Advanced autonomous model-based operation of industrial process systems (Autoprofit): Technological developments and future perspectives. (2016)

• Record Type: Journal Article
• Authors: Özkan, Leyla; Bombois, Xavier; Ludlage, Jobert H.A.; Rojas, Cristian; Hjalmarsson, Håkan; Modén, Per Erik; Lundh, Michael; Backx, Ton C.P.M.; Van den Hof, Paul M.J.
• Is Part Of: Annual reviews in control. Volume 42(2016)
• Page Start: 126
• Page End: 142
• Publication Date: 2016
• Subjects: Model based operation support system -- Autonomous maintenance -- Performance diagnosis -- Hypothesis testing -- System identification -- Experiment design -- Randomized algorithms -- (Auto)Tuning -- Controller matching -- Extremum seeking; Automatic control -- Periodicals; Periodicals; 629.805
• DOI: 10.1016/j.arcontrol.2016.09.015
• Languages: English
• ISSNs: 1367-5788

Abstract

Abstract: Model-based operation support technology such as Model Predictive Control (MPC) is a proven and accepted technology for multivariable and constrained large scale control problems in process industry. Despite the growing number of successful implementations, the low level of operational efficiency of MPC is an existing problem, specifically the lack of advanced maintenance technology. To this end, within the EU FP 7 program, a project (Autoprofit 1 ) has been executed to advance the level of autonomy and automated maintenance of MPC technology. Taking linear model-based technology as a starting point, in the project a philosophy has been developed for autonomous performance monitoring, diagnosis, experiment design, model adaptation and controller re-tuning, that is driven by economic criteria in each step, working towards an operation support system in which effective maintenance and adaptation of MPC controllers becomes feasible. In this development, challenging research questions have been addressed in the areas of on-line performance monitoring and diagnosis, least costly experiment design, automated adaptation of models, and auto-tuning, and new fundamental techniques have been developed. Although a full fledge and industrially proven (semi-)automated system is not yet realised, parts of the on-line system have been implemented and validated on real life cases provided by the industrial partners, showing that the formulated objectives are within reach.