Design of the particle size composition of an alumina powder matrix for maximum flowability and minimum water content.

Transcription

1 Design of the particle size composition of an alumina powder matrix for maximum flowability and minimum water content Abílio Silva 1, Ana M. Segadães 2, Tessaleno Devezas 1 1 Univ. Beira Interior, Dept. Electromechanical Engineering, Covilhã, Portugal 2 Univ. Aveiro, Dept. Ceramics and Glass Eng. (CICECO), Aveiro, Portugal segadaes@cv.ua.pt Keywords: Refractory castables, Powder self-flow, Statistical modelling Abstract: In this work, commercial alumina fine powders were used as raw materials, namely two tabular alumina fractions ( 500 mesh and 230 mesh) and a reactive alumina. Statistical modelling and the Response Surface Methodology (Statistica, Mixtures Designs and Triangular Surfaces module) were applied to three-component mixtures and used to calculate the various property-composition surfaces. To that aim, the various mixtures were prepared, cast, dried, fired and characterised. The particle size distribution modulus, q, was determined for all mixtures using the software LISA. The various response surfaces were then combined, so that the water content in the mixture could be minimised and the matrix flowability maximised. The properties of the resulting test-bricks (linear shrinkage, mechanical strength, apparent density and porosity) were also modelled and response surfaces were obtained. Combined results enabled the definition of an optimised particle size composition range, which guarantees the presence of a low water flow-bed that enables the aggregate self-flow. Introduction The water content in refractory castables, particularly in no-cement self-flowing castables, must be kept to a minimum, which means that the flowability and the paste-like fluid state of the castable have to be provided by a matrix of fine powders. Thus, the fine powder matrix performs the filler function and acts as a binder, but is also responsible for the wet castable flowability characteristics. The obvious way to reduce the water requirement is to maximise the powders packing density, i.e. to have an Andreasen particle size distribution modulus, q, near 0.37 [1,2]. However, maximum particle packing translates into minimum flowability and the ideal Andreasen q values should be near 0.22, which corresponds to a higher Maximum Particle Thickness (i.e. reduced interferences between the largest particles) [3,4]. Given the low water level and the fineness of the powder matrix, there is a competition between the Van der Waals interaction forces and the capillary forces between particles, and even the water mixing procedure has to be optimised [5-7]. When the water mixing procedure is kept constant, so that all mixtures reach a similar paste-like condition, important properties of the wet castable, such as its flowability, are basically determined by the combination (or mixture) of particle sizes in the fine powder matrix. After firing, that fine powder matrix also determines the final properties of the refractory lining. That is the basic assumption in the design of mixtures experiments to obtain a response surface using mathematical and statistical techniques [8]. Using the software Statistica (Mixtures Designs and Triangular Surfaces module) [9], a system can be defined comprising three independent size fractions (components or ingredients of the mixture) and,

2 on this composition triangle, a regular array of uniformly spaced points (simplex lattice) is set. The value of the property of interest is experimentally determined for each of the simplex compositions and, based on those results, a mathematical equation (model or response surface) can be calculated and used afterwards (after statistical validation of the model) to estimate the value that the property will assume for any composition [8]. The various response surfaces can then be combined [10], so that an optimum composition range (i.e. combination of particle sizes) can be determined, to produce a castable with the desired performance. Materials and Methods Commercial alumina fine powders were used as raw materials, namely the Alcoa T60 tabular alumina in two size fractions ( 500 mesh and 230 mesh) and the Alcoa CT3000SG reactive alumina (AR). A detailed description of the powders characteristics can be found elsewhere [1,2]. These selected size fractions define the composition triangle on which a simplex lattice with ten points (A, B, J) was set (Fig. 1). Afterwards, three extra points (K, L and M, also in Fig. 1) were added to better characterise the self-flow composition range. Figure 1. Composition triangle and simplex points. Shaded area is the self-flow range. The approximate Andreasen particle size distribution moduli, q, were calculated for all simplex mixtures using the software LISA [1,2,11]. The powders with the specified simplex compositions were mixed in a mortar-blender and anhydrous citric acid was added as a deflocculant, in the proportion of 0.36 mg/m 2 of solids specific surface area [5,6,12]. An intermittent water addition mixing procedure was used (the detailed description can be found elsewhere [7]) until the mixtures reached a paste condition (in the self-flow composition range this means that the mixture is near the turning point ). The flowability index (FI) was then determined (ASTM C-230) and the amount of water used was referred to the solids specific surface area (SSA). The various castables were poured into casting boxes (150x25x25 mm 3 test-pieces) and cured in the open for 24 hours. After de-moulding, the specimens were dried at 110ºC for 24 hours and then sintered at 1600ºC in a muffle oven. After sintering, the linear shrinkage (LS) was measured (ASTM C-113) and the modulus of rupture (MoR) was determined in three-point bending tests (ASTM C-133). The resulting broken pieces were used to determine (ASTM C-20) the apparent density (AD) and porosity (AP). The values of the various properties were used to calculate, with the software Statistica, the property-composition surfaces, subjected to a 5% significance level and experimental validation [8-10].

3 Results and Discussion Wet matrix properties. The water requirements and mixing time, needed to reach the paste-like condition, is not the same for all mixtures. During the preparation of the castables it was clear that there is a composition range (shaded area in Fig. 1) where mixtures display a self-flow character (i.e. show a turning point ). The mixing time (min) and added water (mg/m 2 ) surfaces, constructed with the results obtained, are presented in Fig. 2. Figure 2: (a) mixing time surface (min), and (b) added water surface (mg/m 2 ). See Fig.1 for mixture identification. It is clear that mixtures with a turning point need shorter mixing times and require less added water to reach the fluid state (generally below 30 mg/m 2 of solids SSA). Mixtures within the high water composition range will also show high flowability due to the lubricating effect of a high amount of water. Since the objective is to minimize the added water, the restricted composition area in which mixtures have a turning point was investigated in greater detail. Fig. 3 shows the surfaces, for this restricted area, for the FI (Fig. 3-a) and the Andreasen distribution modulus, q (Fig 3-b). Figure 3: Response surfaces for the composition range of mixtures with a turning point : (a) Flowability index (%), and (b) Andreasen distribution modulus, q. Fig 3-a shows that mixtures that have a turning point (with low water content and short mixing time) present the highest flowability (FI > 150 %) and are, thus, ideal for a future use as a castable matrix. Fig 3-b shows that these mixtures have Andreasen distribution

4 modulus q between 0.21 and 0.28, which is within the usual range of optimised particle size distributions for maximum flowability [3]. Combining the property-composition surfaces obtained for mixing time, added water, flowability and distribution modulus, shown in Figs. 2 and 3, it becomes clear that a particle packing optimised for highest flowability combined with the appropriate mixing procedure lead to minimum added water requirement and a wet castable with a turning point. Sintered Body Properties. The results obtained for the properties of the fired bricks produced with the various castables can be found elsewhere [7]. Again, all these properties were modelled using Statistica and response surfaces were calculated and are shown in Fig.4. For the composition range of mixtures with a turning point, the linear model was always the best one. Figure 4: Response surfaces for the composition range of mixtures with a turning point : (a) Linear shrinkage (%), (b) Modulus of Rupture (MPa), (c) Apparent density (g/cm 3 ) and (d) Apparent porosity (%). Fig.4 shows that it is possible to choose a high flowability matrix with ~6% shrinkage, MoR of at least 60 MPa, apparent density between 3.42 and 3.58 g/cm 3, and apparent porosity between 4.1 and 9.7 %. Fig. 4 also shows that the reactive alumina content is the parameter that most affects all the properties. The wet matrix behaviour also has a strong bearing on the fracture characteristics of the sintered body. Mixtures with a turning point (i.e. with higher MoR) produced smooth surfaces and the transgranular fracture progressed into diagonal directions (Fig.5-a). Mixtures

5 without a turning point (i.e. with lower resistance), fractured intergranularly along the applied load direction and produced much rougher surfaces (Fig. 5-b). Figure 5: Fracture surfaces as seen on SEM: (a) for mixtures with a turning point, and (b) for mixtures without a turning point. Although the final castable behaviour is mostly controlled by the finer sized particle fraction (matrix), the presence of aggregate coarser particles will have some advantageous effect on the ultimate castable properties, such as higher mechanical resistance and better dimensional stability. Nevertheless, the introduction of coarse particles should not destroy the matrix characteristics. In other words, a commercial distribution of particle sizes, as produced in normal grinding operations, can be theoretically separated into aggregate and matrix fractions, the missing finest part of the matrix is brought in by added reactive alumina so that it falls within the optimum property range, and the appropriate amounts of each size class can be calculated with LISA so that the overall distribution modulus is the same as (or very close to) that of the matrix. Table 1 shows an example of such an exercise, for a distribution modulus q=0.22. Table 1. Properties of the full castable, prepared with as-ground commercial alumina. Aggregate SSA [m 2 /g] Matrix SSA [m 2 /g] Total solids SSA [m 2 /g] Added water [mg/m 2 of total solids SSA] 29 Flowability index [%] Linear shrinkage [%] 1.93 Modulus of rupture [MPa] 74.4 Apparent density [g/cm 3 ] 3.32 Apparent porosity [%] 8.68 The wet matrix characteristics were in essence extended to the full castable, in particular the high flowability of the matrix was preserved. As expected, the fired body properties improved with the introduction of the aggregate, namely the linear shrinkage was reduced and the MoR increased, and are within the usual range for self-flow castables. Conclusions Statistical modelling and the response surface methodology were used to determine the best combination of three commercial alumina fine powders, likely to produce a self-

6 flowing castable matrix. The results obtained showed that there is a reasonably broad composition range, rather then a single solution, where powder mixtures present the desired characteristics: good particle packing, low water requirement, short mixing time and high flowability with a turning point. The preparation of a full castable, using as-ground commercial alumina, can be based on such an optimised matrix and its distribution modulus q, with the help of the software LISA. With the introduction of the aggregate, the wet matrix characteristics were in essence extended to the full castable, in particular the high flowability of the matrix was preserved. As expected, the fired body properties improved and are within the usual range, resulting in a self-flow castable with high mechanical resistance and good dimensional stability. References [1] A.P. Silva, A.M. Segadães, T.C. Devezas: Relações Entre Distribuição Granulométrica, Morfologia e Empacotamento de Partículas Num Sistema Real: Alta- Alumina, Proceedings of 47 CBC, João Pessoa-PB, Brazil (2003), p (in Portuguese). [2] A.P. Silva, A.M. Segadães, T.C. Devezas: Aplicação de Métodos Estatísticos na Optimização da Densidade de Empacotamento de Distribuições de Pós de Alumina, Cerâmica, 50 [316] (2004), p (in Portuguese). [3] J.E. Funk, D.R. Dinger: Predictive Process Control of Crowded Particulate Suspensions Applied to Ceramic Manufacturing, Kluwer Academic Plub., [4] I.R. Oliveira, A.R. Studart, R.G. Pileggi, V.C. Pandolfelli, Dispersão e Empacotamento de Partículas Princípios e Aplicações em Processamento Cerâmico, Faz. Arte Editorial, S. Paulo, 2000 (in Portuguese). [5] R.G. Pileggi, A.R. Studart, J. Gallo, V.C. Pandolfelli: How Mixing Affects the Rheology of Refractory Castables, Part 1, Am. Ceram. Soc. Bul., 80 [6] (2001), p [6] R.G. Pileggi, A.R. Studart, J. Gallo, V.C. Pandolfelli: How Mixing Affects the Rheology of Refractory Castables, Part 2, Am. Ceram. Soc. Bul., 80 [7] (2001), p [7] A.P. Silva, A.M. Segadães, T.C. Devezas: Statistical modelling of the particle size composition of an alumina matrix for no-cement self-flowing refractory castables, Mater. Sci. Forum, 2005 (accepted). [8] J. Cornell: Experiments with Mixtures, 3rd Ed. (John Wiley & Sons, NY 2002). [9] Statistica for Windows 5.5, Comp. Prog. Manual, StatSoft, (USA 2000). [10] S.L Correia, D. Hotza, A.M. Segadães: Application of mathematical and statistical strategies to optimize ceramic bodies: effects of raw materials on the technological properties, Ceram. Forum Int., 82 [1-2] (2005) p. E39-E43. [11] Lisa Size Distribuition Analyser 2.0, Elkem ITS & Elkem Materials Refractories, [12] A.R. Pardo, R.G. Pileggi, V.C. Pandolfelli: Novo método para seleção de dispersantes de última geração para concretos refractários, Cerâmica, 48 [308] (2002), p (in Portuguese).