P.R. Krishnaswamy

Multi-linear model-based fault detection during process transitions

Abstract Process transitions due to startup, shutdown, product slate changes, and feedstock changes are frequent in the process industry. Experienced operators usually execute transitions in the manual mode as transitions may involve unusual conditions and nonlinear process behavior. Processes are therefore more prone to faults as well as inadvertent operator errors during transitions. Fault detection during transition is critical as faults can lead to abnormal situations and even cause accidents. This paper proposes a model-based fault detection scheme that involves decomposition of nonlinear transient systems into multiple linear modeling regimes. Kalman filters and open-loop observers are used for state estimation and residual generation based on the resulting linear models. Analysis of residuals using thresholds, faults tags, and logic charts enables on-line detection and isolation of faults. The multi-linear model-based fault detection technique has been implemented using Matlab and successfully tested to detect process faults and operator errors during the startup transition of highly nonlinear pH neutralization reactor in the laboratory.

An adaptive internal model control strategy for pH neutralization

Abstract An adaptive internal model control strategy that is capable of providing effective and robust control for pH neutralization has been developed. The approach is based on combining the concepts of non-linear internal model control, strong acid equivalent (Wright and Kravaris, 1991, Ind. Engng Chem. Res. 30 , 1561–1572) and a simplified adaptive mechanism (Huberman and Lumer, 1990, IEEE Trans. Circuits Systems 37 , 547–549). The resulting control system is extensively tested via simulation for acid-base neutralization. The results show the excellent capabilities of the proposed method for disturbance rejection, servo control and tolerance to model uncertainty.

Fault detection during process transitions: a model-based approach

Abstract Startup, shutdown and other transitions are integral to batch and continuous process operations. Operators usually execute transitions in manual mode. Processes are therefore prone to operator errors in addition to process faults during transitions. If undetected, such abnormalities can lead to process downtime and in the worst case, accidents. Although essential, fault detection during transitions has received little attention in literature. This paper presents a novel multiple filters and observers based fault detection scheme using (i) a nonlinear process model, and (ii) knowledge of the standard operating procedure for executing the transition. Extended Kalman filters, Kalman filters, and open-loop observers are used to estimate process states during the transition and generate residuals. These residuals indicate deviations from normal operation due to process faults and operator errors. The model-based scheme has been implemented in Matlab/Simulink and found to successfully detect faults during the startup of an experimental pH neutralization CSTR.

Estimating second-order dead time parameters from underdamped process transients

Two simple methods for determining second-order dead time model parameters from underdamped process transients are presented. One of the methods relies on three characteristic points of the oscillatory step response curve. These three points attempt to minimize the integral of absolute error between process and model responses. The other method is based on just two points of the step response. Illustrative examples show that the proposed techniques allow rapid and reliable estimate of parameters.

Closed-loop tuning of process control systems

A technique for tuning PID controllers, involving a single dynamic test (such as a step or pulse change in set point) implemented during closed-loop operation, is proposed. The resulting transient data are used to predict the frequency response of the open-loop process which in turn is used to compute the optimum controller settings. Simulation results reveal that this technique provides reliable tuning constants even when such practical problems as process noise and unanticipated load upsets arise during implementation. A comparison with the recently proposed tuning alternative (Yuwana and Seborg, A.I.Ch.E. J.28, 434, 1982; Jutan and Rodriguez, Can. J. Chem. Engng62, 802, 1984) confirms that in general the suggested procedure, apart from being more flexible, yields also relatively better results.

Application of a plant-wide control design to the HDA process

Abstract With the increasing interest in and the need for plant-wide design and control, it is desirable to develop rigorous models for realistic and large-scale processes with recycle streams, energy integration and non-ideal systems, and use them for evaluating plant-wide control schemes. In this study, a rigorous model for the hydrodealkylation of toluene (HDA) process is developed using the commercial software, hysys.plant . After reviewing the reported methods for plant-wide control, a systematic method, namely, Control Configuration Design (CCD) method, is selected for application to the HDA process. The resulting control structures from the application of this method are evaluated and compared through rigorous dynamic simulation. The results show that the CCD method successfully yields workable base-level regulatory control structures for the HDA process. The control structures obtained by the CCD method are also compared with that reported by Luyben et al. (Plant-wide Process Control, McGraw-Hill, 1998) using their nine-step method.

Use of alternate process variables for enhancing pH control performance

Abstract In pH control, the process variable that is very often used is pH itself although use of other alternate variables may provide better performance. In this paper, use of three easily implementable process variables (namely, difference in hydrogen and hydroxyl ion concentration η , hydrogen ion concentration C H and process pH) are experimentally investigated in the context of two control strategies, namely, PI control and adaptive nonlinear internal model control. Experimental results show that, in general, use of η provides superior control performance. In the case of adaptive nonlinear internal model control, significantly improved control is obtained with η as the process variable. This improvement is attributed to the reduction in process static nonlinearity on account of η . Besides, η can also be easily calculated from measured pH for any type of neutralization system. Thus, there is a strong case for using η instead of pH as the process variable in pH control.

Experimental application of robust nonlinear control law to pH control

An experimental application of RNCL to pH control is presented. The experimental setup consists of a computer controlled CSTR (continuous stirred tank reactor) and the control objective is to maintain the pH of a strong-acid strong-base system at the neutralization point in the presence of disturbances. The algorithms considered are PI (proportional-integral) control, nonlinear IMC (internal model control), and RNCL (robust nonlinear control law). The experimental results point to the excellent capability of RNCL for controlling pH comparison to nonlinear IMC and PI control.

On-Line Process Identification from Closed-Loop Tests Under PI Control

A method for establishing second order plus dead time model parameters under closed-loop PI control is proposed and its performance compared with the recent method of Park et al. 1 . It can be readily applied regardless of whether the response obtained during closed-loop identification is under- or over-damped. The drawbacks of the other methods in existence, i.e., poorer model fit due to first order plus dead time approximation and/ or the need for a step test under P-control, are overcome. The required closed-loop response data can be obtained while the process is in normal operation since many industrial controllers are of the PI-type. Simulation results reveal that the proposed method produces reliable parameter estimates and its accuracy is comparable with the method of Park et al . 1 . The relative advantages of both these methods are discussed.

Set point weighting for simplified model predictive control

Abstract Set point weighting (SPW) is a technique for achieving optimum performance of feedback control for both load and set point disturbances. It involves assigning a weighting factor to the set point in the proportional action part of the control algorithm. In this paper, simplified model predictive control (SMPC) is modified to include SPW. The modification consists of first tuning the SMPC controller for load disturbances and then optimizing the weighting factor for set point disturbances. The performance of the resulting feedback control system is studied through simulation and the results compared with those in the absence of SPW. In general, the results demonstrate that SPW offers significant benefits at practically no additional expense to digital control systems experiencing both load and set point variations.