Norberto Fueyo

Modelling the Temporal Evolution of a Reduced Combustion Chemical System With an Artificial Neural Network

The present work introduces a way of embedding a combustion chemical system in a neural network, in such a way that it can be used, with considerable CPU time and RAM memory savings, in fluid-flow-simulation codes. The system is composed of four neural networks, with three of them simulating the evolution of the reactive species and one providing density and temperature as a function of composition. The performance in terms of accuracy of the networks is assessed by comparison with the results of the direct integration of the thermochemical system for a large number of random samples. Error measurements are reported, and sample evolutions of the chemical system with both methods are compared. It can be summarized that the results of this exercise are satisfactory, and the CPU-time and memory savings encouraging.

Combustion characteristics of heavy oil-water emulsions

The combustion of heavy oil and its emulsions with water was investigated in experiments on a semi-industrial scale. Two comparisons between heavy oil and oil-water emulsion flames are presented that, due to the different initial conditions of the spray, provide complementary information. Reported results include spatial distributions in the flame of temperature and species concentrations (O2, CO, UHC, NOx) as well as gaseous and solid emissions in the flue gases. The measurements inside the emulsion flame display a remarkable improvement in the combustion process with respect to that of the neat oil with poor atomization; differences are much less important if a fine spray is achieved with the heavy oil. Solid emissions are significantly reduced in the emulsion tests and the morphology of the particle samples demonstrates the fragmentation of the drops and/or the coke particles initially formed. The flame temperatures are reduced by ∼65 K. The heat absorbed by the water injected in the emulsion and enhanced radiative heat transfer due to the higher particle number density could explain this difference. The spatial distribution of NOx indicates that a significant reduction is obtained in the final part of the flame; this may be attributed to a decrease in the rate of thermal-NO formation as a consequence of lower gas temperatures. No measurable difference in NOx concentration is found in the inner core of the flames.

An economical strategy for storage of chemical kinetics: Fitting in situ adaptive tabulation with artificial neural networks

Reducing the computational time of chemical kinetics is essential for implementation of realistic chemistry into large-scale numerical simulations. Among the storage-based techniques, the in situ adaptive tabulation (ISAT) method features storing and retrieving data during the simulation: therefore, only the needed data are stored. As ISAT is based on linear approximation, the required storage can grow rapidly when a wide range of chemical states is involved, such as occurs in turbulent flames. An economical strategy for storing chemical kinetics data is proposed here by fitting results obtained from ISAT with artificial neural networks (ANN). This concept is explored in this study using a partially stirred reactor (PaSR) with two reduced chemical mechanisms of 9 and 17 reactive scalars. The performance of the ANN fitting is assessed on the basis of accuracy, memory, and CPU time. Test results based on PaSR demonstrate that a significant saving in memory can be realized with the ANN. Both the accuracy and CPU time with the ANN are found comparable with those of ISAT, suggesting great promise for use of ANN in large-scale computations.

Detailed modelling of a flue-gas desulfurisation plant

Abstract This paper presents a CFD model for a flue-gas desulfurisation plant, and its application to an operating plant. The FGD plant is of the wet-scrubber type, with co-current and counter-current sections. The sorbent used is limestone, and, after cleaning the flue gases, the limestone slurry is collected in an oxidation tank for the production of gypsum. The model uses an Eulerian–Eulerian treatment of the multiphase flow in the absorber and the tank. The essential mass-transfer mechanisms (such as SO2 and O2 absorption and CO2 desorption) are accounted for, as are also the main chemical kinetics leading to the formation of gypsum. Given the different nature of the flow in the absorber and tank, two separate simulations are conducted for each of these domains, and the solutions are iteratively coupled through boundary conditions during the calculations. The model is applied to the FGD plant of the Teruel powerstation located in Andorra (Teruel, Spain). The powerstation is fired with a high-sulfur coal (up to 4.5 percent), and the FGD system has been designed for a desulfurisation capacity of 1.4 million N m3/hr for a desulfurisation efficiency in excess of 90 percent. Validation of the model is conducted by comparison with available plant data for two design coals and two desulfurisation efficiencies. The model accuracy is reasonable, given the complexity of the aero/hydrodynamical and thermo-chemical phenomena involved.

Computational Evaluation of Low NOx Operating Conditions in Arch-Fired Boilers

In the present paper, a computational model is used to simulate the aero-dynamic, thermal, and chemical conditions inside an arch-fired coal boiler. The model is based on the Eulerian-Eulerian concept, in which Eulerian conservation equations are solved both for the gas and the particulate phases. A NO{sub x} formation and destruction submodel is used to calculate the local concentration of NO. The model is used to simulate a range of operating conditions in an actual, 350 MW, arch-fired boiler, with the aim of reducing, using primary measures, the emissions of NO{sub x}. The model results shed some light on the relevant NO{sub 2}-formation mechanisms under the several operating conditions. Furthermore, they correlate well quantitatively with the available field measurements at the plant, and reproduce satisfactorily the tendencies observed under the different operating modes.

STATISTICAL DESCRIPTION OF THE TURBULENT MIXING OF SCALAR FIELDS

A formulation in terms of probability density function (PDF) transport equations is presented for inert and reactive scalar fields undergoing turbulen mixing. The PDF methodology is related to the classical moment equations. The hierarchy of PDF transport equations resembles the BBGKY equations in statistical mechanics. Closure hypothesis, approximating the molecular mixing term, are described and their predictions for simple systems are compared with direct numerical simulations (DNS). Solution algorithms in terms of Monte Carlo particles are also discussed.

A self-organizing-map approach to chemistry representation in combustion applications

Several alternative techniques have been proposed in the literature in order to avoid the CPU-intensive numerical integration of the thermochemical equations in the simulation of combustion processes. The present paper introduces a new approach, which is based on two artificial neural-network (ANN) paradigms, namely the self-organizing map (SOM) and the multilayer perceptron (MLP). The SOM is first employed for the automatic partitioning of the thermochemical space into subdomains. Then, a specialized MLP is trained in order to fit the thermochemical points belonging to a given subdomain. The presented strategy is tested on a partially stirred reactor (PaSR) with a reduced methane-air mechanism, and encouraging results are reported. The relatively modest CPU-time and memory requirements of the method make the SOM-MLP approach a promising technique for the inclusion of large chemical mechanisms in the context of complex applications, such as the multidimensional simulation of combustion.

A single-step time-integrator of a methane–air chemical system using artificial neural networks

Abstract The present paper reports a novel method for embedding a reduced chemical system, suitable for the simulation of methane–air combustion, in an artificial neural network (ANN). The use of ANNs as a means of storing in a compact manner the chemical kinetics of a system is an emerging alternative to other methods, the full potential of which remains to be exploited. The current contribution introduces two novelties: firstly, the compositional domain is split into subdomains, for each of which an ANN fitting is attempted; and secondly, the timestep is introduced as an additional input to the network, thus increasing the accuracy and speed of the method. The paper introduces three alternative types of network, and describes in detail the methodology used for their construction and validation, as well as the validation results. The level of accuracy attained is at least one order of magnitude better than with previously-published ANN approaches.

An efficient particle-locating algorithm for application in arbitrary 2D and 3D grids

Abstract Several alternative techniques have been proposed in the literature to efficiently locate particles within unstructured grids. The present paper reviews two recently published, particle-locating algorithms, and introduces a new approach which improves the performance of the previous methods. The proposed algorithm is valid for arbitrary two-dimensional (2D) or three-dimensional (3D) grids, it is simple to implement, and results in fairly small CPU-time requirements. Furthermore, the directed-search feature of the present method offers shorter search paths and allows the detection of walls during the tracking of the particle trajectory. The performance of the proposed particle-locating strategy is compared with existing ones, and is evaluated on two tests, namely a 2D/3D random particle-locating test and a 2D Lagrangian simulation of a premixed turbulent flame.

An open-source library for the numerical modeling of mass-transfer in solid oxide fuel cells ☆

The generation of direct current electricity using Solid Oxidize Fuel Cells (SOFCs) involves several interplaying transport phenomena. Their simulation is crucial for the design and optimization of reliable and competitive equipment, and for the eventual market deployment of this technology. An open-source library for the computational modeling of mass-transport phenomena in SOFCs is presented in this article. It includes several multicomponent mass-transport models (ie Fickian, Stefan-Maxwell and Dusty Gas Model), which can be applied both within porous media and in porosity-free domains, and several diffusivity models for gases. The library has been developed for its use with OpenFOAM(R), a widespread open-source code for fluid and continuum mechanics. The library can be used to model any fluid flow configuration involving multi-component transport phenomena and it is validated in this paper against the analytical solution of one-dimensional test cases. In addition, it is applied for the simulation of a real SOFC and further validated using experimental data