Control Of Multivariable Plants Research Articles

Design and implementation of decoupled periodic control scheme for a laboratory-based quadruple-tank process

It is well known that linear time-invariant controllers fail to provide desired robustness margins (e.g. gain margin, phase margin) for plants with non-minimum phase zeros. Attempts have been made in literature to alleviate this problem using high-frequency periodic controllers. But because of high frequency in nature, real-time implementation of these controllers is very challenging. In fact, no practical applications of such controllers for multivariable plants have been reported in literature till date. This article considers a laboratory-based, two-input–two-output, quadruple-tank process with a non-minimum phase zero for real-time implementation of the above periodic controller. To design the controller, first, a minimal pre-compensator is used to decouple the plant in open loop. Then the resulting single-input–single-output units are compensated using periodic controllers. It is shown through simulations and real-time experiments that owing to arbitrary loop-zero placement capability of periodic controllers, the above decoupled periodic control scheme provides much improved robustness against multi-channel output gain variations as compared to its linear time-invariant counterpart. It is also shown that in spite of this improved robustness, the nominal performances such as tracking and disturbance attenuation remain almost the same. A comparison with [Formula: see text]-linear time-invariant controllers is also carried out to show superiority of the proposed scheme.

Reconfigurable controller design for linear multivariable plants

Decentralised control is widely used for the control of multivariable plants. Prior to the design of the decentralised controllers, input-output pairing is an important step in the design procedure. In the face of unknown, uncertain or time varying plant parameters, the input-output selection may endure fundamental changes, which will severely degrade the decentralised controller performance. This paper proposes a reconfigurable structure for the design of the decentralised controller based on the adaptive control strategies. Simulation results are provided to show the effectiveness of the proposed methodology.

Adaptive Decentralized Control of Multivariable Plants with Scalar Input and Output and Nonminimal Realization of the Reference Model

The problem of designing an adaptive control system with reference models of local subsystems was solved for the case where only the controlled variable and the scalar control action are measurable. It was proved to be operable if the controlled plant is subjected to the action of unmeasurable bounded perturbations. At that, only the measurable variables of the local subsystems are used to generate the control actions and run the adjustment algorithms, that is, control is completely decentralized.

Adaptive Decentralized Delayed Control of Multivariable Plants

Consideration was given to the adaptive delayed-control system with the reference models of local subsystems. The adaptive predictor forecasts the controlled value, and the proposed control algorithms provide zero errors of tracking the local reference models in the a priori uncertainty of the parameters of subsystems and interconnections. Only the measurable variables of the local subsystems are used in generation of the control actions and in the adjustment algorithms, that is, the control is fully decentralized.

Model reference adaptive control of multivariable plants with delays

The design and complete stability analysis for the model reference adaptive control (MRAC) of multivariable plants with input and/or output delays are presented in this paper. A modified estimation error, which enables us to solve the non-commutativity problem and the stability issue in the multivariable case, is introduced in the adaptive law for updating the controller parameters. A generalized controller structure, which achieves stable model-plant transfer matrix matching in the presence of delays, is developed for the multivariable MRAC scheme. Based on a unified framework it is shown that the proposed adaptive control scheme guarantees the global stability and asymptotic zero tracking error for the closed-loop system.

Interaction Indices for the Decentralized Control of Unknown Multivarible Systems

A set of steady-state interaction indices is defined and analyzed in this paper. These indices are found to play an important role in characterizing steady-state control loop interaction in the decentralized control of multi variable plants. The indices are simple to use and do not require that a mathematical model of the system to be controlled should be available. The only assumptions made re the plant are that: (i) it can be described by a linear time-invariant model and (ii) it is open loop asymptotically stable. Under these assumptions, it is shown that the indices may be determined experimentally by applying a number of simple steady-state input/output experiments on the plant. The indices can be used to indicate when “undesirable interaction effects” will arise in controlling an unknown multivariable system using decentralized control. A comparison with the relative gain array [10] method, shows that the indices are more complete and accurate to apply. A number of examples are included to illustrate the type of results which may be obtained.

Synthesis of controllers for uncertain multivariable plants for prescribed time-domain tolerances†

Abstract The paper considers the synthesis of controllers for the feedback control of multi-variable plants with large parameter ignorance. Here, the output response of each loop is required to satisfy some a priori specified time domain tolerance bounds irrespective of the parameter variations of the plant and the interaction carried through other loops. By an interplay between the Nyquist array and Bode array, it becomes possible to obtain a design that not only satisfies the required specifications but also offers some advantages over other design techniques for the same class of problems.

Design of error-actuated controllers for multivariable plants with unknown dynamics and immeasurable outputs

It is shown that, for a large class of multivariable plants, the design of error-actuated controllers which ensure that excellent set-point tracking occurs simultaneously with disturbance rejection can be readily effected even though the detailed dynamical properties of the processes involved are unknown and the controlled outputs are unmeasurable. The theory is illustrated by the design of analogue and digital controllers for a modern gas-turbine engine.

Linear programming design of multivariable feedback controllers for the NERVA nuclear engine

The application of Linear Programming (LP) to the design of state variable feedback controllers for multivariable plants with known deterministic inputs is described. The design problem is solved in two steps. The first step consists of an open-loop optimization in which LP is used to determine the plant input vector profile which optimizes the plant response to a specified demand. The second step, also implemented by LP, produces the controller parameters (a matrix of constant gains) and consists of an optimum fitting, through the feedback controller, of the input and feedback error vector profiles determined in the first step. The performance criterion used is the minimum Integral Weighted Absolute Value (IWAV). The technique is illustrated with an application to the NERVA nuclear rocket engine controller design.

Analysis of the Accuracy of Parametric Adjustment of a System for Adaptive Control of Multivariable Plants in Modified Sensitivity Models

Analysis of the Accuracy of Parametric Adjustment of a System for Adaptive Control of Multivariable Plants in Modified Sensitivity Models