Design and Certification of Industrial Predictive Controllers
De Keyser, Robin
Wolfson College Cambridge
Doctor of Philosophy (PhD)
MetadataShow full item record
Dutta, A. (2014). Design and Certification of Industrial Predictive Controllers (Doctoral thesis). https://doi.org/10.17863/CAM.15887
Three decades have passed since milestone publications by several industrial and academic researchers spawned a flurry of research and commercial, industrial activities on model predictive control (MPC). The improvement in efficiency of the on-line optimization part of MPC led to its adoption in mechanical and mechatronic systems from process control and petrochemical applications. However, the massive strides made by the academic community in guaranteeing stability through state-space MPC have not always been directly applicable in an industrial setting. This thesis is concerned with design and a posteriori certification of feasibility/stability of input-output MPC controllers for industrial applications without terminal conditions (i.e. terminal penalty, terminal constraint, terminal control). MPC controllers which differ in their modelling and prediction method are categorized into three major groups, and a general equivalence between these forms is established. Then an overview on robust set invariance is given as it plays a fundamental role in our analysis of the constrained control systems. These tools are used to give new tuning guidelines as well as a posteriori tests for guaranteeing feasibility of the suboptimal or optimal predictive control law without terminal conditions, which is fundamental towards stability of the closed loop. Next, penalty adaptation is used as a systematic procedure to derive asymptotic stability without any terminal conditions and without using set invariance or Lyapunov arguments. This analysis however is restricted to repetitive systems with input constraints. Then, predictive control without terminal conditions is considered for nonlinear and distributed systems. The invariance tools are extended to switching nonlinear systems, a proof of convergence is given for the iterative nonlinear MPC (NMPC), and a guarantee on overall cost decrease is developed for distributed NMPC, all without terminal conditions. Reference generation and parameter adaptation are shown to be effective mechanisms for NMPC and distributed NMPC (DNMPC) under changing environmental conditions. This is demonstrated on two benchmark test-cases i.e. the wet-clutch and hydrostatic drivetrain, respectively. Terminal conditions in essence are difficult to compute, may compromise performance and are not used in the industry. The main contribution of the thesis is a systematic development and analysis of MPC without terminal conditions for linear, nonlinear and distributed systems.
This work was supported within the framework of the LeCoPro project (grant nr. 80032) of the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen).
This record's DOI: https://doi.org/10.17863/CAM.15887
Recommended or similar items
The following licence files are associated with this item: