Showing papers on "Smith predictor published in 1989"

Deadtime compensation for nonlinear processes

Abstract: Many industrially important processes feature both nonlinear system dynamics and a process deadtime. Powerful deadtime compensation methods, such as the Smith predictor, are available for linear systems represented by transfer functions. A Smith predictor structure in state space for linear systems is presented first and then directly extended to nonlinear systems. When combined with input /output linearizing state feedback, this Smith-like predictor makes a nonlinear system with deadtime behave like a linear system with deadtime. The control structure is completed by adding an external linear controller, which provides integral action and compensates for the deadtime in the input/output linear system, and an open-loop state observer. Conditions for robust stability with respect to errors in the deadtime and more general linear unstructured multiplicative uncertainties are given. Computer simulations for an example system demonstrate the high controller performance that can be obtained using the proposed method.

Extension and partial optimization of a modified Smith predictor and controller for unstable processes with time delay

Abstract: Extension and partial parameter optimization are applied to a modified Smith predictor and controller, designed by de Paor( 1985) for a class of unstable processes having transfer functions of the form k1A exp(—sT)/B, where A is a monic, Hurwitz polynomial of degree (n — l) and the nth degree monic polynomial B has a single right half-plane root at s=λ It is found that for l = 1,2,3 asymptotic stability can be achieved for λ T> 1, which more than doubles the best result obtained earlier. For a specific process, responses are compared with those obtained using the optimum stability PI controller of de Paor and O'Malley (1989). The advantages of the optimized, modified predictor in terms of tightness of control and faster settling time are highlighted

Dynamics and control of quasirational systems

Abstract: Systems having transfer functions of the form where P1(s), P2(s) and Q(s) are polynomials, are called quasirational distributed systems (QRDS). They are encountered in processes modeled by hyperbolic partial differential equations. QRDS can have an infinity of right half-plane zeros which causes large phase lags and can result in poor performance of the closed-loop system with PID controllers. Theory on the asymptotic location of zeros of quasipolynomials is used to predict the nonminimum phase characteristics of QRDS and formulas are presented for factoring QRDS models into minimum and non-minimum phase elements. A generalized Smith predictor controller design procedure for QRDS, based on this factorization, is derived. It uses pole placement to obtain a controller parameterization that introduces free poles which are selected to satisfy robustness specifications. The use of pole placement allows for the design of robust control systems in a transparent manner. Controller selection is generally better, simpler and more direct with this procedure than searching for optimal PID controller settings.

Control of processes with noise and time delays

Abstract: A stable Kalman filter predictor (KFP) is developed which generates minimum variance estimates of the future outputs {y(t + i | t), i = 1, … d} of stochastic, single-input/single-output processes with time delay, d. The predicted outputs are used for time delay compensation and in the design of a predictive feedback controller. An innovation model analysis is used to convert the state space formulation to transfer function form and to show the relationship between the KFP, the Smith predictor, and the internal model controller. A modified KFP includes a disturbance model, and eliminates offset due to deterministic disturbances (e.g., steps) and modeling errors. Simulation results show that the modified KFP also predicts the disturbances and gives significantly better performance than the Smith predictor, particularly in the presence of process and measurement noise.

A Closed Loop Parameter/Deadtime Estimation Technique

Abstract: Processes with large deadtime to time constant ratios have traditionally proven difficult to control effectively. In practice, PID controllers (tuned to assure robustness at the expense of performance) are generally employed for the control of these processes. Conceptually, better performance is achievable through deadtime compensation using a Smith Predictor or Model Based Controller. Unfortunately, these techniques are rarely used in practice due, in-part, to unfamiliarity of technicians and operators with the tuning of these controllers. Recent work in self-tuning control has had little impact on the control of large deadtime processes. This can be attributed to reliance of most self-tuning controllers on the PID algorithm and the absence of techniques for the identification of process deadtime. This paper describes a model based control algorithm that directly incorporates deadtime compensation, and a novel closed loop process identification technique capable of identifying process deadtime directly. This combination provides an effective system for the control of large deadtime process. The resulting controller is easy to implement and commission. Its auto-tuning feature provides a simple mechanism for retuning the controller on a periodic basis. Extensive simulation results and preliminary field test results are presented in addition to a brief description of the controller and auto-tuner.

Generalised input-output compensators for multivariable processes with time-delays

Abstract: Generalized input-output compensation of MIMO (multiinput, multioutput) systems with time delays is considered. It is shown that any of the three basic features of the Smith predictor can be extended to MIMO systems in the input-output configuration. A selection scheme that does not require the inverse of the plant transfer matrix is devised for plant delays in reducing diagonal delay elements in the decoupling response bound. It is believed that the scheme simplifies implementation of delay selection procedures. >

Neural network architecture for slow nonlinear time-delayed processes

Abstract: An experimental multilayered neural network architecture for the control of a given plant is presented. It will be implemented using multi-microcomputing techniques, based on synchronized parallel processing, using associative shared memory message passing, and the retrieval of knowledge of plant and controller parameters. This new concept of control is used to solve the difficulties present in the digital control of slow nonlinear time-delayed processes, especially the case of systems with long dead times. This paper explains how the Smith predictor algorithm was modified in order to adapt it to a multilayered neural network architecture.