Packed bed combustors burn fairly large solid fuel particles within confining walls, with air supplied from below the grate. As combustion occurs and particles are consumed, fresh particles are fed onto the bed so the level is kept roughly constant. Packed bed combustion is used for wood and biomass combustion in small-scale power plants, wood waste combustion in pulp and paper plants, and trash incineration. The structure of a packed bed is very important to the combustion process and can be defined by particle shape and size, sphericity, particle overlap (decreasing area availability) and chiefly by void fraction. Void fraction has already been proven of great influence in packed beds – it is raised to the third power in the pressure loss equation, and it can also affect heat and mass transfer and surface reaction rates. This thesis presents results of several experimental combustion tests that were performed in a packed bed combustor, using commercial spruce lumber particles of parallelepipedal geometry as fuel. At the end of each test the bed contents were removed, taking care to preserve their structure, and fixed with liquefied wax. The solidified bed was then cut into circular cross sections at different heights of the bed, and photographs of the cross sections were taken so the local void fraction could be estimated using image analysis. The bed sampling led to the discovery that, surprisingly, the actual bulk void fraction in the combustor, which is the average of local void fraction measurements, is less than that of the unburnt particles, varying from 19% to 30% in decrease in void fraction depending on the particle type used. Local measurements allowed the development of an empirical linear equation model to represent the variation of void fraction with height above the grate. Each combustion test had measurements of gas volume fractions and temperatures at different heights above the bed grate to be compared with the results of a numerical model simulation. The numerical model used in this work is an existing numerical model of all the relevant processes in packed bed combustion. Previously, the numerical model had assumed the void fraction to be constant and equal to that of the unburnt fuel, since no information on local variation was available, and the packing geometry remained self-similar as particles are consumed. Three models for void fraction were then compared in the combustion model: a constant void equal to that of the unburnt particles, the empirical linear fit of void fraction with height, and a constant void equal to the measured bulk void fraction. Maximum temperatures were higher using the unburnt fuel void fraction because of a thicker oxidation zone, whereas the void fraction model iii based on experiments generated a thicker reduction zone and therefore higher CO concentrations. CO concentrations were experimentally measured and agreed quite well with the CO concentration from the model. Local void fraction differences had the most impact in the diffusion-controlled zone, as shown by comparing the empirical void model and the measured bulk void fraction. How lowering the void fraction can increase gas velocities, heat and mass transfer coefficients, and burning rates is also discussed in this work.

This thesis constitutes fundamental experimental and computational investigations on gasification and combustion in a packed bed of biomass. Packed bed gasification-combustion in counter-current mode is used in two applications -(1) Gasifier stove in reverse downdraft mode (or equivalently, top-lit updraft mode) that constitutes the idea behind efficient and clean burning domestic stoves. (2) Combustion-on moving grate for boiler application, studied widely in Europe. While a large part of the present study is around domestic stoves, a crucial part of the study aims to address the second application as an extension of the approach taken in the first part to clarify conflicting conclusions of earlier studies and explain the aero-thermochemical behavior over the entire range of superficial velocities, V s (this is velocity of air through the empty cross section of the reactor). Operational differences between the two applications lie in the range of superficial velocity -3.5 to 6 cm/s for domestic stoves and 15 to 30 cm/s for grate combustion. Lower values of Vs are chosen for domestic stoves to limit the particulate emissions; higher values of V s for combustion-on-grate to maximize the conversion rate. Present work deals with a fan based gasifier stove, Oorja, built by BP, India (currently transferred to FEPL, Pune) and disseminated to over 400,000 households between 2005 and 2009. The technology was developed at CGPL, IISc and transferred to BP for commercialization. Work reported in this thesis was started to resolve issues of higher CO emissions in char mode operation and occasional smoking during transition from flaming to char mode. The contribution of the thesis is split into two parts. (a) Use of the principles of gasification to improve the performance of the stoves to the highest possible level, balancing between efficiency and ash fusion issues for domestic and industrial applications and (b) fundamental studies to unravel the flame structure in the two-phase gasification-combustion process over the entire range of Vs. Improving the stove performance It has been known that in most free-convection based stoves, like three stone fire and others developed over the last two decades, the amount of energy extracted from the stove by a cooking pot, usually measured as water boiling efficiency, is between 15 to 35 % with CO emissions of more than 1.5 g/MJ. Oorja stove had demonstrated water boiling efficiency of 50 % and CO emissions of 0.75 g/MJ. Operational issues noticed in the field provided an opportunity to further improve the performance by conducting detailed thermo-chemical studies. Towards this, the components of water boiling efficiency in different phases and from different modes of heat transfer were determined. Optimizing the ratio of air flow rate between combustion air from top and gasification air through the grate (denoted by R) was the key to improving the performance. The maximum water boiling efficiency obtained was 62% with 0.53 g/MJ CO for a 320 mm diameter vessel; under these conditions, the first phase, termed flaming mode, involving pyrolysis-gasification-gas phase combustion contributed 45% to the total efficiency and 0.4 g/MJ CO at R = 4.8 and the second phase, termed char mode, involving char surface oxidation-gasification-gas phase combustion contributed 17% and 0.13 g/MJ CO at R = 1.9. Under optimal air flow conditions, efficiency depends on the size of the vessel used; reactive flow calculations were performed with fast chemistry (using mixture fraction approach) in a zone that includes the free space of the combustion chamber and the vessel to obtain the heat transfer efficiency and bring out the effect of vessel size. Experiments aimed at evaluating the performance of the stove on either side of stoichiometry, revealed that while the stove could be operated on the rich side, it was not possible to operate it on the lean side -it was always tending towards the stoichiometric point with enhanced power. Computational studies showed that increased air flow from the top caused enhanced recirculation around the fuel bed bringing more oxygen that reacted closer to the surface and transferred additional heat enhancing the pyrolysis rate, explaining the observed shift towards stoichiometry. An examination of literature showed that the energy balance for stoves had long remained unexplained (unaccounted losses in stoves were up to 40 %). Using the different components of efficiency obtained from experiments and computations, a heat balance was established to within 5%. This vast improvement in the heat balance is due to the fact that the unaccounted loss in the earlier estimates was essentially due to poor combustion, but was not so recognized. The very significant increase in combustion efficiency in this class of stoves allowed the possibility of estimating other components reasonably accurately. This is a direct consequence of the two stage gasification-combustion process yielding steady flow of gases which contain 80% (gasification efficiency) of the input energy enabling near-stoichiometric combustion with the help of controlled supply of combustion air. Fundamental studies Experiments with wood chips (615 kg/m3) and pellets (1260 kg/m3) showed that particle density has no effect on single particle and packed bed combustion in flaming mode beyond the role played through the surface energy balance (involving the product of fuel density and propagation rate, ˙r). Same is true for single char particles. A transport controlled combustion model taking into account the ash build up over the char surface confirmed this behaviour and showed that the phenomenon follows d2 law, where d is the equivalent diameter of the fuel particle, consistent with the experimental results. But packed bed of char particles showed distinct dependence on particle density. Differences were traced to poor thermal environment faced by low density wood char pieces compared to pellet char leading to variations in the volumetric heat release rate. A composite picture of the operational behaviour of the packed bed flame propagation was obtained from the measurements of exit gas composition, bed temperature, temperature of gas phase and condensed phase surface using 100 µm thermocouples, O 2 drop across the flame front using lambda sensor as a function of Vs. The packed bed studies were conducted in insulated steel and glass reactors. These studies clearly showed distinctive regimes in the bed behavior. In the first regime from Vs = 3 to 17 cm/s, (a) the propagation rate increases with Vs, (b) the fractions of CO, H2 are at least 10%, CH4 drops from 3 to 1%, (c) the oxygen fraction is near zero, (d) the gas phase temperature in the bed is constant at about 1600 K, (e) the condensed phase surface temperature increase from 850 K to 1600 K and (f) oxygen fraction drops from 0.21 to 0.0 within a single particle depth and coincides with the gas phase ignition. The inferences drawn from these data are that (i) the process is diffuusion controlled (ii) the bed operates in fuel rich mode, (iii) char participates only in reduction reactions. In the second domain from V s = 17 cm/s up to about 50 cm/s, (a) the propagation rate is nearly constant (b) the mass fractions of CO and H2 drops to less than 5%, CH4 decreases further, (c) oxygen fraction remains near zero, (d) CO 2 increases, (e) gas phase and surface temperatures are nearly equal and increase from 1600 K to 2200 K and match with the equilibrium temperature at that equivalence ratio, (f) oxygen fraction drops from 0.21 to 0 in one particle depth like in the first regime indicating diffuusion limitedness in this regime as well, (g) unlike in the first regime, volatiles from biomass are convected up to the next layer suppressing a local flame and char oxidation dominates. Beyond Vs = 50 cm/s, the propagation ceased to occur. The precise value of the extinction V s depended on the rate of increase of Vs in this range. A faster change initiated the extinction earlier. Observations showed that extinction began at some location around the periphery and spread laterally. Extinction at one layer was adequate to complete the extinction process. To explain the observed behaviour a simple zero-dimensional model tracking the heating of a fresh biomass particle upstream of the propagating flame front because of radiative heat transfer was set up. This equation was coupled with the equation for single particle flaming combustion to explain the behavior in the first regime. In order to explain the observed flattening of propagation rate in the second regime, it was found essential to account for the effect of the ash layer building on the oxidizing char particle and the temperature dependence of ash emissivity, on the radiative heat transfer to fresh biomass. The results of the model coupled with the experimental data from all sources on a corrected propagation rate vs. V s showed a universal behaviour that is considered a very important recognition of the packed bed propagation behaviour. Combining theory and experiments was essential to explain the extinction. The features are: (a) the presence of ash layer over the surface is shown to be responsible for maintaining a steady char conversion in a single particle at low stream speeds, (b) the feature that the ash layer would be blown away at stream velocities of 2.5 to 3 m/s in a single particle combustion, (c) with V s close to 50 cm/s, local velocities of air flow through the bed can reach 2 to 3 m/s, this value being sensitive to the bed arrangement (with slight shifting or settling of one particle leading to increase of the local velocity at the periphery). Thus, the high local speeds of flow through the bed (more than 2 m/s) was considered responsible for removal of ash layer such that radiation losses would be dominant and cause local extinction of the reaction front at the char surface. Thus, this study has led to a comprehensive understanding of the gasification-combustion behavior of packed bed in stoves and on grates. It has also led to the evolution of parameters for obtaining high efficiency and low emissions (HELE) from stoves -both domestic and industrial. Most interestingly, it has resulted in recognition of an universal behavior of flame propagation rate through packed bed of biomass.