Natércia C.P. Fernandes

Modeling the effect of mixing in biodiesel production

The transesterification reaction models available in the literature are valid only for one particular mixing condition. In this work, a modeling strategy is presented in order to predict the effect of mixing conditions in the transesterification process. The proposed methodology was applied to independent sets of experimental data available in the literature that show the dependency of the transesterification reaction on the frequency of rotation of the stirrer. The accuracy of the developed models corroborates the validity of the proposed modeling approach.

Stiction Detection and Quantification as an Application of Optimization

Stiction is a major problematic phenomenon affecting industrial control valves. An approach for detection and quantification of valve stiction using an one-stage optimization technique is proposed. A Hammerstein Model that comprises a complete stiction model and a process model is identified from industrial process data. Some difficulties in the identification approach are pointed out and strategies to overcome them are suggested, namely the smoothing of discontinuity points. A simulation study demonstrates the application of the proposed technique.

Simulation of the transient behaviour of a digester used in the pulp and paper industry

Abstract The development of mathematical models that describe industrial processes is playing an increasing role in industrial context, because such models can replace, in some cases, experimental simulation, in a more inexpensive and flexible way. In this paper is presented a preliminary study of a transient model of a continuous moving bed reactor - the digester - used in the pulp and paper industry. In this complex and heterogeneous digester a moving bed of wood chips, containing cellulose, hemicellulose and lignin, reacts with sodium hydroxide and sodium sulfide - dissolved in a liquid phase - to remove the lignin from the cellulose fibers. The mathematical model studied here has been developed by Fernandes in a forthcoming Ph.D Thesis. The model is derived from the fundamental principles of mass and energy balance and is represented by a system of 15 partial differential equations (PDE) of convection-reaction type. The numerical methods used in the discretization of this PDE system are based on operator splitting which essentially consists in considering separately convection and reaction phenomena. This approach allows the use of methods with different properties - explicitness, implicitness and order - for each subprocess. The final global class of methods represents in some sense a “patching of tailored methods” well adapted to the feature of individual phenomena. Numerical simulations of the concentrations of organics and inorganics are presented.

Parameter estimation of a pulp digester model with derivative-free optimization strategies

The work concerns the parameter estimation in the context of the mechanistic modelling of a pulp digester. The problem is cast as a box bounded nonlinear global optimization problem in order to minimize the mismatch between the model outputs with the experimental data observed at a real pulp and paper plant. MCSFilter and Simulated Annealing global optimization methods were used to solve the optimization problem. While the former took longer to converge to the global minimum, the latter terminated faster at a significantly higher value of the objective function and, thus, failed to find the global solution.The work concerns the parameter estimation in the context of the mechanistic modelling of a pulp digester. The problem is cast as a box bounded nonlinear global optimization problem in order to minimize the mismatch between the model outputs with the experimental data observed at a real pulp and paper plant. MCSFilter and Simulated Annealing global optimization methods were used to solve the optimization problem. While the former took longer to converge to the global minimum, the latter terminated faster at a significantly higher value of the objective function and, thus, failed to find the global solution.

Simulation and Advanced Control of the Continuous Biodiesel Production Process

The biodiesel industry is characterized by high fluctuations of the prices and a multiplicity of biological raw material sources. On the other hand, there exist strict quality standards imposed on the final product. Because of these factors, it is important for biodiesel plants to run their processes in the most efficient manner in order to stay competitive. One of the ways to achieve this is the use of model based approaches for design, operation, and control. In this work, that focuses on the latter two areas, a first-principle dynamic model of the main units of a biodiesel plant is developed and applied in two situations: for open-loop simulation as well as for process optimization. The former demonstrates the response observed in the process variables when the plant is subjected to a series of disturbances in the input variables. The later is built in the context of nonlinear model predictive control that determines the optimal profiles of the manipulated variables taking into account process and quality constraints as well as the associated reactant and energy costs.

Development of a Numerically Efficient Biodiesel Decanter Simulator

This chapter deals with the modelling, simulation, and control of a separator unit used in the biodiesel industry. While mechanistic modelling provides an accurate way to describe the system dynamics, it is an iterative and computationally burdensome process that arises from the need to determine the liquid-liquid equilibria via the flash calculation. These disadvantages would preclude the use of mechanistic models for process optimization or model based control. In order to overcome this problem, an alternative strategy is here suggested. It consists of maintaining the mechanistic model structure and to approximate the iterative calculations with an artificial neural network. The general approach for dataset consideration and neural network training and validation are presented. The quality of the resulting neural network is demonstrated to be high while the computation burden is significantly reduced. Finally, the obtained grey-box model is used in order to carry out dynamic simulation and control tests of the unit.

System identification as an application of optimization

The work concerns the system identification of industrial processes via the Sequential Quadratic Programming algorithm. The proposed approach, testing scenarios, and the system identification results are discussed. The tool is tested with two datasets, the first one collected in loco from an industrial process and the second one generated with a plant simulator of a continuous stirred tank reactor, a system widely used in industry. In both cases, the resulting models capture well the process dynamics.

Modeling intra-particle convection in heterogeneous non-catalytic reacting systems

Abstract A dynamic mechanistic model of a multi-phase heterogeneous batch reactor processing a non-catalytic fluid–solid reaction is developed and numerically solved. The model takes into account both intra-particular convection and diffusion phenomena, as well as chemical reaction between the fluid and the solid. By addressing the until now ignored case of convection originated by the continuous increase of the particle porosity, this work enables a more realistic modeling of many non-catalytic fluid–solid reaction systems. The dependency of the convective term on the chemical reaction rates gives rise to a space and time dependency of the intra-particle velocity thus increasing the mathematical complexity of the problem. The model is characterized by a set of PDEs coupled with adequate boundary conditions that are solved by the method of lines. An example of application of great importance to the chemical pulping of wood is used, emphasizing the differences with the typical situation where only diffusion is considered.