Hsiao-Ping Huang

A direct method for multi-loop PI/PID controller design

Difficulties caused by the interactions are always encountered in the design of multi-loop control systems for MIMO processes. To overcome the difficulties, a multi-loop system is decomposed into a number of equivalent single loops for design. For each equivalent single loop, an effective open-loop process (EOP) is formulated without prior knowledge of controller dynamics in other loops, and, hence, controller can be designed directly and independently. Based on the derived EOPs, a model-based method aims at having reasonable gain margins (e.g. 52) and phase margins (e.g. � 60 � ) are presented to derive multi-loop PI/PID controllers. This proposed method is formulated in details for the EOPs of 2-loop systems. Extension to higher dimensional systems needs further simplification and is illustrated with formulation for 3-loop systems. Simulation results show that this presented method is effective for square MIMO processes, especially, for low dimensional ones. # 2003 Elsevier Ltd. All rights reserved.

A sliding observer for nonlinear process control

Sliding observers are considered as nonlinear state estimators with good robustness to bounded modeling errors. In this paper we have developed sliding observers for process control. The observer is hence designed so as to possess invariant dynamic modes which can be assigned independently to achieve the desired performance. Convergence of the estimating algorithm is formulated by using Lyapunov stability theorems. Conditions for robustness to modeling errors are derived by analyzing the norms of estimation errors. For process control, servo-tracking and disturbance rejection for chemical reactors have been discussed by making use of this sliding observer. Simulation examples to demonstrate the construction and performance of this proposed sliding observer for chemical process control are also presented.

A simple method for tuning cascade control systems

Abstract A simple method for tuning controllers in a cascade system is presented, In this method, all the relations that facilitate the tuning procedures are well prepared in terms of figures or simple equations. Using these figures and equations, the controller tuning for different configurations of cascade systems becomes easy and straightforward when process models are available. On the other hand, when process models are not available, a simple method that uses one single run of step input experiment to develop such models is proposed. Based on these developments in the controller tuning and process models, an autotuning system that uses relay feedback is presented. Unlike the existing autotuning systems, this proposed system conducts identification and controller tuning in a decoupled manner. As a result, no excessive trial- and-error efforts for modeling and tuning are required. Simulation results show the potential usage of such a method, It is interested to see that the resulting systems have almo...

Deadtime compensation for nonlinear processes with disturbances

Abstract The control problem of time delay non-linear systems that are perturbed by disturbances is discussed. By coordinate transformations and feedback and prediction of the transformed states, we first linearize the perturbed non-linear systems into controllable quasi-linear systems with disturbances. Then, we can apply the well-developed linear control theory to stabilize the transformed systems. Thus, stable quasi-linear systems with time delay can be obtained. Furthermore, we may implement powerful deadtime compensation methods to study the performance of the proposed dynamic compensators, a Smith predictor and a new modified Smith predictor, for disturbance rejection. Finally, a typical non-linear chemical process, a continuous stirred tank reactor, is used as an example system to demonstrate the effectiveness of these deadtime compensators

THE ADAPTIVE AUTOREGRESSIVE MODELS FOR THE SYSTEM DYNAMICS AND PREDICTION OF BLAST FURNACE

The blast furnace can be viewed as a time-varying stochastic system. The Adaptive Autoregressive (AAR) models are proposed to characterize such systems. AAR identification is a method of successive parameter estimation by using recursive formulas with variable forgetting factors to closely track time-varying parameters. A simple example is presented to illustrate the parameter tracking capability of the AAR models. Based on the prediction errors, the AAR models of blast furnace are compared with the conventional time series models. Through this comparison, the AAR models prove to be superior to the other time series models, since the latter are suitable only for time-invariant systems. It is concluded that during smooth operation, just the AAR scalar model is required for forecasting as operational guide. When the operation is uneven, the AAR vector model provides the better results. To control the performance of this process the data should be sampled under uneven operating condition, where the AAR vecto...

Assessment of controller performance: a relay feedback approach

Assessment of control loop performance is one of the important tasks to be carried out in a plant, regardless of the control strategy. The present work utilizes the shape information from single relay feedback test to assess the performance of a feedback system. The process considered is a typical first-order plus dead time process and the controller used is PI type of controller. An ideal relay is introduced in the feedback loop before the controller. The shape of the relay output characterizes the performance of the controller. The mismatch in the integral time can also be observed from the shape. Based on the shape information, analytical expressions for relay feedback responses are derived and straightforward procedures are evolved to assess the performance of the controller. The proposed scheme assesses the controller performance and computes the new tuning parameters, in case retuning of the controller is necessary. The robustness of the method is also tested for second-order plus dead time as well as for higher order systems. More importantly the present approach employs only one relay test for (1) assessment and (2) retuning of the controller. It provides a reliable way that is compatible even to a non-expert operator to assess the controller performance.

Recursive parameter estimation of transfer function matrix models via Simpson's integrating rules

Recursive formulae for repeated integration of a continuous-time function with uniformly sampled data using Simpsons 1/3 and 3/8 integrating rules are derived. Combined with the recursive algorithm of the least-squares solution, a method for recursive parameter estimation of transfer function matrix models in multiple-input-multiple-output systems is proposed. It is demonstrated that the use of the popular integrating rules for parameter estimation can be as effective as sophisticated methods that use orthogonal functions and the associated operational properties reported in the literature. The proposed algorithm is suitable for on-line applications and computer programming. Three numerical examples are included to illustrate the applicability of the proposed method.

Inverse-based Design for a Modified PID Controller

A modified PID (m-PID) controller design using low order dynamic models is presented. In contrast to the one of Haggland and Astrom (1985), the controller uses model-based parameters. Synthesis of such a controller is by way of an inverse-based approach. By this approach, a pure lead is introduced into a l=p to make the system equivalent to a modified Smith predictor (MSP) control system. Tuning rules for this m-PID controller are then derived from this equivalent system. For robustness, gain and phase margins are analyzed. It is found that this proposed m-PID control system has good responses to both set-point and load changes. Tuning rules derived from reduction of this m-PID controller for conventional PID control have also been presented.

Grade transition using dynamic neural networks for an industrial high-pressure ethylene―vinyl acetate (EVA) copolymerization process

In this paper, estimating of melt index (MI) in an industrial ethylene and vinyl acetate (EVA) copolymerization process will be studied. Three products with their melt indexes ranging from 2.49 to 167.21 are dynamically estimated by an artificial neural networks (ANN) model based on available plant measurements. With this dynamic estimator, a simple MI controller with only proportional-integral mode can be established for the purpose of grade transition by suitably adjusting the chain modifier feed rate. Simulation results demonstrate that significant reduction in the grade transition time can be gained in comparison with the base strategy by step-changing of the operating recipe. A simple model updating algorithm is also proposed for the adjustment of the predicted MI using infrequent lab measurement to handle plant-model mismatches.

A mathematical model for the catalyst deactivation in a commercial residue hydrodesulfurization reactor system

A mathematical model that describes catalyst deactivation in a commercial Residue Hydrodesulfurization (RDS) system is proposed. In this model, coking, metal deposition and the decrease of effective diffusivities are all considered, so that it can predict the dynamics of catalyst activity during a long run of an RDS system. The RDS system under consideration here is a complex trickle bed reactor system with four adiabatic reactors in series, between which the oil is quenched by hydrogen to prevent the catalysts in the next reactor from being damaged by high temperature. This article shows that by incorporating a catalyst deactivation model into the overall reactor system model, the predicted catalyst aging curve assumes the “S” shape, characteristic of real RDS system. The model prediction are compared with data from an operating refinery and the agreement is pleasing. KEYWORDS Residue Hydrodesulfurization(RDS) Apparent activity Catalyst deactivation curve Pore radius.