R.L. Tousain

Design strategy for iterative learning control based on optimal control

This paper deals with the analysis and synthesis of iterative learning control (ILC) systems using a lifted representation of the plant. In this lifted representation the system dynamics are described by a static map whereas the learning dynamics are described by a difference equation. The properties of the lifted system and in particular the role of nonminimum phase zeros and system delays are investigated. Based on the internal model principle a general, integrating update law is suggested. Next, a new multiobjective design method is proposed for the design of the learning gain, based on optimal control theory. The convergence speed is optimized subject to a bound on the closed loop variance due to stochastic initial conditions, process disturbances and measurement noise. An efficient tailor-made solution to the design problem is presented, making optimal use of the specific and nice structure of the lifted system ILC representation. The potential of the design method is demonstrated on a realistic example.

SSQP for the solution of large-scale dynamic-economic optimization problems

Abstract Dynamic optimization of industrial processes modeled by complex differential-algebraic equation systems is still a major challenge from an algorithmic point of view. We consider a special class of these problems, in which the objective is economic. The severe nonlinearity of this objective function relative to the smooth process behavior motivates the development of a new optimization algorithm called Successive Sequential Quadratic Programming (SSQP), which manages to drastically reduce the number of model integrations required to reach an optimum operating trajectory. The algorithm solved a 3800-variable (100 states) optimal grade transition problem for a HDPE reactor in less than 25 CPU minutes on a personal computer.

Suppressing non-periodically repeating disturbances in mechanical servo systems

Non-periodically repeating (NPR) disturbances are fixed-shape disturbances that occur randomly in time. We can provide a control system with the capability to suppress this type of disturbance by adding in parallel to the input of the nominal feedback controller a learning look-up-table based feedforward controller that is activated using an NPR-disturbance detector.

Closed loop controllability analysis of process designs:Application to distillation column design

Publisher Summary This chapter discusses a new approach toward the integration of process design and control. The approach compares alternative process designs based on the optimal closed-loop performance in the presence of stochastic disturbances. The most important contribution is that a clear relation is established between the degrees of freedom in the process design and a closed-loop performance measure. The approach is illustrated with a case study of a distillation column. The dynamic model of the new process is available in the design stage. The behavior of the plant in or close to its steady state operating points is described using linearized models. The general idea of all design methods that take process control into consideration is that the plant should be designed such that, using some kind of controller, some kind of operating performance can be achieved or optimized. The most important features of this approach are that it can deal with stochastic disturbances and that the analysis is based on the optimal closed-loop behavior of the system. An inherent limitation of the method is that it can only deal with linear models; however, the behavior of process systems around an operating point can often be described accurately with a linear model.

Closed-loop controllability analysis of process designs: Application to distillation column design

Abstract This paper presents a new approach towards the integration of process design and control. The approach compares alternative process designs based on the optimal closed-loop performance in the presence of stochastic disturbances. The most important contribution is that a clear relation is established between the degrees of freedom in the process design and a closed-loop performance measure. The approach is illustrated with a case-study of a distillation column.

A MILP approach to the optimization of the operation procedure of a fed-batch emulsification process in a stirred vessel

This work addresses the model-based optimization of the operation procedure of a fed-batch emulsification process in a stirred vessel. The computation of the input trajectories (i.e. the stirrer speed and the oil flow addition rate as function in time), for reaching a certain predefined, terminal, drop size distribution (DSD) in minimum time, is studied. It is explained that general optimization techniques do not give satisfactory results for this optimization problem. It is suggested to approximate the original minimum time optimization problem as a Mixed Integer Linear Program (MILP). The MILP can be solved for its global solution, which is a good solution of the original optimization problem. The feasibility of the approach is illustrated by means of several optimization studies. The optimization results indicate that the operation time can be decreased by applying non-conventional input trajectories.

Efficient dynamic optimization for nonlinear model predictive control-application to a high-density poly-ethylene grade change problem

Nonlinear model predictive control (NMPC) is believed to play an important role in improving the quality and flexibility of the production of many chemical plants. More widespread application can be expected when systematic solutions are found for modeling large-scale nonlinear processes and for efficient solution of the dynamic optimization problems NMPC entails. The control parametrization approach to dynamic optimization solves the dynamic optimization problem as a nonlinear program using e.g. the sequential quadratic program (SQP) in the outer loop optimization problem. In the SQP approach, a reduced space quadratic program is set up based on a quasi-Newton method estimate of the Hessian. We propose, based on an investigation of the structure of the Hessian of the NMPC problem, a different Hessian update procedure: part of the Hessian is calculated explicitly and only the part that relates to the second derivatives of the dynamics is estimated using a Hessian update. The proposed method shows a large improvement in computational efficiency for a semi-large-scale poly-ethylene reactor NMPC problem with 27 states and 6 inputs with 15 parameters each.

An Optimal Control Approach to Preventing Marine Diesel Engine Overloading Aboard Karel Doorman Class Frigates

Abstract The aim of this research is to analyse and find possible solutions for the occurence of diesel engine overloading aboard the Karel Doorman class frigates in high sea states and during acceleration. In this paper an H ∞ control approach is used to include the knowledge on the model, the disturbances and the performance criteria (disturbance suppression and preventing overloading) in the controller design in a systematic way. If only the fuel rack position is used for control, then the effect of wave disturbances on the shaft revolutions cannot be reduced significantly without causing overloading of the engine. The incorporation of a second control degree of freedom, the propeller pitch angle, makes it possible to meet all the control objectives. A constrained, time optimal control approach is taken to solve the servo problem, i.e. acceleration and deceleration of the vessel. The main benefit of this approach is that constraints on the operation of the propulsion plant, such as the overloading criterion, can be taken into account explicitly. The optimal controls derived from the constrained optimization are combined with the tracking controller in a reference model tracking configuration.

Market-Oriented Scheduling, Economic Optimization and Stochastic Constrained Control of Continuous Multi-Grade Chemical Processes

Abstract In this paper an approach for flexible production scheduling for continuous multi-grade chemical processes is proposed. The approach integrates the economics of production and of company-market interaction for single-machine multi-grade continuous processes. The resulting grade transitions are realized using a newly developed closed-loop stochastic MPC framework, that decomposes this task into a deterministic feed forward constrained trajectory optimization task and a stochastic feedback disturbance suppression task. The back-off used in the former is provided by the latter. The approach is demonstrated on a gas phase HDPE manufacturing plant.

Scheduling Continuous Multi-Product Processes in Compliance with Process Dynamics and Changing Market Conditions

Abstract Advanced process control and optimization is believed to be the key in enabling more flexible operation of continuous multi-product plants. A major benefit of increased operating flexibility is to be able to respond faster to changing market conditions. To investigate how this can be realized, we consider the problem of production scheduling in compliance with plant dynamics and changing market conditions. Plant dynamics are included via the introduction of transition tasks, the attributes of which can be calculated using dynamic optimization. A discretized description of company-market interaction is introduced which enables to include most relevant aspects of market behavior.