.