Lets make a castable!!

Bjrn Myhre
Elkem Materials, P.O. Box 8126, Vaagsbygd, 4675 Kristiansand Norway

Since the beginning of the 1980s, Elkem Materials has been manufacturing microsilica for refractory purposes. During these years, in spite of numerous attempts to link performance in castables and chemical analyses, no definitive correlations have been found that could replace the sometimes tedious testing of new microsilica in castables. Over the years the test castable developed to what we now use, a low-cement composition based on white fused alumina. Since our task is to test variability of components in the castable, it is a point that the castable should consist of well-defined ingredients from reliable suppliers. There are a number of decisions that have to be made before making the castable; first of all, one must decide what is the intention of making this castable?. In our case, the intention is to produce a castable in which new ingredients could be tested for performance in the laboratory. For us this is normally microsilica, but could just as well be one of the other ingredients, such as cement, calcined alumina, dispersants, etc.

Mixing equipment: Since we like to do the testing in our laboratory, we would probably like to use a relatively small batch size, e.g. 3 kg. In many castable laboratories the mixer equipment consists of Hobart mixers capable of mixing approximately 3.5 kg of dense castable. This is a paddle-type mixer with a revolving paddle that rotates at a speed determined by fixed gears. Normally a speed of approximately 60rpm is chosen. The paddle and bowl must be of stainless steel. Quite often the paddle is made from aluminium, which wears out quickly and also, more importantly, influences set time of the castable. So, use stainless steel. The size of the mixer makes use of particles larger than 4-5mm difficult, so the test mix is restricted to aggregates smaller than e.g. 4mm by the mixer. Larger particles will jam the motion of the paddle and make lots of squealing noises when the aggregates get trapped between the bowl and paddle. Better avoid that

Hobart Mixer.

A few words about ambient conditions: The testing should be performed in a room with moderate temperature and humidity that should be kept as constant as possible. This is not essential, although it may make interpretation and comparison of the results easier. The components of the mix should be stored though, in this environment.

The ingredients:
We have already indicated that such a test castable should be based on reproducible and well-defined raw materials like white fused alumina, tabular alumina, etc. The most important parameters here are absence of water solubles, or at very low and constant levels, and a defined internal porosity and shape of the aggregates. We need several sizes of aggregates, from the top size and down to milled fractions, additionally cement, calcined alumina, microsilia and dispersant. Sometimes retarder or accelerator are also needed to adjust set time to within acceptable limits. Below follows a short description of desired characteristics and suggestions for the various ingredients of the test castable. Aggregates: Fractions from 5mm and down to 200mesh (-74micron). The aggregates should be dense and have a minimum amount of open porosity. Shape should approach equiaxed or rounded which will impart flow. No or insignificant content of leachable components is desired.

Elkem Materials Page 1

Cement: This is one of the components frequently tested in such mixes. We need a reference though, and a good choice is to use a well-known international grade without additives from one of the major international manufacturers. Recommendations are Secar 71 from Kerneos and CA14M from Almatis. Calcined alumina: Used to fill up the gap between the milled fraction (-74micron) of the aggregate and microsilica which is essentially submicron. Look for particle sizes between 1-10 micron. Ease of dispersion is important here, so also use a well-known supplier. Almatis has some qualities that work well, like CT9FG, HVA SG and CT800FG. There are also many other qualities available from other suppliers that can be used. Microsilica: This is another component that is commonly tested. Microsilica is the finest fraction in the castable make-up and is probably, together with the cement, the component that influences properties the most. More than 50% of the surface area of a mix is normally found originating from the microsilica. When water is added to the castable, microsilica and the other fine ingredients join into a liquid slurry-phase of which amount and properties form the origin of the desired flow-properties of the castable. The extremely small size of the individual microsilica particles fills voids down to nano-scale and provides proper dispersion, ensuring a castable with the expected properties. The state of dispersion is however extremely complex and is easily influenced by any impurities in the system. Any leachable component may cause the delicate balance to be disturbed and cause a rapid coagulation, resulting in loss of flow or even solidifying. Such loss of flow is a normal initiation of the setting of a castable, but is a problem when it starts at an undesired time, either too soon or too late. Here we are aiming to make a test castable, and need to pay particular attention to the choice of the microsilica in the reference mixture. There are only a few high-grade microsilicas commonly available, 971U from Elkem Materials is here the preferred choice. Dispersants: If microsilica and cement are mixed in an aqueous media, immediate coagulation will be seen. Microsilica coagulates in contact with water soluble ions like calcium and aluminate from cement. Without control of this coagulation, microsilica-containing, cement-bonded castables would not have been viable without use of excessive amounts of water. This is why we use dispersants. A well-working dispersant prevents the microsilica and the other fines from coagulating, but also influences setting time. Proper use of additives requires the correct addition level, and there is a range of additives available today that works well. Some of them are easily available and cheap, others are more sophisticated and more expensive. The correct dosage is somewhat dependent on the mixture and has to be tested individually. Good examples working in WFA-mixes with microsilica and high-grade cement are: Sodium hexametaphosphate ( approx. 0.2% by weight addition). Brand name Calgon; many producers, cheap. Polyacrylates: Darvan 811D from Vanderbilt is recommended. Approximately 0.05% by weight Polycarboxylate ethers : Castament FS20 from BASF gives good results at 0.05% by weight Other additives: Retarders are frequently necessary. One easy and good example is citric acid at levels <0.1wth%. Boric acid is very powerful. Accelerators: Lithium carbonate and Sodium aluminate. Sodium carbonate, Partly hydrated aluminas. (Alphabond, Rho-alumina) Levels are variable, lowest for the carbonates, highest for the aluminas (up to >1%).

The making of a castable:

When we have chosen our ingredients, it is time to compose the castable. Elkem Materials has, since the beginning of the 1990s, used Particle Size Distribution analyses in the making of new mixes based on new raw materials. The reason for doing this is that it saves us a lot of testing in the laboratory while enabling us to predict flow properties of the finished castable. The methodology is based on the perception that with perfect packing of the individual particles, a minimum amount of water is needed to make it a castable. During the last 100 years or more, quite an extensive amount of work has been undertaken in order to make theoretical packing models, but we have chosen to use the one that i was proposed by Andreassen in 1930, due to its simplicity and proven results. Below is the equation given: Andreassen: CPFT =

d 100 D

CPFT: Cumulative Percent Finer Than d: Particle size D: maximum particle size q: distribution coefficient (q-value)

Elkem Materials Page 2

Andreassen assumes an infinite distribution, i.e. no minimum particle size which has been held up against the model, but according to our experiences this is merely a psychological problem with mixes containing microsilica. If this equation is plotted in a double logarithmic (log-log) plot it equals a straight line with the q-value as the slope. Experiments using different particle size distributions based on the Andreassen distribution, have shown some very interesting relationships which concern flow of the castables at a fixed amount of water (in volume). ii Results obtained on castables based on White Fused Alumina (WFA) are shown below at a fixed water addition of 13 volume% water (approximately 4% by weight). This Figure shows a strong relationship between q-value and self-flow, but not so much between q-value and vibration flow.

180 160 140 120 Flow (%) 100 80 60 40 20 15 12 10 8 Microsilica wt% 0,20 0,22 0,24 q-value 0,26 0,28 6 4

180

Vibra-flow

160 140

Self-flow

120 Flow (%) 100 80 60 40 20

0,30

Figure 1 The influence of q-value on flow of a castable based on white fused alumina (4mmtop size) after (ref 2) Corresponding amount of microsilica is also indicated. By choosing the appropriate q-vale, one is in other words, able to design a castable with either self-flow or vibration flow purposely. Self-flowing castables have lower q-values than vibratables, or higher amount of microsilica. LISA software To ease the use of PSD analyses, Elkem Materials has developed downloadable software for calculating PSD of castable mixes and Andreassen distributions. The software is found at http://www.materials.elkem.com and the software is named LISA. From this site more than 50 other relevant publications by Elkem Materials staff can also be downloaded. After installing the software, the user threshold is comparatively low and provided the new user is equipped with individual PSDs and densities for his individual fractions, he should be able to start designing castables right away! In Figure 2, a screenprint of the LISA software is shown, here we have already made a composition that matches the Andreassen distribution with a q-value of 0.25. This is seen in the graph to the left.

Elkem Materials Page 3

Figure 2: Screen print of the dialogue window of LISA, showing distribution to the left and recipe to the right. Because of a q-value of 0.25 at a top size of 4mm, this castable is, according to Figure 1, expected to show a semi-self-flowing behaviour at 8% microsilica and 4.15 wt% water. Comments on top-size and density: It will always be possible to add coarser particles to a well-working matrix. The use of larger aggregates is one of the variations commonly used to make impressively low water additions. If we take the above castable as an example, one may add up to 25% coarser material like Tabular alumina to the mix with 4.15 wt% water. Thereby we have 125% dry matter. By doing this small operation the water addition has decreased from 4.15% to 3.32% (4.15/1,25). Looks much more impressive, doesnt it? Another trick is to use high density aggregates like zirconia. If aggregate density gets 50% higher, then seemingly only 2/3 of the water is needed. In volume% it is the same, though. That is why we have been focusing on volume in our work with packing; it is the volume that fills voids, not the weight.

Then, to the lab!

Figure 3: Our test recipe, - which we have to modify slightly.

Elkem Materials Page 4

The recipe we developed in the LISA software contains both BACO MA95 and 6% cement. BACO is no longer existing,- at least the product MA95 no longer exists, so we replace that with CT9 FG from Almatis which has a PSD very close to MA95. In the following, we will play around with the castable so it is unwise to let it contain cement this time. We replace the cement with the milled -75micron fused alumina. But before starting playing around, a few words about general procedure: General procedure when making test mixes in the laboratory: 1) Always make a reference mixture based on known ingredients and with known properties if new components are to be tested. Use cement from the same bag, properly stored inside a closed plastic bag at even temperatures. The same applies for the microsilica and additives. Use microsilica from one particular bag. Do always use the same microsilica for the reference, since there are differences between lots and sometimes between bags. As there are slight variations from day to day, always make a reference mix. The results should be judged considering the score of the reference. 2) When preparing the mixes, it is best (but not mandatory) to weigh every fraction directly into the bowl placed on a scale. By using the tare function weighing is simple and accurate. This ensures correct amount of all ingredients going into the mix. Total amount is normally 3500g. Additive has to be separately weighed on a finer scale due to small amounts and need for precision. All ingredients should be room tempered!!

3) After all dry ingredients have been added, dry mixing is performed for 4 minutes. This duration may be reduced, but be consistent!

Elkem Materials Page 5

4) A stopwatch is an indispensable tool in doing this testing. Use it to keep track of time!

5) During the 4 minutes dry-mixing, potable, tempered water are weighed to a pre-determined amount. We use 4.15 wt% or 13 vol% in these mixes.

6) After water has been added, mixing continues for another 4 minutes. Things to note are wet-out time, loss of flow during mixing, flash set, jammed mixer, etc. In case of heavy loading, it is better to lower bowl in order to ease the movement of the paddle than to break the gear.

Elkem Materials Page 6

7) Flow testing: For the flow testing we use a flow cone as described in ASTM C230. Its base is 100mm and height 50mm. The measurement is performed on a vibration table connected to a timer so that accurate vibration can be undertaken. The castable is placed into the cone and after the cone has been removed, the castable is allowed to spread out by gravity alone. When spreading stops, the diameter of the patty is regarded as free-flow value; after subsequent vibration (15 seconds) one obtains the vibraflow value. The flow value may be expressed either as percentage increase of the diameter, or in mm. Then the mm reading minus 100 will equal to the flow value in %.

ASTM flow cone (height 5cm, base 10cm): This flow-cone is probably the most wide-spread cone, but in Europe (PRE), for testing self-flowing castables, another cone may be used. The European cone has a height of 8cm as opposed to ASTMs 5cm,and the volume is some 40% higher. Free-flow values obtained by using this cone will therefore be higher, one can expect some 40% more. Outside Europe, the use of the tall cone has not become common. For simplicity most users tend to stay to one cone in order not to mix up.

The effect of microsilica on flow of a castable:

We have carefully followed the instructions given in items 1 though 6 above and have weighed the recipe given by LISA. BACO MA95 is replaced with CT9 FG and this time the cement has been replaced by unreactive, milled alumina. The microsilica has been omitted for the time being, and dispersant has been added (0.05wt% Castament FS20). An amount equivalent to 4.15 wt% water has been added and the mix has been mixed for 4 minutes. In this experiment we will investigate what effect microsilica has on the flow.

1) 0% microsilica Castable mixture mixed with the water but without microsilica.

There is no self-flow whatsoever During vibration, the castable starts to slide over the table while disintegrating and crumbling. One more second of vibration and it would have fallen down from the table. No flow with or without vibration. This is not a castable.

Elkem Materials Page 7

Let us look at the PSD:

Figure 4: PSD of the microsilica-free castable composition. Andreassen distribution with q=0.30 and D=4000 as straight line. In Figure 4, we see that particularly in the lower size classes there is a big deviation from ideality represented by the Andreassen distribution, even at a q-value of 0.30. We are facing a situation where there is a lack of microsilica, and this creates a situation where there is an undersupply of liquid to fill the internal pore system and promote flow. The castable appears as moist sand.

3 wt% microsilica. The above castable gets 3% microsilica addition After the testing depicted above, the castable was placed into the bowl, mixing was continued and 3 wt% microsilica was added. The appearance changed, the castable got stickier and started looking more like a castable. No self-flow was recorded however, but under vibration a decent flow was obtained. Vibra-flow values were close to 60% or so.

No self-flow with 3 wt% microsilica either

But with vibration, flow is attained.

Elkem Materials Page 8

Figure 5: PSD of Castable with 3 wt% microsilica addition together with Andreassen distribution with q=0.30 and D=4000 The flow of this castable shows approximately what we could expect from the results shown in Figure 1, although vibra-flow is lower than can be read from that graph. The main lesson here is that we may get acceptable vibration flow, and no self-flow, with low microsilica contents; i.e. high q-values. In terms of packing, we here have an aggregate structure where there is just enough liquid phase to fill voids in the structure, but not enough to make the individual particles slide relatively to each other (flow) without help(vibration). Internal friction due to particle-particle contact is too high. 8 wt% microsilica add another 5 wt%! The procedure of collecting the castable into the bowl and mixing in another dose of microsilica was repeated. This time the castable with 3 wt% microsilica had an additional 5 wt% added so that it contained a total of 8 wt% microsilica. Already during mixing after adding the extra 5 wt% dry microsilica, it became clear that great things were happening. The surface became shiny and the whole castable looked loose in structure. This is shown in the pictures below that have been made during mixing. The shiny surface indicates the ease at which the mixer mixes this easy flowing substance with no hints of dilatancy or thixotropy.

The test castable with 8 wt% microsilica. The picture is taken during mixing, and there is no hint whatsoever of the dryish appearance that would be attributed to dilatancy and thixotropy

Another close-up of the same castable during mixing.

Elkem Materials Page 9

The cone is removed and the castable allowed to spread out by gravity alone. Free-flow value attains some 80%

15 seconds vibration brings the patty to flow almost out of its edges. Vibra-flow around 160%!

If we compare the obtained flow values of 80% self-flow and 160% vibra-flow with those shown in Figure 1, there is a remarkably good agreement. Even though the results were obtained with similar materials, the results span some 10 years in time. If we take a look at the particle size distribution shown in Figure 6, together with the Andreassen distribution for q=0.25, first of all what we note is a very good match, down to approximately 1micron. Such a good match together with a non-dilatant viscous matrix phase makes full utilisation of the liquid phase that is confined in this system. With 8 wt% microsilica there is excess of the liquid that facilitates selfflow. Based on earlier investigations the below mechanism attempts to explain the phenomena of vibra and self-flow: The dependence of the flow on microsilica content is a two-stage process. The first step is depicted in Figure 7i). If microsilica is added to a castable mixture without (or deficient in) superfines, the first step is that microsilica enters the void structure and replaces water on a volume-to-volume basis. One volume of microsilica replaces one volume of water. However as the aggregates touch, some external energy input is still needed to make them move relatively to each other. This energy is normally vibration, and what we have is typical vibratable. Figure 7ii) shows the next step. Substitution of the water proceeds up to a certain limit, when the liquid of microsilica and water reaches a critical density, or crowdedness is perhaps a better term. When the microsilica addition exceeds this limit (normally around 10 vol% or 5 wt%microsilica), the total volume of the liquid system starts to increase, it is not possible anymore to replace water by microsilica and still maintain flow. An increased volume of liquid phase increases the total volume of the castable. Because of the increased volume of the liquid phase, the particles move apart and become free to move relatively to each other. This is when free-flow commences (and it happens at around 5-6 wt% microsilica and higher).
iii

Elkem Materials Page 10

Figure 6: PSD of the castable with 8 wt% microsilica obtaining high flow at 4.15 wt%water.
i) Microsilica replace water (volume by volume)

ii) The total volume expands Self-flow

Figure 7: Schematic representation of microsilica additions to a castable

Elkem Materials Page 11

Part 2: The use of a castable in QC/QA.

If we accept the perception that by using impeccable ingredients, the expected properties will be gained in an industrial mix, then it should be ok to test each individual ingredient in a castable that is intended to detect deviations from normal. The quest is then to make a test-mixture and to execute the testing in a fashion that will detect all undesired deviations.

What if we want to make comparative tests?

Then we follow the recipe given in Figure 3, and perhaps add other additives than dispersants. Normally we try to avoid adding too many chemicals though, and often dispersant alone does the trick if room temperature is around 20C. At higher temperatures, retarder has to be added in order to not lose flow too early. What we do in practise, is to make up two (ore more) castables according to the procedure described earlier. One of these is the reference sample that is made out of a mixture of well-known ingredients. This is there to detect variations due to humidity and small deviations in aggregate quality, etc. The reason for having the reference is in other words, to detect any deviation from normal. The other castables have each one component replaced. If we have 5 cement qualities to test, we make 6 castables, 1 with the known cement and the other 5 containing the different cements to test. Likewise with microsilica, additive or any other component we would like to test. Test elements: Additional to flow, free-flow and vibra-flow, a few other items are essential to test in this kind of testing. A controlled flow is essential for a castable, and so is the setting or hardening. Wet-out: An important measurement is the wet-out time. Wet-out indicates the duration of mixing it takes for the castable to become plastic after water is added. The measurement itself is subjective and the precision is perhaps not so good, but a long wet-out time indicates that dispersion may be a problem. With no wet-out the problem is bigger. An acceptable wet-out time is depending on the mix, but in general, the faster the better. If a castable needs a long mixing time at a site, the chances of extra (too high) water additions increase significantly. No one likes to wait for a castable to get wet, particularly if you have 20 more mixes waiting for you during the job. The properties of the castable will of course suffer by this extra water. So, short wet-out time is preferred. In our mixes we like them to be shorter than approximately 30 seconds. Flow measurements: We have touched flow measurements earlier, we measure the diameter (mm or % increase) of the cone after spreading by gravity alone (free-flow) and after subsequent 15 second of vibration (vibra-flow). The measurements are best performed making 4 cross-sections of the patty, and then average the number. This would adjust for incircularity of the patty. The tool for this is frequently a ruler, but better is a special calliper as described in ASTM C230. What flow one should aim for is a matter of discussion, but freeflows in the range 40-80% (140-180mm) should be adequate for most applications and vibra-flows around 90130% should give good results. Very high flow values are normally not sought for as this will impart segregation. Set time: Besides flow, this is perhaps the most important property for a castable. In modern refractory castable it is a requirement that set time is controlled to within predetermined limits, and deviations are frequently causing problems for both end-user and manufacturer. Setting can be judged in many ways, by simple touch, so called squeeze tests where a plastic beaker or bag of sample is squeezed from time to time by the operator to judge setting, by Vicat needle and similar instruments. The latter are based on measuring the resistance when a thin needle is pressed into the setting castable, but is best suited for mortars without aggregates. Recently devices calculating Youngs modulus by sonic measurements have come into use as well. Although all of these give acceptable results, they are not easy to use, particularly for a multitude of samples, and particularly the latter gets prohibitively expensive in standard quality control. At Elkem Material we have been using the exothermic reactions of the castables to monitor set.

Elkem Materials Page 12

The test equipment consists of a multi-channel strip-chart recorder (left), connected by cheap thermocouple leads to a number of insulated plywood boxes (seen to the right here).

An appropriate amount of sample to be tested is placed into a plastic bag after measuring flow

A thermocouple is made by removing insulation and twisting the leads of the cable (thermocouple type K), connecting the recorder to the plywood box.

The thermocouple is inserted into the plastic bag with the castables and the sample is placed into the insulated plywood box. The lid gets properly into place and temperature measurements can start.

This is how the recording may look like after testing. After a certain induction period, the temperature of the sample starts increasing; when we have a temperature increase of 1C we record the time and name it TTS (Time To Start) Time To Maximum temperature is also recorded (TTM)

After measurement is finished, the thermocouple lead is simply cut and a new thermocouple may be twisted for use next time.

Elkem Materials Page 13

Comments to the method: The physical and chemical significance of TTS is not well understood, but TTM is associated with the formation of hydrates that are responsible for the strength of the castable. TTS is normally associated with end of working time and TTM with final set. Not uncommonly, castables harden to some extent before TTS and one may be misled to believe that the castable is ready for demoulding. Demoulding before TTM is reached is however a risky business, because the strengthening processes, i.e. the formation of hydrates, have not finished or may not even have started,, and special care should be taken before demoulding or moving shapes before TTM is attained. We have chosen to use TTS as set time for two reasons. One is that at TTS there is normally no way to reverse setting, the second is that we have discovered that once TTS is reached, the time to TTM is quite constant in our system. It takes approximately 6 hours from TTS to TTM. One drawback with this method is that it cannot be used on ultralow cement castables. There is not enough cement to generate the required amount of heat necessary for our system. Perhaps with better insulation, one could go lower on cement, but presently the practical lower limit seems to be 4-5% cement. The use of a multichannel strip chart recorder is also outdated; nowadays one would set up a data logger and do the measurements through a computer. Further tests: Occasionally, the test sample is cast into shapes for strength testing. Earlier, cold compressive and flexural strength were tested as an integrated part of the set-up, but when it was realised that we never got deviations unless setting or flow was out of specification, it was decided to omit this labour-intensive part of the testing. Reporting: After wet-out, flow and set are measured, the results are reported and stored electronically for later look-up. Below an example of such a test report is given:
ELKEM MATERIALS To :BM FRom : rh Testing Microsilica wt% Grade Lot.nr Seq Cont. CA-14 % Fused Al. 3-5mm % Fused Al. 0,5-3 mm % Fused Al. 0-0,5mm % Fused Al. -74 mic % CT 9FG Castament FS20 % Water wt% Weight (g) Free-flow % Vibra-flow % Wet-out time (seconds) Temperatures Start temperature (C) Time to start TTS(h:min) Time to max. Temp TTM(h:min) Max temperature (C) g TTS corrected to 22C(hours) : 968 : : : Ref. 8 971U 3032985 1 8 968U 3040080 19590 APLU890533-0 6 10 35 20 12 9 0.05 4,15 3500 40 98 15 24,2 05:30 12:30 31 7,04 2 8 968D 3040080 19590 APLU888732-3 6 10 35 20 12 9 0.05 4,15 3500 44 116 15 23 06:00 12:30 30,2 6,7 3 8 968U 3040058 19591 DBOU512204-5 6 10 35 20 12 9 0.05 4,15 3500 48 120 15 23,7 06:30 12:45 29,6 7,69 18.03.2007

: : : : : : : : : : : : : : : :

6 10 35 20 12 9 0.05 4,15 3500 84 128 15 23,4 02:00 06:30 30 2,98

TTS is normally adjusted to 22C, in order to be able to compare more accurately. 22C is chosen because that is the temperature the castable most often attain during mixing at 20C room temperature. The temperature increase comes from the mixing energy imposed on the castable. Empirically, we have found that the temperature sensitivity of our castable is approximately 0.7hour/C for TTS, and the correction is done accordingly.

Five Guidelines:
Now as we have obtained results, we might sometime in the future want to compare old and new results? If the object is to compare results, the following guidelines may help to produce results that have lasting qualities.

Elkem Materials Page 14

1) Compare apples with apples. Do not try to compare results obtained with different ingredients without having full control. Do not change more than one ingredient at a time (typical if one has to change source of aggregates). If the aggregates have to be changed, then compare with a reference based on the old aggregate. In Elkem Materials we frequently find deviations in the results when a new supply of fused alumina arrives. We may then adjust the recipe somewhat (mostly based on careful PSD analyses) to gain same results as with the old aggregates, or choose to continue with the old recipe but with lower (flow) values. The flow of the castable system is sometimes remarkably sensitive to variations in aspects like PSD of the finest fraction, porosity, and (possibly) soluble elements. 2) Use synthetic raw materials in your test mix. Be particularly cautious with using fines based on natural raw materials. If it cannot be avoided using some natural raw materials, then at least try to use synthetic fines. The fines of natural raw materials are often loaded with impurities that influence flow, in bauxite, aluminium carbide and other carbides may be found; in fines of uncalcined natural minerals, residue from flotation aids can be found, and so on. Our preferred choice is to use white fused alumina in the milled fraction as this has given best results for us. 3) Use calcined alumina. Use a good quality calcined alumina that fits the gap between the milled fraction (-75micron) and microsilica (submicron). Do not use natural, milled, raw materials as the risk of contamination is high. With microsilica it should not be necessary to use sub-micron alumina (reactive) in this test mix, unless reactive alumina is to be tested. 4) Always pay attention to the PSD. You cannot interchange microsilica and reactive alumina unless the volumes are taken into consideration. Alumina is 1.8 times more dense than is microsilica (S.G. of 3.9 and 2.2 g/cc respectively) and in order to fill the same volume, 1.8 times as much should be used (based on weight). Think volume! 5) Obtaining best values is not a competition. The mix is there to be able to do a sober comparison between alternative sources of additives, microsilica, calcined alumina, cement etc. Also the aggregates can undergo similar treatise but do not change more than one component each time. It must be admitted though, that for use in an industrial environment, this treatise is dependent on the perception that any castable composed of impeccable components performs according to expectations.

Five ways of fooling yourself:

In case you are dissatisfied with your results, particularly the flow and the water addition, the following variations have tremendous effects: 1. Decrease your water addition by adding coarser grains. If you add 25% coarse, the relative water addition drops by almost 25%. E.g. from 4 wt% to 3 wt%. Expanding your PSD to bigger top sizes is very effective, but may be hard for the mixer. Industrial mixes normally do contain larger aggregates (up to 10-12mm) to impart properties like thermoshock resistance and slag resistance, etc. but here we are dealing with a laboratory mix for testing flow and set. 2. Add some really heavy aggregates or fines. Addition of zirconia or other heavy minerals decreases your water requirement significantly. At least it looks so in weight%. The volumes remain the same though, but the numbers look more impressive. 3. Improve free-flow by using the PRE-cone of 8cm instead of the ASTM-cone of 5cm height. A significantly improved free-flow may be attained, and if you are addicted to free-flow values in the range of 100% or more, you have succeeded. A free-flow of 100% for the ASTM cone is probably a bit too high, though. 80% is more desirable if segregation should be avoided. 4. Add more water. This is quite effective. Earlier, unpublished results have indicated that increased water addition by 1 volume%, increases flow by approximately 25% (or mm). For a castable like that described above (density 3.2g/cc), 1 more weight% water could result in up to 75-80% extra flow. That much water would possibly result in separation and segregation though, together with increased porosity (by approximately 3%). 5. Do a combination of the above. A combination of all elements above will certainly boost your numbers and make your mixture look invincible. It does not improve your basic castable though, only makes it look more attractive.

Elkem Materials Page 15

References: Most of these and many others may be found and downloaded from http://www.materials.elkem.com, Look under Refractories
A.H.M. Andreassen and J. Andersen: Kolloid Z. 50 (1930) 217-228 B. Myhre and K. Sunde, "Alumina based castables with very low contents of hydraulic compound. Part I: The effect of binder and particle-size distribution on flow and set.", in Proc. UNITECR95, Kyoto, Japan, Nov. 19-22 1995, p. II/309-16 iii B. Myhre: "Particle size distribution and its relevance in refractory castables", in Proc.2nd India Internat. Ref. Con., New Delhi Feb. 8-9, 1996, p. I/47-56.
ii i

Elkem Materials Page 16