Search

Determining the optimal operating mode (batch or continuous-flow) for chemical manufacturing remains a pressing open problem. Focusing on reactors, an accurate determination must account for fundam...

The relay (on–off) controller can stabilize wide ranges of processes including open-loop stable, integrating, and unstable processes, producing sustained oscillations. For improved proportional-integral-derivative controller tunings, methods to find process models with mixed closed-loop tests of relay feedback and proportional-derivative (PD) controllers are proposed. For unknown processes with arbitrary initial states, relay feedback tests are first applied and, after cyclic steady states are obtained, PD controllers or other relay feedback tests with set point changes are followed. This full closed-loop operation is desirable for integrating and unstable processes and will be useful even for stable processes when processes are far from their desirable operating points. Refined methods to find exact frequency responses of processes from initial and final cyclic steady states are derived. Whole relay feedback responses need not be saved. Several integrals at the relay switching times are used without iterative tests or computations.

Step and pulse responses have long been used to identify process dynamics and design control systems such as PID (proportional-integral-derivative) and model predictive control. They are simple and...

Over the past two decades, the pharmaceutical and specialty chemical industries have led a considerable effort toward transitioning from existing batch to continuous flow processes. The transition ...

This paper applies the concept of dynamic intensification (defined as changes to the dynamics, operation strategy, and/or control of a process that lead to a substantially more efficient processing path) to binary distillation columns. The resulting strategy consists of manufacturing a target product as a blend of two auxiliary products, both having lower energy demands than a reference value, which corresponds to producing the target product(s) in a column operating at steady state. A discussion of the appropriate control structures and switching strategies between the two auxiliary products is provided. An extensive case study concerning the separation of a methanol–1-propanol mixture was carried out, demonstrating that energy savings in the order of 1.4% are possible with no disruption in product quality or production rate.

Most process intensification research and its applications have focused on process and equipment design modifications. In this paper, we present an overview of existing developments and opportunities in dynamic process intensification (DI), which comprises changes to dynamics, operational and control of a chemical process, that result in substantial efficiency improvements. After reviewing a series of examples, we identify several fundamental principles underlying DI, including equipment design and operational approaches. We conclude with a number of challenges and opportunities related to DI system design, operation and control, and fault detection and isolation.

The goal of quantifying the monetary value of process control has been a target of much research since the inception of the field, and methods have been developed for quantifying the value of control in the case of predominantly steady-state processes. However, there has been no attempt to quantify the monetary value of control for predominantly transient processes. In this work, we utilize the general framework of integrated scheduling and control to develop novel performance functions that enable the quantification of the monetary value of control from a scheduling perspective for a predominantly transient process. Specifically, we posit that the transition time between one product and the next in a production sequence can be used as a performance metric over which the value of control can be quantified. We demonstrate the utility of the developed performance functions using a case study of the scheduling of a multi-product CSTR.

Three dimensional data structures such as batch process data or infra-red spectral measurements usually contain inconsistent trajectories of various durations and quality. In the case of batch process data, most modeling methods require the data from all batches to be of same duration. For spectral data, peaks might be shifted from one sample to another due to unaccounted sources of variation. These inconsistencies are usually resolved through trajectory alignment (or synchronization) methods. In this paper, we first review the deficiencies of existing approaches. Next, a Constrained selective Derivative Dynamic Time Warping (CsDTW) method is proposed to perform automatic alignment of trajectories. Different from conventional methods, CsDTW preserves key features that characterizes the batch and only apply warping to regions of least impact to trajectory characterization. The proposed warping technique is applied to both industrial and simulated datasets to demonstrate its effectiveness.

The first-order plus time delay (FOPTD) model-based method is a standard approach to tune proportional–integral (PI) controllers in plants. The FOPTD model can be obtained easily from step responses. However, because of their structural limitations, FOPTD models suffer from difficulties in approximating step responses for some processes including processes with overshoot, resulting in PI controllers with unacceptable performance. To remove these drawbacks, models combining two FOPTD models that can be obtained easily from step responses are proposed to tune PI controllers. Several simulations and experimental examples are given, illustrating the improved performance of the proposed method.

Methods based on the first-order plus time delay (FOPTD) model are very popular for tuning proportional-integral (PI) controllers. The FOPTD model-based methods are simple and their utility has been proved with many successful applications to a wide range of processes in practice. However, even for some overdamped processes where the FOPTD model seems to be applied successfully, these empirical FOPTD model-based methods can fail to provide stable tuning results. To remove these drawbacks, a PI controller tuning method based on half-order plus time delay (HOPTD) model is proposed. Because FOPTD model-based methods can be applied to higher order processes, the proposed HOPTD model-based method can be applied to higher order processes as well. It does not require any additional process information compared to the FOPTD model-based method and hence can be used for overdamped processes in practice, complementing the traditional FOPTD model-based methods. © 2016 American Institute of Chemical Engineers AIChE J, 63: 601–609, 2017

Plant-wide oscillations generated in a single unit can negatively affect the overall control performance of the process; thus, it is necessary to detect them and diagnose their root cause. However, the interference of noise and the need for oscillation propagation routes pose more challenges for process engineers. In this paper, the concept spectral transfer entropy is proposed and its connection to the spectral Granger causality is derived. Moreover, an information transfer method incorporating spectral envelope algorithm and spectral transfer entropy is applied to provide new diagnostic guidance, whose feasibility and effectiveness have been demonstrated by both simulated and industrial case studies. Compared with current methods, the new procedure enjoys the following advantages: (a) performing oscillation detection and diagnosis within a targeted frequency range and mitigating the effects of measurement noise outside the bandwidth; (b) provides an nominal causal map reflecting the oscillation propagat...

First order plus time delay (FOPTD) models obtained from the cyclic steady state of the relay feedback oscillations can suffer from poor performances for some processes such as first and second order processes with fast parasitic dynamics. Poor FOPTD models are due to high frequency oscillations that activate parasitic dynamics. Dynamic elements such as hysteresis and integrator added in the feedback loop can reduce the oscillation frequencies but they increase the experimental times. Here, to relieve these problems without increasing the experimental time and computational complexity, relay feedback autotuning methods that apply the method of weighted moments to relay feedback transients are proposed. The proposed methods use the same relay feedback tests as the conventional relay feedback methods and are simple to use computationally, providing conservative PI controllers for all the test batch processes and the above first and second order processes with fast parasitic dynamics.

The question of how to assess the monetary value of process control has been of significant interest from both the academic and industrial points of view since the inception of the field. While metrics exist for addressing this problem in the case of predominantly steady-state processes, no work has been reported for predominantly transient processes. In this paper, we develop two performance metrics that allow the evaluation of the monetary value of improving process control for predominantly transient processes. We demonstrate the utility of the metrics in assessing (1) whether improvement of process control would yield any worthwhile monetary benefits, and (2) for control structures that carry out multiple production transitions within a production schedule, the specific areas where improvement of process control would yield the highest economic benefits.

Relay feedback method is one of the standard methods for PID controller autotuning in field. Relay feedback oscillations provide the ultimate gain and period of processes to design PID controllers with Ziegler-Nichols type tuning rules. Since the original relay feedback method uses only two process data of ultimate gain and period, its performances can be limited. For wider applications of the relay feedback autotuning method, methods using additional process data such as the steady state gain have been available. They often use the first order plus time delay (FOPTD) models and the FOPTD model-based tuning rules. Here it is shown that FOPTD model-based methods can provide closed-loop systems with very oscillatory responses for some processes. To relieve such drawbacks, the critically damped second order plus time delay (C2PTD) model that uses the same process data of previous FOPTD model-based methods is proposed. When three process data are extracted from the relay feedback oscillation, the C2PTD model-based method instead of the usual FOPTD model-based methods is shown to be better for wider applications of the relay feedback autotuning.

A nonlinear process with input multiplicity has two or more input values for a given output at the steady state, and the process steady state gain changes its sign as the operating point changes. A control system with integral action will be unstable when both signs of the process gain and the controller integral gain are different, and its stability region will be limited to the boundary where the process steady state gain is zero. Unlike processes with output multiplicities, feedback controllers cannot be used to correct the sign changes of process gain. To remove such stability limitation, a simple control system with parallel compensator is proposed. The parallel compensator can be easily designed based on the process steady state gain information and tuned in the field. Using the two time scale method, the stability of proposed control systems for processes with input multiplicities can be checked.

Over the years, there has been a consistent increase in the amount of data collected by systems and processes in many different industries and fields. Simultaneously, there is a growing push towards revealing and exploiting of the information contained therein. The chemical processes industry is one such field, with high volume and high-dimensional time series data. In this paper, we present a unified overview of the application of recently-developed data visualization concepts to fault detection in the chemical industry. We consider three common types of processes and compare visualization-based fault detection performance to methods used currently.

This paper describes nonlinear methods in model building, dynamic data reconciliation, and dynamic optimization that are inspired by researchers and motivated by industrial applications. A new formulation of the l1-norm objective with a dead-band for estimation and control is presented. The dead-band in the objective is desirable for noise rejection, minimizing unnecessary parameter adjustments and movement of manipulated variables. As a motivating example, a small and well-known nonlinear multivariable level control problem is detailed that has a number of common characteristics to larger controllers seen in practice. The methods are also demonstrated on larger problems to reveal algorithmic scaling with sparse methods. The implementation details reveal capabilities of employing nonlinear methods in dynamic applications with example code in both Matlab and Python programming languages.

Soft sensors (or inferential sensors) have been demonstrated to be an effective solution for monitoring quality performance and control applications in the chemical industry. One of the key issues during the development of soft sensor models is the selection of relevant variables from a large array of measurements. A subset of variables that are selected based on first principles and statistical correlations eases the model development process. The resulting model will perform better and will be easier to maintain during the deployment stage. In the current literature, data-driven variable selection methods have been investigated within the context of spectroscopic data and bioinformatics. In these studies, the variable selection methods assume that the inherent correlation in the entire data set remains fixed. This is not the case in common industrial processes. In this paper, existing variable selection methods based on partial least squares (PLS) will first be evaluated. Second, we will present a new approach called moving window variable importance in projection (MW-VIP) to target the selection of correlations present in segments or small clusters. Finally, a set of new evaluation criteria will be presented along with industrial data set modeling results to demonstrate the effectiveness of our proposed approach.

This paper proposes a model-based detection and isolation (FDI) system based on nonlinear state estimation that can be applied to nonlinear systems. The proposed FDI system uses an extended Kalman filter (EKF), in which conditions based on high filtering are defined to best serve the FDI objectives. A better understanding of the residual trends, calculated from the difference between measurements and the EKF estimates, can be obtained when a fault occurs by developing a model that is able to predict the behavior of the residuals. This model is utilized as the basis for detection and isolation of single and multiple faults. Comparisons with data driven techniques, specifically principal component analysis (PCA) and Kernel PCA, show superior isolation results, having the advantage of distinguishing single and multiple faults from a diverse array of possible faults, a common occurrence in complex processes. The proposed approach is validated using an experimental air heater.

Very simple proportional-integral-derivative (PID) controller tuning rules for a wide range of stable processes are available. However, for unstable processes, the design trend is for controllers to be more complex for better performances. Here, the design concept of “simplicity” is extended to unstable processes. Simple desired closed-loop transfer functions for the direct synthesis method and simple approximations of the process time delay are utilized for unstable processes. Very simple tuning rules for PID controllers and set-point filters are obtained, yielding similar or even improved performances over previous more complicated PID controller tuning methods.

This paper presents an integrated virtual metrology (VM) and quality excursion detection framework for online implementation on a plasma etch tool. Traditional external metrology have inherent dela...

High-mix manufacturing in the semiconductor industry has driven the development of several nonthreaded state estimation methods. These methods share information among different manufacturing contex...

The SIMC method by Skogestad (J. Process Control 2003, 13, 291–309) to tune the PID controller is revisited, and a new method (K-SIMC) is proposed. The proposed K-SIMC method includes modifications of model reduction techniques and suggestions of new tuning rules and set point filters. Effects of such modifications are illustrated through simulations for a wide variety of process models. The proposed modifications permit the SIMC method to be applied with more confidence.

The issue of dynamic batch profile synchronization is addressed. By converting synchronization into a dynamic optimization problem, dynamic time warping (DTW) and derivative DTW (DDTW) show the best performance by far. To deal with the singularity point and numerical derivative estimation problems of DTW and DDTW in the presence of noise, a robust DDTW algorithm is proposed by combining the Savitzky–Golay filter and DDTW algorithm together. A comparative analysis of robust DDTW and available methods is performed on simulated and real chemical plant data.

Batch process monitoring methods, such as multiway PCA and multiblock multiway PLS, make use of process variable time profiles to normalize and define most likelihood trajectories for statistical process control. Nevertheless, a continuous process analytics counterpart has not been developed, nor addressed in the literature. This paper presents a novel methodology that defines “state variables” to determine the multiple operating points around which a continuous process operates. In this manner, the operating region is divided into multiple regions (states) and shifts in operating conditions are captured by such state variables. Transition trajectories between states are calculated to determine the most likely path from one state to another. This methodology is referred as multistate analytics and can be implemented in the context of empirical monitoring methods, named multistate PLS and multistate PCA. A case study using data from carbon dioxide removal process shows that multistate analytics is beneficial for statistical monitoring of continuous processes.

Parallel coordinates is a recognized visualization technique in which data points, each defined by multiple coordinates, are represented by an unlimited number of adjoining parallel axes. This type of visualization technique is suitable for process monitoring applications in industrial facilities where a significant number of sensors are used to detect and identify abnormal operating conditions. This work makes use of principal component monitoring methods implemented in parallel coordinates plots, named PC2. The PC2 capabilities to visualize confidence regions of operations, evaluate models with different number of principal components, compare faulty events and determine the frequency of false alarms are here demonstrated. The monitoring visualization technology presented by PC2 was successfully used for early detection of compressor surge and column flooding using actual process data. © 2012 American Institute of Chemical Engineers AIChE J, 59: 445–456, 2013

In situ adaptive tabulation or ISAT based moving horizon estimation (MHE) is suggested as a fast and robust approach for state estimation. Computational issues with a moving horizon constrained state estimation technique like MHE are discussed. Implementation of storage and retrieval approach of ISAT for state estimation is proposed to maintain the accuracy and robustness of MHE, while generating the estimates at a reduced computational cost (∼300 times faster). Comparison with the widely used nonlinear state estimation technique of extended Kalman filtering (EKF) shows better performance using ISAT–MHE. Case studies with nonlinear discrete-time and continuous-time systems are carried out using ISAT, which is tailored for solving the optimal estimation problem.

The relay feedback system often shows an asymmetric response because of initial transient states, disturbances, and process nonlinearity. To restore a symmetric response of the relay feedback system, an iterative adjustment of input or output bias is required. Instead of trying to obtain a symmetric response, the asymmetric response can be analyzed to estimate the ultimate data or a parametric model of process. However, the asymmetric response causes additional errors in estimating the ultimate properties of the process. An area method is proposed to reduce these errors. Because integrals (areas) of the relay responses reduce the effects of the high-order harmonic terms significantly, the proposed method shows better accuracy in obtaining frequency response data and parametric models compared with previous approaches.

Industrial process monitoring tools require robust and efficient estimation techniques that maintain a high service factor by remaining online during abnormal operating conditions, such as during loss of measurements, changes in control status, or maintenance. Constraints incorporate additional process knowledge into estimation by bounding estimated disturbances within feasibility limits thereby providing robustness to faulty measurements or conditions that violate process models. Moving horizon estimation (MHE) and unscented Kalman filtering (UKF) are two estimation techniques that permit incorporation of constraints prior to evaluating the a priori estimate. This paper evaluates both constrained nonlinear estimators versus the extended Kalman filter (EKF) using industrial process data provided by ExxonMobil Chemical Company. Results provide short-term insight into the fouling process, and parameter estimates produced by UKF and MHE are shown to be more accurate than EKF.

In situ adaptive tabulation (ISAT) is applied to dynamic process simulators for reducing computational run-time. Several enhancements of previous approaches are presented here, including a method for estimating the sensitivities using input−output data, along with different strategies for record distribution. A modified version of the original algorithm (mISAT) to improve performance of ISAT is also suggested. Case studies for first-principles and data-driven models using ISAT are performed to generate accurate trajectories, which are essentially the same as those obtained by direct integration. Computational speed-up using ISAT is also shown for these studies.

The amplitude and period of the relay feedback oscillation are used to obtain the ultimate parameter of a dynamic process and to tune a proportional−integral−derivative (PID) controller automatically. Equations for the ultimate gain and period are based on ignoring higher harmonic terms in the relay feedback response. The integral of the relay feedback response can suppress the higher harmonic terms, and the amplitude of this integral has been found to be better for estimating the ultimate gain of a process. As extensions, simpler methods that use several area calculations of the relay feedback response are proposed and shown to provide more accurate frequency responses, ultimate gain and period, and parametric models.

As the standard size of silicon wafers grows and performance specifications of integrated circuits become more demanding, a better control system to improve the processing time, uniformity and repeatability in rapid thermal processing (RTP) is needed. Identification and control are complicated because of nonlinearity, drift and the time-varying nature of the wafer dynamics. Various physical models for RTP are available. For control system design they can be approximated by diagonal nonlinear first order dynamics with multivariable static gains. However, these model structures of RTP have not been exploited for identification and control. Here, an identification method that iteratively updates the multivariable static gains is proposed. It simplifies the identification procedure and improves the accuracy of the identified model, especially the static gains, whose accurate identification is very important for better control.

The traditional way of state estimation in semiconductor manufacturing, known as “threaded” state estimation, segregates the process data into different bins and uses the ones that match the current event of the specific context information (such as tools, layers, products) to update the process state. The limitation of threaded state estimation is that a narrowly defined process stream can result in too many different threads and insufficient data for each thread. This limitation becomes more severe in high-mix manufacturing, where there can be many products and many tools. Hence there is great interest in estimation methods that utilize all available data in the analysis. In this work, the characteristics inherent in state estimation of high-mix semiconductor manufacturing processes are analyzed, and a general framework is introduced for the non-threaded state estimation methods, i.e., state estimation without segregating the process data into different bins. The framework is based on the best linear unbiased estimate (BLUE) of a simplified stationary singular Gauss–Markov process, and non-threaded state estimation methods based on the Kalman filter, least squares and recursive least squares (RLS) are analyzed using the general framework. Simulation examples are presented to illustrate the equivalence between different algorithms. As real processes are rarely stationary, modifications to the Kalman filter and RLS are discussed. We show that in non-threaded state estimation, how to regulate the estimate covariance plays a significant role in estimation performance. To handle nonstationary disturbances that often occur in semiconductor processes, Bayesian-enhanced adaptive versions of the Kalman filter and RLS are proposed. Both simulated and industrial nonstationary processes are used to demonstrate the effectiveness of the proposed adaptive methods.

Design of experiments (DOE) for parameter estimation in dynamic systems is receiving more attention from process system engineers. In this paper, a principal component analysis (PCA)-based optimal criterion (P-optimal) for model-based DOE is proposed that combines PCA with information matrix analysis. The P-optimal criterion is a general form that encompasses most widely used optimal design criteria such as D-, E-, and SV-optimal, and it can automatically choose the optimal objective function (criterion) to use for a specific differential and algebraic (DAE) system. Two engineering examples are used to validate the algorithms and assumptions. The advantages of P-optimal DOE include ease of reducing the scale of the optimization process by choosing parameter subsets to increase estimation accuracy of specific parameters and avoid an ill-conditioned information matrix.

We have analysed several aspects of insulin-dependent diabetes mellitus, including the glucose metabolic system, diabetes complications, and previous and ongoing research aimed at controlling glucose in diabetic patients. An expert review of various models and control algorithms developed for the glucose homeostasis system is presented, along with an analysis of research towards the development of a polymeric insulin infusion system. Recommendations for future directions in creating a true closed-loop glucose control system are presented, including the development of multivariable models and control systems to more accurately describe and control the multi-metabolite, multi-hormonal system, as well as in-vivo assessments of implicit closed-loop control systems.

A fundamental dynamic model was successfully used in the implementation of multiple control methodologies via a novel inferential control strategy using HYSYS to treat missing process measurements. Results from steady state and dynamic testing of an azeotropic distillation system of methanol, normal pentane, and cyclohexane were obtained. Steady-state equilibrium and nonequilibrium models were developed and validated with experimental data from a packed distillation unit operated at finite reflux. Dynamic multicomponent distillation experiments were also carried out, and experimental process data were collected using the pilot scale experimental unit. Two different variable pairings were studied, and the results from individual control loop configurations were compared with a multivariable control strategy using the model predictive control (MPC) software Predict Pro.

Multiloop control systems are often used for industrial multivariable processes because of their simplicity. To design multiloop control systems, single-input single-output (SISO) methods that guarantee specified closed-loop characteristics can be applied. Because the design equations are nonlinear due to loop interactions and may cause computational difficulties, a continuation method is proposed to obtain the solutions. By choosing the interaction level as a continuation parameter, we can design multiloop control systems without worrying about numerical difficulties such as divergence. To illustrate the approach, the modified Ziegler−Nichols method is applied to design multiloop control systems.

This paper outlines a method to implement nonlinear model predictive control (NMPC) in real-time control applications. Nonlinear model identification is generally seen as a major obstacle to implementing NMPC. However, once an accurate nonlinear model is identified the computational effort is often too great to implement the model in a real-time application. The approach in this paper is a two step process, model reduction followed by computational reduction. Model reduction is accomplished by computing balanced empirical gramians. Computational reduction is accomplished by using the method of in situ adaptive tabulation (ISAT) which was previously developed for computational reduction of turbulent flame direct numerical simulations and is extended to the sequential NMPC framework in this work. A case study is performed with a binary distillation column model with 32 states. By computing balanced empirical gramians the number of states is reduced to five. With ISAT, the computational speed is 85 times faster than the original NMPC while maintaining the accuracy of the nonlinear model. Since ISAT is a storage and retrieval method, it is compared to artificial neural networks in another case study. This case study is performed with a dual CSTR model with 6 states. Open loop and closed loop step tests are performed to demonstrate the superior quality of ISAT in extrapolating outside of the training domain.

We present some insights on the relative error matrix between the process transfer function matrix and its diagonal matrix, which has been used in analyzing dynamic interactions by several authors. Proposed interaction measures can be interpreted in terms of differences of the complementary sensitivity function matrices when loops are closed, extending the concept of gain changes in the relative gain array. To employ the interaction measures, control systems are designed in advance using internal model control. With changes in the closed-loop time constant, proper pairings based on closed-loop response speeds can be obtained.

Nonlinear model predictive control has become increasingly popular in the chemical process industry. Highly accurate models can now be simulated with modern dynamic simulators combined with powerful optimization algorithms. However, computational requirements grow with the complexity of the models. Many rigorous dynamic models require too much computation time to be useful for real-time model based controllers. One possible solution to this is the application of model reduction techniques. The method introduced here reduces nonlinear systems while retaining most of the input–output properties of the original system. The technique is based on empirical gramians that capture the nonlinear behavior of the system near an operating point. The gramians are then balanced and the less important states reduced via a Galerkin projection which is performed onto the remaining states. This method has the advantage that it only requires linear matrix computations while being applicable to nonlinear systems.

Existing strategies for the solution of the nonlinear dynamic data reconciliation problem use the process model as a constraint which is expressed as a differential−algebraic equation system. Modeling a process using conservation laws may require a considerable number of equations to obtain an accurate representation of the system. It is possible to model a process using commercial dynamic simulation software. However, this also requires the solution of a large number of equations interfaced to reliable optimization software in order to perform data reconciliation. This paper focuses on two new approaches for dynamic data reconciliation using model identification tools and commercial dynamic simulation software. The first one is based on an analogy to the nonlinear dynamic data reconciliation method developed by Liebman et al.1 The second approach uses time series analysis to generate a simplified model of the plant. A simplified process model is generated by a model identification method to replace the s...

This paper presents a computational approach to determine a reduced order model of a nonlinear system. The procedure is closely related to balanced model reduction and introduces the concept of covariance matrices for local controllability and observability analysis of a nonlinear system. These covariance matrices are an extension of gramians of a linear system and are used to determine unobservable and uncontrollable parts of the system for a given operating region. Additionally, an algorithm is introduced that eliminates these nonminimal parts of the model and can further reduce the model, i.e., the number of state variables. This minimal realization/model reduction procedure is simple to implement and can be applied locally to any stable system without making any assumptions about observability and controllability. Examples are presented to demonstrate the procedure. When the algorithm is applied to linear systems, it reduces to well-known techniques for minimal realization and balanced model reduction.

We show that subspace identification in the frequency domain can be effectively used to simulate the time-domain response of large dimensional closed-loop systems with time delays. The proposed method can also be applied to large dimensional discrete-time systems.

Recent approaches for nonlinear and dynamic data reconciliation suffer from inapplicability and infeasibility for large systems. Because these systems are expressed by differential and algebraic equations, the complete problem definition requires a considerable number of equations that need to be solved simultaneously during the solution of the nonlinear programming problem. One way in avoiding this is to use a commercial software package to model a process and to reduce the size of the model by generating an input-output model from the simulation results. In this research two different approaches are presented to describe dynamics of the system and reduce the size of the model by model identification techniques.

This paper analyses the sodium nitroprusside infusion rate control problem that occurs in patients after surgery. The process includes several parameters that can vary over a significant range. Analysis of the model parameters has shown that an internal model controller (IMC) that can meet robust stability and performance criteria can be designed for variations in all of the parameters but one. This model parameter is found from an adaptation law that is introduced in this paper. The resulting controller was shown to meet the robust performance criteria for several simulated patients including some that showed extreme reactions to the drug infusion.

In this paper model reduction methods are used to obtain a nonlinear process model for designing a model predictive controller (MPC). The corresponding controller and its closed-loop response is then compared with controllers that are determined from the original model and a linearized version of this model. The reduced dimensional nonlinear MPC controller performs almost as well as the nonlinear MPC controller that is based on the original model and considerably better than the linear MPC controller.

It is hard to design robust control systems for interacting multivariable processes. When processes are ill-conditioned, difficulties become more serious due to sensitivities on plant variations. For robust control systems, we decompose systems to slow subsystems and fast subsystems. When the slow subsystem is detuned and maintained to be slow, the fast subsystems can be made to be robust to plant variations. On the other hand, the slow loop is sensitive to plant variations. Adaptive control technique is introduced to the slow loop to maintain performances of the slow loop. Because the slow subsystem is chosen to be single-input single-output and controlled to be slow, the adaptive control scheme is very simple. Simulations show that the proposed control system has nominal performances similar to usual multivariable control systems while maintaining better robustness.

For some ill-conditioned processes, small element errors due to the classical element-by-element identification cannot be tolerated and control systems based on models with such errors can suffer f...