Robert Haber

Followers

Following

Co-author

Public Views

Interests

Uploads

Papers by Robert Haber

An alternative for PID control: Predictive Functional Control - a tutorial

Download

The Effect of Coincidence Horizon on Predictive Functional Control

Processes, Jan 8, 2015

This paper gives an analysis of the efficacy of PFC strategies. PFC is widely used in industry fo... more This paper gives an analysis of the efficacy of PFC strategies. PFC is widely used in industry for simple loops with constraint handling, as it is very simple and cheap to implement. However, the algorithm has had very little exposure in the mainstream literature. This paper gives some insight into when a PFC approach is expected to be successful and, conversely, when one should deploy with caution.

Download

Predictive functional control based tuning of PI and PID controllers for aperiodic processes

2020 IEEE 16th International Conference on Control & Automation (ICCA)

The PFC (Predictive Functional Control) algorithm is a very effective predictive controller with ... more The PFC (Predictive Functional Control) algorithm is a very effective predictive controller with controller parameters having a physical meaning. The equivalence of PFC to PI and PID controllers with Smith predictor is shown for delayed first-, second- and higher-order process models. PI(D) controllers are tuned based on the equivalent PFC algorithms. Temperature control of a hot air blower pilot plant illustrates the proposed tuning method.

Equivalence of different two-degree-of-freedom control structures

1999 European Control Conference (ECC)

Three different approaches in control design are considered, which correspond to different contro... more Three different approaches in control design are considered, which correspond to different control structures: (1) feedback controller with filtering of the reference signal, (2) two-degree-of-freedom internal model control and (3) generic control scheme. It is shown that all these structures are equivalent if the control aim e.g., the tracking and the disturbance rejection behaviors, respectively, are the same in the three cases. Also inverse unstable and dead time processes are dealt with.

Estimation of the Predictive Equations

Predictive Control in Process Engineering, 2011

Long-range Predictive Control of Nonlinear Systems given by Volterra Series

IFAC Proceedings Volumes, 1988

The unknown nonlinear process is approximated by a Volterra series model, whi ch is a nonparametr... more The unknown nonlinear process is approximated by a Volterra series model, whi ch is a nonparametric description. The parameters of the Volterra series can be appr oximated recursively. Because of the large number of the parameters the usage of scalar convergence coefficients is recommended. The control algori thm i s an extension of Brujin's predictive control algorithm, based on the weighting function series of linear systems, to the Volterra wei ghting function ser ies mode l. The control signal is assumed constant during the control horizon and the calculation is repeated in every control interval. The above strategy results in a very robust algorithm .

Download

Extended Horizon Predictive Control of Non-Linear Systems - Multi-Dimensional Optimisation and Suboptimal Solution

IFAC Proceedings Volumes, 1996

Predictive control strategies have been proved effective and robust in practice. These algorithms... more Predictive control strategies have been proved effective and robust in practice. These algorithms were developed mainly for linear processes. The main idea is to determine the control signal minimising the deviation between a reference signal and the predicted process output in a given prediction horizon. A simplified version is one-stepahead predictive control, where prediction horizon is restricted to one future point. Extended horizon control proposed by Y dstie ( ) is enlengthened one-step-ahead control which lets more time for the settling process than the dead time. Extended horizon predictive control strategy is derived here for a class of nonlinear systems. Some simulation results show the effect of the tuning parameters for the reference signal tracking performance of the nonlinear system.

Download

Robustness Properties of Extended Horizon Nonlinear Predictive Control Strategies Applied for the Hammerstein Model

IFAC Proceedings Volumes, 1997

Abstract Predictive control algorithms are promising also in the case of nonlinear systems. Two v... more Abstract Predictive control algorithms are promising also in the case of nonlinear systems. Two versions of extended horizon predictive control algorithms are given here for the nonlinear simple Hammerstein model. A quadratic cost function is minimized, which considers the quadratic deviation of the reference signal and the output signal predicted in a future point and also the squares of the control increments. The system model is transformed to a predictive incremental form. Two versions of suboptimal extended horizon control algorithms are given with different assumptions for the control signal during the control horizon. Robustness properties of these algorithms are considered in case of plant-model mismatch through some simulation examples.

Robust Design of PID and IMC-Based Controllers in the Time Domain

IFAC Proceedings Volumes, 1997

A third-order proportional process model with parameter uncertainty is controlled by different si... more A third-order proportional process model with parameter uncertainty is controlled by different simple controllers. First the robustness of four different PIDcontrol designs is investigated. Then internal model control (IMC) is designed in the time domain for the exact process model and for its approximating model with a fIrst-order lag and a delay time. Finally, the PID-control of the nominal model is extended by an IMC-like feedback, which makes the control especially robust.

Download

Long-Range Predictive Control of the Parametric Hammerstein Model

IFAC Proceedings Volumes, 1998

Predictive control algorithms are promising also in the case of nonlinear systems. Long-range pre... more Predictive control algorithms are promising also in the case of nonlinear systems. Long-range predictive control algorithms are derived here for the nonlinear Hammerstein model. A quadratic cost function is minimized, which considers the quadratic deviation of the reference signal and the output signal predicted in a future horizon and punishes also the squares of the control increments. The predictive incremental form of the Hammerstein model is used to predict the output signal. Suboptimal versions of the control algorithm are given with different assumptions for the control signal during the control horizon. Some properties of these algorithms are shown through simulation examples. Behaviour of long-range predictive control algorithms is compared with the performance of one-step-ahead predictive control algorithms.

Download

Predictive Control of a Two Tank System

IFAC Proceedings Volumes, 2000

Predictive control algorithms calculate the consecutive values of the control signal minimising a... more Predictive control algorithms calculate the consecutive values of the control signal minimising a quadratic cost function defined in a finite future horizon. Some types of predictive control algorithms are applied for a non linear model of a two tank system. The system is identified in the form of a linearised model, as a nonlinear parametric Hammerstein and as a Volterra model, respectively. Adaptive predictive control algorithms are applied to control the level of the lower tank while the reference working point is changed. Simulation results show the behaviour of the control.

Download

A Phd Course on Predictive Control

IFAC Proceedings Volumes, 2000

The material of an intensive two weeks predictive control course has been worked out. The course ... more The material of an intensive two weeks predictive control course has been worked out. The course was held for PhD students, in February 1999, for the invitation of Process Control Laboratory, Helsinki University of Technology. The theoretical material covered basic concepts of predictive control of linear and nonlinear plants. Every day the 3 hours lectures were followed by 3 hours computer laboratory exercises using software Matlab/Simulink. The students worked out a project and presented it at the end of the course as a part of the exam. A short description of the course is given here. The experiences are summarised. Copyright <&l 2000IFAC

Download

Long-range optimal model and multi-step-ahead prediction identification for predictive control

IFAC Proceedings Volumes, 2003

Download

Reduction of the Optimization Efforts with Predictive Control by Using Non-Equidistant Prediction Points

IFAC Proceedings Volumes, 2003

With predictive control most of the computation time is spent for the simulation of the predicted... more With predictive control most of the computation time is spent for the simulation of the predicted variables and for the optimization if constraints or non linear processes are assumed. In addition to the known blocking technique for the manipulated variable another possibility is calculating the control error not in each sampling point of the prediction horizon but only in some coincidence points. It will be shown that the best choice is to allocate the coincidence points exponentially thus that with small prediction steps more and with increasing prediction steps less coincidence points are considered. As a practical example the multivariable control of a distillation column model illustrates the benefits of the method presented.

Download

Nonlinear process identification and predictive control by multiparameter models

IFAC Proceedings Volumes, 2004

According to the multi-model approach a nonlinear dynamical process is approximated in different ... more According to the multi-model approach a nonlinear dynamical process is approximated in different working points by local valid linear models. The global valid model output is calculated as the weighted sum of the sub-model outputs. The parameters of the Gaussian weighting function can be chosen by optimization. The computation time can be reduced if instead of the model outputs the parameters (for example static gain and time constant) of the local valid models are merged. The global valid nonlinear model can be used e.g., for model based predictive control. The new, multi-parameter method is illustrated by a heat exchanger example.

Download

Predictive On-Off Cost Minimizing Control of a Municipal Waste Water Treatment Plant

IFAC Proceedings Volumes, 2005

With intermittently operated municipal waste water plants the aeration switching sequence should ... more With intermittently operated municipal waste water plants the aeration switching sequence should be optimized under the constraints that the solved nitrogen concentration in effluent has to be kept within its environmental limits. This is a highly nonlinear control problem because of the nonlinear biological process model ASM1 (Activated Sludge Model 1), the on-off state of the manipulated sequence and the concentration constraints. The constrained optimization was transformed to the unconstrained optimization of the predictive on-off control, as the violations of the constraints were converted to their economical costs. Furthermore, the cost of the aeration was considered in the cost function.

Download

An alternative for PID control: Predictive Functional Control - a tutorial

2016 American Control Conference (ACC), 2016

Download

Pole-placement Predictive Functional Control for over-damped systems with real poles

ISA Transactions, 2016

This paper gives new insight and design proposals for Predictive Functional Control (PFC) algorit... more This paper gives new insight and design proposals for Predictive Functional Control (PFC) algorithms. Common practice and indeed a requirement of PFC is to select a coincidence horizon greater than one for high-order systems and for the link between the design parameters and the desired dynamic to be weak. Here the proposal is to use parallel first-order models to form an independent prediction model and show that with these it is possible both to use a coincidence horizon of one and moreover to obtain precisely the desired closed-loop dynamics. It is shown through analysis that the use of a coincidence horizon of one greatly simplifies coding, tuning, constraint handling and implementation. The paper derives the key results for highorder and non-minimum phase processes and also demonstrates the flexibility and potential industrial utility of the proposal.

Download

Mercedes - Interactive Software Package for Identification and Experimental Control of Industrial Plants by a Portable Process Computer Laboratory

Software for Computer Control, 1979

The paper describes the ~mRCEDES program-package su~t8ble for industrial applications. The intera... more

Download

Evolutionary optimization methods for predictive on-off control

2nd IFAC International Conference on Intelligent Control Systems and Signal Processing (2009), 2009

On-off control is still widely used in the industry, especially for discontinuous actuators, e.g.... more On-off control is still widely used in the industry, especially for discontinuous actuators, e.g. compressors, steam turbines or furnace burners. Exact set-point control is not possible by an on-off controller. Lower and upper limits are defined for the control signal which should not be violated. For this predictive gap control quadratic and horizon dependent cost functions are defined. Traditional numerical optimization fails with longer prediction horizons, but a genetic algorithm can be effectively applied even for long horizons.

Download

Robert Haber

Uploads

Papers by Robert Haber

Log In