Volume 4, Issue 2 - Modeling and Control

This special issue collects papers based on the interplay between modeling and control. These features are combined into various aspects for exploiting the potential of physical and mathematical modeling while assessing the practices to attack fundamental issues of control theory, getting a deeper understanding of the related phenomena, and finding useful applications by establishing a common framework to deal with vastly different expressions of modeling and control of industrial processes.

The applicability of pulse respirometry, for the estimation of kinetic and stoichiometric parameters in pure cultures was evaluated by comparison with traditional chemostat method. Pseudomonas putida F1 was cultured in a continuous stirred tank reactor, using glucose as sole carbon source. The reactor was operated under steady-state with six dilution rates, ranging from 0.06 to 0.35 h-1. Substrate and biomass concentration were measured and used to estimate kinetic and stoichiometric parameters, according to the Monod model. An in situ respirometry method was also applied to the reactor, with the injection of pulses of glucose from 19 to 97 mg L-1. The respirograms obtained were used to estimate the kinetic and stoichiometric parameters according to ASM1 and ASM3 models. No significance difference was observed between parameters estimated by chemostat and respirometric methods. The glucose affinity constant was from 0.4 to 0.7 mg L-1, the maximum specific growth rate was from varying from 0.14 to 0.20 h-1, and the growth yield was from 0.41 to 0.67. These results confirm that in situ pulse respirometry is a suitable method for kinetic and stoichiometric parameters estimation.

The objective of this work is to infer the biomass and product concentrations in a class of bioreactor, under the assumption of substrate measurements. This is done by employing a mathematical model of a continuous stirred bioreactor where the corresponding kinetic term is experimentally corroborated. The observer's structure contains, as usual, a predictive term given for nominal model plus a corrective term, which is proposed as adaptive one, where the corresponding observer's gain obey a specific dynamic equation, which provides an exponential convergence characteristic, under the Lipschitz assumption of the system. Numerical simulations show an adequate performance of the proposed observer and a comparison with a finite-time observer is done.

Performance monitoring has the main aim to analyze operating process unit efficiencies especially to lower variable costs, increase profitability, and improve the supply chain optimization in terms of enterprise resource planning (ERP), production accounting, and optimal scheduling of maintenances. Nevertheless, the industrial state of the art is based on approximated techniques for data reconciliation and performance monitoring, which make use of the plant historical database and adopt regressive methods to predict trends of future efficiencies, whereas more reliable performance monitoring techniques are needed to obtain an effective ERP. On this subject, the paper shows how the process performance monitoring is fast moving from reacting towards predicting methodologies and proposes a novel promising technique based on unsteady detailed models to tackle this issue.

Ammonia production is generally carried out in multiphase catalytic reactors. In this paper we analyze an industrial reactor for ammonia production with a dynamic heterogeneous model that accounts for transport resistance both inside and outside catalyst pellets. The work is aimed at investigating the possible appearance of periodic solutions in normal operating conditions for their relevance in terms of safety and control strategy. The analysis is based on a continuation approach to determine the bifurcational characterization of model predictions. This description is compared with that derived in a previous work (Mancusi et al. 2000), which was based on a simpler pseudo-homogeneous model of the same process.

Finding optimal operating conditions fast with a scarce budget of experimental runs is a key problem to speed up the development and scaling up of innovative bioprocesses. In this paper, a novel iterative methodology for the model-based design of dynamic experiments in modeling for optimization is developed and successfully applied to the optimization of a fed-batch bioreactor related to the production of r-interleukin-11 (rIL-11) whose DNA sequence has been cloned in an Escherichia coli strain. At each iteration, the proposed methodology resorts to a library of tendency models to increasingly bias bioreactor operating conditions towards an optimum. By selecting the ‘most informative’ tendency model in the sequel, the next dynamic experiment is defined by re-optimizing the input policy and calculating optimal sampling times. Model selection is based on minimizing an error measure which distinguishes between parametric and structural uncertainty to selectively bias data gathering towards improved operating conditions. The parametric uncertainty of tendency models is iteratively reduced using Global Sensitivity Analysis (GSA) to pinpoint which parameters are keys for estimating the objective function. Results obtained after just a few iterations are very promising.

The dynamic simulation of chemical processes is an indispensable and central tool for process design, analysis, and operation. Compared to a steady state simulation, the dynamic process simulation has some outstanding advantages, such as the assessment of control system reliability before its implementation on the real plant, the planning and testing of start-up and shutdown procedures, the training of operators, the possibility to investigate the causes of previous accidents, and to determine the consequences of possible accidents. In particular, as far as the accidents are concerned, it is not of minor importance that they are an economic liability to the company and the community (Fewtrell and Hirst, 1998). Fewtrell and Hirst highlighted also that high-cost accidents have some common features. The coupled dynamic simulation of process and accidents allows identifying the aforementioned deficiencies while moving towards a more safe and productive management of chemical processes. This manuscript analyzes the possible plant/accident interactions and suggests the simultaneous dynamic integration of process simulators and accident simulators. In fact, to simulate a realistic accident dynamics, both the dynamics of the plant and of the accidental event should be run simultaneously since they are biuniquely interrelated and they influence one another. The manuscript investigates the feedbacks and interactions between the process and the accident by means of a specific case study.

A methodology is proposed to identify both process and disturbance dynamics in order to improve the performance of control loops under PID control. The information recorded by a common DCS after a persistent perturbation has affected the plant from desired operating conditions is sufficient for the application of the procedure and no additional tests are required. The identification is based on a simplex algorithm and performs a numerical approximation; results are compared with a linear least square identification algorithm which allows an analytical solution under the assumption of equal process and disturbance dynamics. Results for a wide class of process dynamics show that the simplex approach allows a more efficient identification and permits a model based design of the controller, resulting in highly superior closed loop performance. Robustness specifications can be included in the design. The relevance of the problem and the reliability of the proposed identification and design procedure are confirmed by applications on industrial data sets. The algorithm is implemented in a performance monitoring system operating in refinery plants.

This paper is focused on the design of an in-line control system for the freeze-drying process of pharmaceuticals in a vial. The goal is to minimize the duration of the primary drying, when most of the water is removed by sublimation while maintaining the product temperature below the maximum value allowed by the product. The pressure in the drying chamber and the temperature of the fluid that is used to heat (or to cool) the product can be manipulated; the pressure rise test and the DPE (Dynamic Parameters Estimation) algorithm are used to estimate the product temperature and the residual ice fraction, i.e. the controlled variables. Both an open-loop controller, using constant values of fluid temperature and chamber pressure, and a closed-loop controller, based on the in-line optimization of the fluid temperature, will be described and tested. Finally, the possibility of manipulating in-line the pressure in the drying chamber will be discussed.

We develop a Takagi-Sugeno (TS) fuzzy model of a concentric-tubes heat exchanger. The model is structured on fuzzy logic reasoning with sets of linguistic rules describing the dynamic characteristics of the thermal system. Using a system identification technique based on adaptive neural networks and subtractive clustering, the fuzzy rules are derived from experimental data of flow rates and fluid temperatures that were previously collected in a heat exchanger test facility. The accuracy of the resulting model is assessed by comparing predictions, versus experimental measurements, of the time-dependent response of the outlet hot- and cold-water temperatures under a step-change in the mass flow rate of the cold fluid. The results indicate that the TS fuzzy model is able to estimate the behavior of the physical system with predicting errors of the order of the experimental uncertainties.

The current contribution is aimed to describe and design the dynamics of the recuperative catalytic combustor. The step responses of the combustible concentration, fuel flow rate, fuel temperature and three manipulated variables are determined for the non-controlled combustor. The time delays, time constants and gains are determined for the temperatures measured in two points in the catalytic zone and its dependencies on the process and geometrical parameters are discussed. The frequency response analysis conducted shows that in the combustor with a higher time constant medium term input disturbances are effectively attenuated. In order to handle long term impulses of lean or rich fuels, the control strategy dedicated to the current combustor is proposed. It is verified in the example in which the frequently used control algorithm is applied to simulate the process control. The results show that the improvement is attained in combustor transient operation and control.

The distributed dynamic model of a circulating fluidized bed combustor in isothermal operation is developed and proposed, consisting of four 1–D evolutionary partial differential equations and relevant boundary and initial conditions. The continuum model is then approximated by a finite–difference method to provide a "reference" solution. Then, proper orthogonal decomposition with Galerkin projection is introduced to derive a reduced order model. The POD modes are then tested in the low-order reconstruction of space profiles of the state variables at different times, during the transient and at steady-state. POD–based models prove to be effective, being able to reproduce steady–state with four basis functions. Quantitative accuracy even during the early transient stage is achieved with as little as 24 modes.