Search

When a fault occurs in a process, it slowly propagates within the system and affects the measurements triggering a sequence of alarms in the control room. The operators are required to diagnose the cause of alarms and take necessary corrective measures. The idea of representing the alarm sequence as the fault propagation path and using the propagation path to diagnose the fault is explored. A diagnoser based on hidden Markov model is built to identify the cause of the alarm signals. The proposed approach is applied to an industrial case study: Tennessee Eastman process. The results show that the proposed approach is successful in determining the probable cause of alarms generated with high accuracy. The model was able to identify the cause accurately, even when tested with short alarm sub-sequences. This allows for early identification of faults, providing more time to the operator to restore the system to normal operation.

This article studies the problem of stabilizing a two-stage continuous bioreactor system. A simple dynamic model of the system is first introduced based on a detailed mass balance model, and then used to derive a constant-yield controller to stabilize the system at given design steady state conditions. Using Lyapunov stability analysis, this control law is proved to guarantee stability of the closed loop system over the entire positive orthant. Simulation results show the successful performance of the controller. The control law is proved to be robust with respect to errors in the kinetic parameters and in the inlet feed concentration, in the sense of preserving its stability region. Performance of the control system can be enhanced if a feed-forward measurement of the inlet feed concentration can be incorporated in the control law. © 2017 American Institute of Chemical Engineers AIChE J, 64: 1295–1304, 2018

A fundamental practice in process engineering is monitoring the state dynamics of a system. Unfortunately, observability of some states is related to high costs, time, and efforts. The mechanistic event recognition (MER) aims to detect an event (defined as a change of the system with specific significance to the operation of the process) that cannot be directly observed but has some predictable effect on the dynamics of the systems. MER attempts to apply fault diagnosis techniques using mechanistic “recognition” models to describe the process. A systematic method for building recognition models using optimal experimental design tools is presented. As proof of concept, the MER approach to detect organic matter depletion in sequencing batch reactors, measuring only ammonia, dissolved oxygen, and nitroxides is applied. The event, that is, consumption of organic matter to a level below 50 gCOD/m3, was successfully detected even though microbial activity is known to continue after organic matter depletion.

PI and PID controllers continue to be popular methods in industrial applications. It is well known that linear PI and PID controllers result from the application of model-based controller design methods to linear first- and second-order systems. It is shown that nonlinear PI and PID controllers result from the application of nonlinear controller design methods to nonlinear first- and second-order systems. As a result, the controllers resulting from nonlinear model-based control theory are put in a convenient form, more amenable to industrial implementation. Additionally, the quantities used in the controller are useful for monitoring the process and quantifying modeling error. Chemical engineering examples are used to illustrate the resulting control laws. A simulation example further demonstrates the performance of the nonlinear controllers, as well as their useful process monitoring quantities.

The experimental work concerns the multivariable nonlinear control of a pilot-size continuous polymerization reactor with generically singular characteristic matrix. The control problem is to control conversion and temperature in a continuous stirred tank reactor by manipulating two coordinated flow rates (reactor residence time) and two coordinated heat inputs. A nonlinear controller is synthesized within the framework of the globally linearizing control (GLC) method and is implemented on a microcomputer. Conversion is inferred from on-line measurements of density and temperature. A key feature of the control problem is that its characteristic matrix is generically singular. Singularity of the characteristic matrix is handled by using a dynamic input/output linearizing state feedback rather than a static feedback. A reduced-order observer is used to calculate the monomer, initiator, and solvent concentration estimates, which are needed for the calculation of controller action. In the presence of active state and input constraints, the reactor-startup and setpoint-tracking performance of the controller is evaluated through experimental runs

This work concerns the synthesis of discrete-time feedforward/feedback control systems for general nonlinear processes with stable zero dynamics. Depending on the process under consideration, the derived feedforward/feedback controllers can completely eliminate the effect of measurable disturbances and produce a prespecified linear response with respect to a reference input, or provide integral-square error optimal response to step changes in the disturbances and a prespecified linear response with respect to a reference input. In either case, the developed feedforward/feedback controllers allow for the asymptotic rejection of unmeasurable disturbances. These controllers are derived within the globally linearizing control frame-work, first under full state information and then in the absence of state measurements. The internal stability of the closed-loop system is addressed. The derived controllers are interpreted from a model-predictive point of view, and their connections with the feedforward internal model control and the model algorithmic control are established. The theoretical results are illustrated through a continuous stirred-tank reactor example.

This experimental work concerns the multivariable nonlinear control of a continuous stirred-tank polymerization reactor. The globally linearizing control (GLC) method is implemented to control conversion and temperature in the reactor in which the solution polymerization of methyl methacrylate takes place. Control of conversion and temperature is achieved by manipulating the flow rate of an inlet initiator stream and two coordinated heat input variables. Conversion is inferred from on-line measurements of density and temperature. A reduced-order state observer is utilized to estimate the concentrations of monomer, initiator and solvent in the reactor. The concentration estimates are used in the control law. This study demonstrates the considerable computational efficiency of the nonlinear controller, which is implemented on a microcomputer. The experimental results show the excellent performance of the controller in the presence of active state and input constraints. A systematic approach is also given for the synthesis of output feedback controllers within the GLC framework for processes with secondary outputs.

This work concerns the synthesis of discrete-time nonlinear controllers for nonlinear processes that make the closed-loop system linear in an input/output sense. The synthesis of state feedback controllers is studied first, followed by the synthesis of dynamic output feedback controllers. Both problems are solved within the globally linearizing control (GLC) framework. Precise theoretical connections between the derived controllers and model algorithmic control (MAC) are established. The theory is illustrated by a chemical reactor example.

The design of controllers for nonlinear, nonminimum-phase processes is one of the most difficult control problems currently faced. Current available control algorithms for nonlinear processes rely implicitly or explicitly on an inverse of the process. Linear control methods for nonminimum-phase processes are based on a decomposition of the process into a minimum-phase and a nonminimum-phase part. Such a decomposition is an open problem for nonlinear systems. In this work, a control structure called the minimum-phase output predictor for nonlinear, nonminimum-phase processes is developed. The structure is based on the notion of statically equivalent outputs; a minimum-phase, statically equivalent output is estimated on-line and then an available nonlinear control algorithm is used to control it to set point. The advantage of the proposed formulation is that it is based on the calculation of an output function, not on a decomposition of the process dynamics. The proposed control methodology is applied and its performance is evaluated for a chemical engineering example.

This work concerns general multiple-input/multiple-output (MIMO) nonlinear systems with nonsingular characteristic matrix. For these systems, the problem of inversion is revisited and explicit formulas are derived for the full-order and the reduced inverse system. The reduced inverse naturally leads to an explicit calculation of the unforced zero dynamics of the system and the definition of a concept of forced zero dynamics. These concepts generalize the notion of transmission zeros for MIMO linear systems in a nonlinear setting. Chemical engineering examples are given to illustrate the calculation of zero dynamics. Input/output linearization is then interpreted as canceling the forced zero dynamics of the system, and precise internal stability conditions are derived for the closed-loop system.

This paper concerns general MIMO nonlinear processes, whose dynamic behavior is described by a standard state-space model of arbitrary order, including measurable disturbances. The concept of relative order of an output with respect to an input, extended to include disturbance as well as manipulated inputs, is generalized in a MIMO context and it is used to obtain a characterization of the dynamic interactions among the input and the output variables. A synthesis formula is calculated for a feedforward/state feedback control law that completely eliminates the effect of the measurable disturbances on the process outputs and induces a linear behavior in the closed-loop system between the outputs and a set of reference inputs. The input/output stability and the degree of coupling in the closed-loop system are determined by appropriate choice of adjustable parameters. A MIMO linear controller with integral action completes the feedforward/feedback control structure. The developed control methodology is applied to a continuous polymerization reactor and its performance is evaluated through simulations.

This paper deals with SlSO nonlinear processes and their control with nonlinear static state feedback. The Byrnes-lsidori concept of nonlinear zeros is first reviewed and illustrated with two chemical engineering examples. This concept is then used to interpret input/output linearizing state feedback as a nonlinear analog of placing poles at the process zeros. This interpretation leads to closed-loop internal stability conditions for nonlinear processes under input /output linearizing state feedback.

Mise en place d'un systeme de commande qui permet d'obtenir des limites superieures de la modelisation de l'erreur pour les processus non lineaires, et de concevoir une loi de retroaction qui garantit une stabilite et de bonnes performances pour toute perturbation entrant dans les limites donnees. Ce systeme est applique a la commande d'un reacteur parfaitement agite isotherme ou ont lieu, en phase liquide, des reactions consecutives du type: A⇄B→C ou P→Q→R

A systematic method for synthesizing feedforward/state feedback controllers for a broad class of SISO nonlinear systems with measurable disturbances is presented. Depending on the structural characteristics of the system, the control law can be static or dynamic. The closed-loop system is independent of the measurable disturbances and linear with respect to set point changes. The performance of the proposed control scheme is illustrated through an example of composition control in a system of three CSTR's in series.

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.