Improving the bed movement physics of inclined grate biomass CFD simulations

Highlights

• Particulate bed dynamics are implemented in a Eulerian CFD formulation.

• The movements of the bed include particle crumbling and thrust feeding.

• Model results are tested in an experimental grate.

• The model presents a good behaviour to reproduce the bed shape and slope in the grate.

Abstract

Extensive research has been performed in the field of biomass combustion, particularly in terms of numerical simulations. However, only a few studies have deepened the understanding of the dynamics of packed beds and boiler grates.

This paper presents and assesses a new physical solution under a comprehensive Eulerian model that tackles the complex problem of biomass movements inside a packed boiler bed. In addition, this paper improves the precision of the obtained results, expanding the application cases of the simulation model produced by the GTE research group at the University of Vigo (Spain). The modelling algorithms are based on mechanisms that represent the crumbling of biomass particles due to gravity and the movements caused by the push generated by the feeding system. The movements’ logic affects the variables involved in the biomass thermal conversion formulation, which is integrated with other combustion models previously presented by the authors.

To test the performance of the presented models, a 3D experimental grate was printed and fed with biomass pellets. The performance was compared to that of the evolution of the simulated bed. Several experiments were performed to analyse parameters, such as the bed size, slope and residence time of different fuel batches. The results of the comparison show agreement in the behaviour of the model in terms of the bed size, shape and slope for different instances of the feeding process. Otherwise, the experiments show that the internal distribution of the bed layers of different batches presents some differences with the predictions. Therefore, possible improvements, such as considering a friction factor, may be applied to the model to upgrade its performance.

Introduction

Currently, the use of resources is a strategic factor at every social level. Improving the efficiency of all processes in which energy is involved is essential to the consumption of energy resources due to the increasingly pressing shortage of natural resources for production [1]. Among the natural resources involved in obtaining energy, biomass is an essential material. Biomass represents between 10% and 14% of the global energy supply, only after coal, oil and natural gas [2], [3]. Therefore, biomass is a very important source of energy worldwide, especially in rural areas or in underdeveloped countries [4]. Beyond being a source of energy, biomass has become important in current contexts for the following three reasons: it has a relatively low operating cost, it is an abundant source, and it is closely related to processes of reusing, recycling and reducing waste. These characteristics are considered to be green energy solutions with special relevance in the near future [5], [6]. Therefore, biomass is directly related to the processes of recycling, reusing materials and reducing pollution [2].

Biomass is abundant because it can be obtained from multiple sources, such as agriculture, forestry, by-products, residues and wastes, even algae, etc. [7]. The methods of obtaining biomass depend on the type of the collection of waste produced by industries or individuals, where reduce and reuse are methods to reducing pollution and impacts related to obtaining new resources. Reuse is especially important because predictions for energy consumption are estimated to increase by 40% by 2040, with finite natural resources [4]. The combination of waste reduction and energy generation adds value to biomass as an energy source in the current context.

Most government actions and international institutions, together with the environmental trend of civil society (especially in industrialized states), will inexorably result in profound changes in the energy model in the near future. This gradual but constant and unstoppable evolution of our energy model focuses on the reuse of resources, the reduction of environmental impacts (emissions, occupation of natural spaces, effects on ecosystems, etc.) and the change in sources of energy and materials used. These solutions seek an option that minimizes damage to nature and improves the efficiency of the energy obtaining processes [5]. Therefore, the importance of biomass as a source of energy is based on two factors. First, it is a source with high availability and a relatively low operating cost. Second, biomass is a strong alternative from an ecological point of view because it is closely related to the use of waste, beyond which its emissions are lower or have a minor impact on environmental sources (biomass is relatively neutral in terms of CO₂) compared to other energy sources [2], [8].

However, biomass combustion presents some inherent problems. These issues include slagging in the beds, fouling of the main thermal-exchange surfaces or releasing particle emissions [9], [10], [11], [12], [13]. In addition, the heterogeneous biomass physical-chemical properties. The diversity of its origin (organic waste, pressed sawdust of different wood materials, shapes and sizes, pine leaves, residues, logs, etc.) and the diversity of the processes to which it is subjected beforehand make their parameterization very difficult [14], [15], [16], [17], [18], [19], [20].

The heterogeneity of biomass generates very different situations. For example, clean woody biomass such as sawdust pellets and chips are high-quality fuels that can generally be used with high efficiency and minor problems in current combustion systems; however, the availability of this biomass and its costs make this resource economically unappealing, providing only marginal profits. Otherwise, straw, agricultural and forest residues, wastes, etc. have a much higher availability and lower costs, but they are low-quality fuels that introduce several technical challenges, some of which are intrinsically linked to their behaviour in the bed and their local temperature (slagging, channelling, elutriation, voids, etc.).

Computational simulations in industrial processes have led to reductions in the cost of product and process development, optimization activities (by reducing down time), the need for physical experimentation, and the time to market. The computational models contribute to other benefits, such as improving design reliability, increasing conversions and yields, and facilitating the resolution of environmental, health, and right-to-operate issues [21].

In this sense, a computational approach to combustion processes can contribute to improving the engineering methods for the design of boilers. Issues such as the efficiency of boilers, the reduction in polluting emissions or the reduction in manufacturing costs are closely related to the improvement of simulation systems and their inclusion in industrial processes.

In the field of computational simulation, experimental values are usually used as data to support the development and validation of algorithms that simulate physical or chemical processes. These experimental data are often based on averaged values ​​of fluctuating signals [22]. Thus, these data do not faithfully represent actual situations and can lead to deviations between the simulations and reality. The union between the computational and the experimental worlds allows us to detect areas of uncertainty in the investigations that lead us to improve both the simulation algorithms and the quality of the experimental data that are collected. As small as it may seem, each mechanism can affect the overall resolution and the experimental data accuracy depends on the simulation models. Better simulation models can be developed with better experimental data.

The GTE research group at the University of Vigo (Spain) has been developing a biomass thermal conversion comprehensive model since 2005; this model has been built over time with the contributions of several researchers with different technical backgrounds [23], [24], [25], [26], [27]. The work presented here is framed within this project and focuses on the development of algorithms that simulate the physics of packed beds during their thermal reaction. As a central element, the model uses an Eulerian approach in opposition to other previously published strategies that focus more on particle tracking approaches, such as the Discrete Elements Method (DEM) [28], [29], [30].

The Eulerian approach is computationally less expensive since it does not track individual particles. This framework employs the mean numerical values that represent the space-averaged values of the particles inside each cell of the grid. However, this makes the management of the physics of the bed such as feeding, fuel distribution or local movements due to compaction much more complex because the density of the information is controlled by the size of the grid and not by the number of particles in it. This issue has been addressed in detail in a previous publication [31], as well as in other studies, where some studies were performed with the Eulerian approach with one-dimensional (walking column) [32], [33], [34] and two-dimensional or three-dimensional models; specifically highlighting the work by Y. B. Yang et al. [35] and S. Hermanson et al. [36]. Although these models present very interesting results in relation to the simulation of biomass combustion, they are usually focused on the biomass thermal conversion but not on the particle motion physics.

As multitude of biomass combustion systems are based on inclined-grate packed beds, numerous attempts to implement this beds in CFD simulations have been performed. The main difficulty is to represent the advance of fuel along the grate during the combustion process. Some models used a one-dimensional approach known as “walking column” in which each step of combustion is represented by a column that is consumed while advances along the grate. This method was broadly used to simulate straw and waste combustion [37], [34], [38]. A two-dimensional model was developed by Y. B. Yang et al. [38], [39] through an Eulerian model based on the transport equations of biomass components in a static bed. B. Miljković et al. [40] presented a more advanced two-dimensional Eulerian model for straw or other porous fuels that is adaptable to multiple grate geometries. A totally different approach was used by B. Peters [41], who considers the bed as a set of individual discrete particles by modelling thermal conversion and particle interaction. A similar method was applied by F. Wissing et al. [42] to a 3D boiler through a DEM/CFD implementation which calculates the DEM trajectories an interaction between particles and the results of thermal conversion are introduced in the CFD domain. These models require an important computational effort to compute the particle motion and contact mechanics. R. Mehrabian et al. [43] applied a thermally thick treatment to the bed particles to simulate a 3D circular grate. These previous commented models calculate the packed bed though external variables and results are introduced to the computational domain as boundary conditions or source terms. Karim and Naser [44] implemented a bed model into the computational domain through Eulerian variables. These variables are moved along the grate through lateral and crumbling movements.

Focusing on the previous work of the GTE research group, M. A. Gómez et al. [45] developed a basic crumbling algorithm adapted to different domestic boiler configurations. This algorithm responded adequately to the needs of this specific type of boiler but presented some difficulties when applying it to other types of grates, essentially to industrial grates. In addition to the crumbling algorithm, a fuel feeding model was presented to feed certain types of packed beds through feeding screws [26]. However, this method presents some complex requirements that are difficult to implement. Consequently, a high-viscosity fluid pre-simulation was performed to pre-establish the trajectories described by the fuel particles in the bed. Then, these trajectories were used to generate an advective flux that described the fuel movement produced by the feeding system. Although reasonable solutions were obtained with this method, in some cases, it required certain flow characteristic assumptions for the fluids, which were not identical to the behaviour of the solid. Furthermore, there was a main limitation in this approach when applying it to grates that were not encased by walls; the viscous liquid spread in a different way than the solid particles. Although it has already been mentioned above, a similar algorithm was successfully implemented by Karim and Naser [44], [46] in the CFD software AVL Fire. These publications show good results that include images of the bed evolution.

With the main objective of expanding the model application to a greater number of biomass boiler technologies, especially taking the leap from domestic boilers in the kW range [9], [45], [47], [48] to industrial boilers in the MW range [49], a completely new algorithm was developed. This algorithm must operate on different types of grate, including inclined grates, this type of grate are very common in industrial designs. Based on the previous work of M. A. Gómez et al. [47] but under new operating logic, the new proposed model can replicate the fuel behaviour regardless of the boiler grate morphology. However, this newly developed methodology requires its main parameters to be tuned by experimental validation. The work presented here goes one step forward and proposes a model that connects the physics involved in the fuel-bed dynamics with its numerical implementations. The model is mainly based on the pushing effect of the feeding system on the packed bed and the crumbling of the particles along the grate by the gravity force. After presenting the model (Section 2), it will be applied to inclined grates (Section 3.1) and will be compared with the experimental results obtained in a 3D full-scale boiler mock-up (Section 3.2) to validate its implementation.