D. Semino

Control of the quality of polymer products in continuous reactors: comparison of performance of state estimators with and without updating of parameters

Abstract In the paper the control of the product quality in polymerization reactors is analysed in the presence of persistent perturbations (unmodeled disturbances, modeling errors), as met in industrial reactors. The free radical polymerization of methyl-methacrylate in a continuous stirred tank reactor is studied. It is shown that state estimators (Extended Kalman Filters) with constant parameters cannot give offset free performance. Criteria for the selection of a set of parameters to be updated as additional states in the filter and to evaluate their effectiveness in opposing the action of realistic perturbations are given by an analysis of the linearized model of the system. Performance of different types of estimators, including one and two-time scale filters, with and without updating of parameters, is analysed by simulation on the full order process and the predictions made by previous analysis are confirmed. In the most common case of presence of perturbations affecting the energy balance and the concentration of initiator in the reactor, offset free control of the molecular weight of the product can be achieved by means of a filter which is based only on measurements of temperature and conversion and makes an update of two parameters. In the case that also the kinetic model of the polymerization reaction is affected by errors, a two-time scale filter, which makes use also of Molecular Weight values and updates three parameters, becomes necessary to obtain offset free performance.

Improved identification and autotuning of PI controllers for MIMO processes by relay techniques

Abstract A technique for on-line identification and tuning is proposed to be used in the framework of a MIMO autotuning procedure. The proposed technique does not suffer from the risks of instability and the lack in performance of common tuning techniques in MIMO autotuning. Identification is accomplished through an extension of the well known ATV autotune identification method and requires only few additional tests in order to obtain some more knowledge about the process. The resulting model, which describes with good precision the process in a region of frequencies around the critical point, is then used for tuning: the integral time is found as a function of the model time constants and delay, while the gain is computed in order to give a desired value of the closed-loop resonance peak. Examples of application show that advantages over other proposed techniques can be retained for processes having different dynamic characteristics.

Measurements of particle-size distribution during suspension polymerization

The particle-size distribution is one of the main characteristics of polymers produced by suspension polymerization which determines the final properties of the product. Since the on-line measurement of particle-size distribution during the course of polymerization is not a straightforward task, a technique has been developed to accomplish such a measurement for a predefined polymerization process. In this paper the experimental procedure is described in detail and the experimental data are shown for the batch suspension polymerization of methyl methacrylate using benzoyl peroxide as initiator, water as suspending phase and agarose as suspending agent. The particle-size distributions have been measured both for a base case and also when varying some operating conditions including stirring speed, water-to-monomer weight ratio, and suspending agent concentration. Average values and variances of the surface distribution and of the volume distribution are also computed and shown for comparison. © 1999 Society of Chemical Industry

Parameter estimation in Extended Kalman Filters for quality control in polymerization reactors

The issue of parameter estimation in Extended Kalman Filters is addressed in this paper. It is shown that the use of parameters as additional system states is appropriate whenever there are systematic errors in the model of the process. Such is the case of the filters which are used to estimate quality properties of polymers which are not measured on-line in polymerization reactors. Criteria for the choice of the parameters to be updated are suggested. Moreover, in order to realize the possibility of obtaining offset free response, the effect of errors on both updated and not updated parameters, both on the observable and on the unobservable system states is analysed. Applications to the polymerization of methylmethacrylate (MMA) in a continuous stirred tank reactor confirm the theoretical results.

Optimal operation of a separation plant using artificial neural networks

Abstract An ANN (Artificial Neural Network) based optimizer has been built and applied to the separation of a gas flow coming out of a hydrocracking reactor. The ANN based optimizer is shown to be able to achieve a precision similar to the one of an optimizer based on a first principle model with a very significant decrease in computational time. A central point of the development of the ANN models lies in the choices related to the training procedure; suggestions about the criteria which should lead such choices have been given.

Automatic Tuning of PID Controllers for Unstable Processes

Abstract The tuning of PI and PID controllers for unstable processes is addressed. A well known technique which makes use of the closed-loop dynamics with a relay to find the controller parameters is here extended to unstable processes. The stabilizability of unstable systems with PID controllers is analysed first; then the technique is developed and tested with good results on a strongly nonlinear process which is described by a complex model that cannot be easily used for control purposes. The technique is particularly appropiate for frequent automatic retuning.

Optimal modified models for robust predictive controllers

Abstract Predictive controllers suffer from sensitivity and robustness problems in the presence of strong ill-conditioning such as all model based controllers. In spite of operating on the common robustness parameters of predictive controllers an optimization technique has been developed in order to find the alternative model to be used in place of the nominal one, which offers the best compromise between nominal performance and robustness to uncertainty. Such a modification can be undertaken only on the model gains or also on the dynamic parameters according to the process characteristics. Applications to badly ill-conditioned distillation columns have shown the great potential of the technique.

Control configuration selection in recycle systems by steady state analysis

Abstract Control structures for chemical processes with recycle streams are often affected by the so-called “snowball effect”, which consists in very large variations of the manipulated variables required to reject relatively small disturbances. This occurrence is related to a wrong choice of the control structure, which may be relatively easy to detect for simple system, but is difficult to discover for complex ones. A systematic steady state analysis has been developed for systems with recycle, which is able to classify all possible control configurations which are and which are not affected by snowballing.

Multiloop autotuning using relay feedback: Limits and extensions

Abstract In this paper a recently presented tuning technique for multiple-input multiple-output (MIMO) systems (Loh et al., 1993), which avoids a trial and error approach, is analysed into detail to understand to what systems it can be extended and in which cases it can be unsuccessful. The technique, which is based on sequential and (if it is the case) iterative autotuning of PI controllers using relay feedback, is shown to be applicable also to the control of MIMO systems with negative values of the relative gain array (RGA) and to systems with inner cascade loops. Moreover controllers for MIMO open-loop unstable systems, which are stabilizable by conventional controllers, can be tuned combining the multiloop autotuning with a recently developed autotuning technique for single-input single-output unstable systems (Semino and Gori, 1995). On the other side, it is shown that for some MIMO open-loop stable systems the tuning procedure can be successful only if an ad-hoc refinement of the Ziegler-Nichols tuning method is adopted.

Design of double filter IMC controller for ill-conditioned distillation columns

Abstract A direct method for robust design of double filter IMC controllers is presented. With respect to the general scheme, already proposed to control ill-conditioned plants, this method is based on a simple filter structure depending on fewer parameters. Guidelines for their selection are indicated; in many cases performance is mostly affected by a single parameter, thus making the controller tuning a much easier task to perform.