Search

This paper describes an approach for data-driven generation of structured models of complex and unknown processes by means of genetic programming. The basic approach which is used to generate and to modify symbolic model descriptions represented as block diagrams is introduced and an application for modelling of an industrial biotechnological fed-batch fermentation process is presented. [ABSTRACT FROM AUTHOR]

The development of a methodology for the simulation of structure forming processes is highly desirable. The smoothed particle hydrodynamics (SPH) approach provides a respective framework for modeling the self-structuring of complex geometries. In this paper, we describe a diffusion-controlled phase separation process based on the Cahn-Hilliard approach using the SPH method. As a novelty for SPH method, we derive an approximation for a fourth-order derivative and validate it. Since boundary conditions strongly affect the solution of the phase separation model, we apply boundary conditions at free surfaces and solid walls. The results are in good agreement with the universal power law of coarsening and physical theory. It is possible to combine the presented model with existing SPH models. [ABSTRACT FROM AUTHOR]

Mass transport in textiles is crucial. Knowledge of effective mass transport properties of textiles can be used to improve processes and applications where textiles are used. Mass transfer in knitted and woven fabrics strongly depends on the yarn used. In particular, the permeability and effective diffusion coefficient of yarns are of interest. Correlations are often used to estimate the mass transfer properties of yarns. These correlations commonly assume an ordered distribution, but here we demonstrate that an ordered distribution leads to an overestimation of mass transfer properties. We therefore address the impact of random ordering on the effective diffusivity and permeability of yarns and show that it is important to account for the random arrangement of fibers in order to predict mass transfer. To do this, Representative Volume Elements are randomly generated to represent the structure of yarns made from continuous filaments of synthetic materials. Furthermore, parallel, randomly arranged fibers with a circular cross-section are assumed. By solving the so-called cell problems on the Representative Volume Elements, transport coefficients can be calculated for given porosities. These transport coefficients, which are based on a digital reconstruction of the yarn and asymptotic homogenization, are then used to derive an improved correlation for the effective diffusivity and permeability as a function of porosity and fiber diameter. At porosities below 0.7, the predicted transport is significantly lower under the assumption of random ordering. The approach is not limited to circular fibers and may be extended to arbitrary fiber geometries.

Improving the yield and selectivity of chemical reactions is one of the challenging tasks in paving the way for a more sustainable and climate-friendly economy. For the industrially highly relevant gas-liquid reactions, this can be achieved by tailoring the timescales of mixing to the requirements of the reaction. Although this has long been known for idealized reactors and time- and space-averaged processes, considerable progress has been made recently on the influence of local mixing processes. This progress has become possible through joint research between chemists, mathematicians, and engineers. We present the reaction systems with adjustable kinetics that have been developed, which are easy to handle and analyze. We show examples of how the selectivity of competitive-consecutive reactions can be controlled via local bubble wake structures. This is demonstrated for Taylor bubbles and bubbly flows under technical conditions. Highly resolvednumerical simulations confirm the importance of the bubble wake structure for the performance of a particular chemical reaction and indicate tremendous potential for future process improvements.

Surface science methods originating from analysis of noble metal catalysts are increasingly applied to metal oxides. These methods provide direct access to fundamental structural properties and phase equilibria governing the catalytic properties of metal oxide surfaces. However, no systematic way existed so far for transferring this knowledge to technical catalysts. The aim of this paper is to combine surface science with chemical engineering methods to bridge this gap. Styrene synthesis over pure and K-doped iron oxides is used as an example to develop and to explain the methodology. Single crystal films (SCF), grown epitaxially on a Pt-carrier are considered as ideal model surfaces. Comprehensive UHV analyses yield the structural properties of SCF as well as their interaction with relevant components of the reaction mixture. Their results are combined with conversion experiments to derive a mechanistic catalyst model along with quantitative information on the reaction rates. The activity of SCF as well as their phase transitions under reactive conditions can be described with a continuum model depending on the macroscopic properties of the system. This model forms the crucial link towards technical catalysts. It is shown that the behaviour of a powder catalyst can be described as a superposition of the above kinetic model and an appropriate porous model. In this paper we review the developed methodology and conclude with the evaluation of the concept.