Modelling fuel combustion in iron ore sintering

Abstract

In an iron ore sintering bed, the combustion behaviour of coke particles together with velocity of the flowing gas stream determines the temperature, width and speed of the traversing flame front. A bed heat treatment mathematical model was formulated in an earlier study to describe this complex relationship. An area of improvement in the model is the description of the coke combustion process, which is highly dependent on the resistances controlling the flow of gases to and from the coke particles. These vary for different coke particles because of the prior coarsening of the sinter mix by granulation. The characteristic structure of granules – nuclear particle with an adhering fines layer – indicates that gases have better access to finer coke particles. In this study, an available granulation model is integrated into the heat treatment model to provide a novel description of coke positioning within granules. In addition to this change, two endothermic reactions were introduced into the model. Using the previous and modified models, predicted bed temperature–time profiles as a function of position down the bed, were compared against embedded thermocouples results from seventeen laboratory sinter tests. Generally, the modified definition of coke combustion behaviour resulted in improved comparison with experimental results. In the sintering literature, studies have been reported on: the use of charcoal/biomass char to replace coke, the preferential placement of coke particles on the outside of granules, and varying the size distribution of the coke particles. Improving the access of gases to coke particles and decreasing coke size are comparable to using more reactive fuels. Combustion rate, efficiency and flame front properties are all influenced by fuel reactivity. Model predictions of changes in bed temperatures, flame front properties and sintering performance caused by fuel type, location and size are consistent with reported observations.

Introduction

Iron ore sintering is used all over the world to prepare agglomerates for the ironmaking blast furnace [1], [2]. Figure 1 shows a schematic diagram of the major processes in a typical iron ore sintering plant. The blended sinter mix – typically composed of iron ores, fluxes, returned sinter fines, plant dust and about 4 wt.% coke breeze – is first granulated to coarsen its size distribution. Typically, the particles in a sinter mix would have sizes ranging from 0.0 to 9.0 mm. As water is added to the cascading mix in the granulating drum, the fines (typically smaller than 0.25 mm) adhere onto the surfaces of the coarser particles (typically greater than 1 mm). The granulated sinter mix is then charged onto the moving strand (in the region of around 3 m/min) via a roll feeder to form a bed of up to 0.8 m in height. As the bed travels under an ignition hood, coke particles on the upper surface are set alight to generate a narrow flame front. Air is continuously drawn through the bed by large fans and this causes the flame front to descend down the bed. The speed of the strand is adjusted so that the flame front reaches near the bottom of the bed close to its discharge. Sinter particles, suitable for blast furnace use, are released when the sintered block disintegrates on crushing.

In sintering, coke combustion supplies around 80% of the heat required to generate sufficient melt to form a strong sinter. As coke accounts for only around 4 wt.% of the total sinter mix, it is not surprising that small changes in its properties (e.g., size and ash value) can have a large effect on combustion behaviour, heat generation and sintering performance. Figure 2 summarises the upstream processes that influence coke combustion behaviour. The same figure also shows the downstream effect of coke combustion on the sintering process. The granulation process, in addition to influencing coke location and access to oxygen during combustion, will determine the structural properties of the bed formed for sintering. At a fixed suction, properties such as bed porosity and height will determine airflow rate through the bed and, consequently, the coke combustion process and flame front speed. If sufficient fan capacity is available, airflow rate can be increased through increasing suction. For a fixed airflow rate, bulk density changes – for example through increasing the level of porous ore and granule size – will influence heat transfer down the bed and flame front speed.

On the downstream side (right side of Fig. 2), producing a sinter of suitable quality at the lowest fuel rate and the highest productivity is the ultimate goal of most operations. The properties of the flame front have a major effect on these three plant performance indicators. The three critical flame front properties are flame front temperature (FFT), flame front width (FFW) and flame front speed (FFS) – the latter two determining the (flame front or heating) residence time as the front traverses a local region in the bed. The properties of the flame front are strongly dependent on the combustion behaviour of coke particles in the bed.

Mathematical models are available to describe many of the processes in Fig. 2, for example: granulation, bed permeability before and during sintering, and bed temperatures as the flame front traverses [3], [4]. Clearly, an important factor in predicting heat transferred to the bed from the flame front is the combustion efficiency of the coke combustion process. To better define the combustion behaviour of coke particles, the effect of granulation has to be taken into account because of its influence on coke particle location.

The combustion and gasification of coke/char has been extensively considered in fixed beds (e.g., [5], [6]), moving beds (e.g., [7]), pulverized coal combustors (e.g., [8], [9]) and circulating fluidized beds (e.g., [10]). In contrast, only a limited number of coke combustion models for iron ore sintering have been reported since the 1970s [11], [12], [13], [14], [15]. A simple rate-controlling model which takes into account the resistance of the gas film diffusion and reaction at the interface is most widely adopted [13], [15]. Yang et al. [16] introduced another resistance – that of diffusion through the ash layer. Toda et al. [17] considered that with the generation of melt, coke combustion rate reduces because the diffusion efficiency of oxygen to coke particles declines. To reflect changing gas concentration within a granule, Cumming et al. [11] proposed the inclusion of an empirical availability factor into the model. To describe the coke combustion process, Zhou et al. [4] combined the resistance of the adhering fines layer with the diffusion resistance of the developing ash layer. But it is more reasonable to consider the two resistances separately.

None of the above coke combustion models take into account the explicit influence of granulation. A single coke combustion mechanism is inadequate because coke particles do not all have the same access to oxygen – being positioned at different locations within granules and encapsulated by varying amounts of adhering fines. It is to be expected that ignoring the diffusion resistance associated with adhering material will result in higher than expected combustion rates – and this has been confirmed by simple bench-scale studies. Hida et al. [18] studied this relationship using coke and alumina particles of varying sizes. Four states or configurations were considered: (a) S – single coarse coke particle coated with fine alumina, (b) S′ – free coke particle, (c) P – pellet (sphere) of well-mixed fine coke and fine alumina, and (d) C – large alumina particles coated with a mix of fine coke and fine alumina. They considered that a typical granulated mix would have 70%, 10% and 20% material simulated by the S, P and C states, respectively. Their sintering simulation studies indicated that increasing the proportion of S′ and C type coke particles increased combustibility and decreased NO emission.

Likewise, Kasai et al. [19] studied the S and P type quasi-particles made of coke and alumina and found that the formation of melt can decrease combustion rate. Tobu et al. [20] suggested that C type can have higher combustion efficiency η – defined as the volumetric concentration ratio of [CO₂]/([CO₂] + [CO]) – than other configurations. Ohno et al. [21] investigated the combustion of tablets comprised of very fine coke (<125 μm) and alumina (<250 μm) to simulate the P configuration. From the work, a combustion rate model was derived.

On determining the proportion of each configuration in a sinter mix, it is possible to model them separately and then combine the components to obtain a composite description of coke combustion behaviour during sintering. However, this is a tedious task. In this study a granulation model was used to provide information on the presence and thickness of solid materials around coke particles. The influence of these layers in determining the diffusion of oxygen to the burning coke particle was then incorporated into the coke combustion model. Essentially, the modified combustion model takes into consideration the partitioning of coke particles of different sizes in granules. Through considering the combustion of coke nuclear particles and coke particles in the adhering fines layer an overall coke combustion rate for the full size distribution was derived. The combustion model was then incorporated into the bed heat treatment model developed by the current authors.

Clearly, the heat treatment model is extremely complex and improving the accuracy of the outputs is an on-going process. Another aim of this study is to incorporate two additional reactions – the dehydration of goethites and the dehydroxylation of serpentine – to improve the model. The level of goethite in many iron ore blends has continued to rise and the impact of the dehydration on the heat load in the bed must be considered. Also serpentine is a popular source of magnesia in sinter mixes and its endothermic reaction needs to be taken into account in the heat treatment model.

Having made these modifications, the model was run using different scenarios and the outputs compared against:

• previous laboratory pot test results [3], [4] to show the effect of these changes,

• information in the literature on replacing coke with charcoal and biomass char,

• information in the literature on the effects of intentionally placing coke particles on the outside of granules, and

• information in the literature on the effects of varying the size distribution of coke.