ALUMINA DUSTINESS RELATED TO PHYSICAL QUALITY PARAMETERS - USER EXPERIENCE AND R&D IN HYDRO ALUMINIUM

Dag Olsen

Norsk Hydro Research Centre

P.O.Box 2560

N-3901 Porsgrunn

NORWAY

ABSTRACT

Dust emission during unloading, handling and pot operation is the main alumina related environmental problem experienced in Hydro Aluminium's smelters in Norway. Our system uses alumina from more than 10 different sources. The dustiness of each alumina is one of the main criteria for approving alumina sources. The dustiness of various alumina qualities has been investigated through sampling and measurements during unloading and pot operation at two of our smelters, as well as analyses and tests in the laboratory. Our data shows that fines content is the most important of the traditional alumina quality parameters, although no general correlation has been found. However, Perra Pulvimeter tests using test samples from one single source yield a strong correlation between dustiness and fines content. The effect of attrition index depends strongly on the handling system at each plant. According to our data the Perra Pulvimeter provides a better prediction of alumina dustiness than any of the traditional quality parameters, not only during handling, but also in Soederberg potlines. There are indications that alumina dustiness is more closely linked to properties of the dust fraction itself, than to bulk alumina properties. Data from refinery plant tests shows an almost perfect correlation between hydrate strength and fines content of the alumina. Our working hypothesis is that the strength and structure of the hydrate is the major key to alumina dustiness. A better understanding of the link between White Side conditions, hydrate properties and further on to alumina properties and dustiness is fundamental to achieve the ultimate goal; to produce a less dusty alumina at the refineries.

KEY WORDS:

alumina quality, dustiness, fines, perra, emissions

Alumina Dustiness Related to Physical Quality Parameters - User Experience and R&D In Hydro Aluminium

Dag Olsen

1.0 INTRODUCTION

Hydro Aluminium is one of the largest buyers of alumina in the open market, and is regularly using alumina from 9-12 different sources in the Norwegian smelters and several others as part of tolling agreements. Our smelters’ main environmental problem linked to the alumina is dusting and dust emission during handling and smelter operation. Most of them are located in narrow fjords between steep mountains, and approximately 40% of our metal are produced in open Soederberg pots. Combined with strict environmental regulations, this makes it important to allocate the right alumina to each smelter.

It is widely assumed that alumina handling dustiness in some way is linked to fines content, attrition and particle shape. The subject has been discussed in several papers, for instance Hsieh (1987), Authier-Martin (1989) and Roach and Cornell (1990). Although clear trends have been found for single alumina sources, no general correlation between alumina quality parameters and dustiness has been found.

Hydro is investigating the dustiness of various aluminas in order to improve our understanding of the relationship between alumina quality parameters and the observed dustiness of the various aluminas used in our system. Our ambition is to improve our ability to allocate the right alumina to each smelter and to predict the dustiness of potential new alumina sources. An important element in this is to gain experience with respect to the ability of various types of dustiness test equipment to predict the actual dustiness of a given alumina. Further on we need to know more about the relationship between refinery technology, mode of operation and the dustiness of the alumina product. This combined knowledge and understanding is fundamental to the achievement of our ultimate goal of producing a less dusty alumina at the refineries.

2.0 LABORATORY ANALYSES AND TESTS

2.1 Analyses

Samples of all alumina shipments to Ardal (since late 1995) and Sunndal (since 1997) have been analysed at the Norsk Hydro Research Centre (HRE). The analytical results quoted in this paper are based on the following main techniques; Granulometry down to 45 Ám is done by dry sieving (Rotap). Sizes 20 Ám and smaller are analysed by laser (LALLS) on the –45 Ám sieve fraction. Due to small sample volumes, the granulometry of dust samples has been analysed directly by laser. Our alumina and Bayer process related analyses are mainly based on Alcan Standard Methods. All samples since June 1997 have been tested on the Perra Pulvimeter.

Table 1 shows number of shipments and averages for some main physical parameters for each source. Please note that some of the average values include shipments back to 1995 and may include data not longer representative of the refineries.

Table 1

Average alumina properties by source

Source

No. of shipments

-20 Ám

[%]

-45 Ám

[%]

A.I.

[%]

BET

[m2/g]

L.O.I.

[%]

A

8

0.41

5.0

22.1

81.6

0.88

B

12

1.09

5.9

19.6

75.7

0.87

C

7

0.47

3.2

18.5

78.0

0.73

D

8

1.77

8.0

14.0

68.8

0.78

E

6

0.83

5.8

15.4

71.8

0.77

F

5

0.42

3.8

23.8

72.6

0.76

G

2

0.40

4.9

N.A.

N.A.

N.A.

H

9

1.01

8.3

13.5

71.5

0.72

2.2 Dustiness Tests

A laboratory instrument that could give a reliable estimate of the dustiness during handling and smelter operations would be very useful both for R&D work, for evaluating potential new alumina sources and for detecting trends or changes in our commonly used sources. The Perra Pulvimeter (Perra 1984) is the most frequently mentioned instrument for this use in the alumina industry and is used extensively in our work. Our instrument was fabricated in-house as a copy of a unit previously borrowed from POSTEC Research in Porsgrunn. All our Perra data is obtained from this single instrument operated by the same person. Average relative standard deviation is 9.7%. We are also working on a dustiness tester based on measuring the weight of dust escaping from a fluid-bed of alumina, but we have experienced problems with instability due to fluctuations in moisture content of the fluidizing air. Due to this, all data quoted in this paper is based on the Perra instrument. However, in the periods when the fluid-bed unit has been stable, the results have been very promising, giving lower standard deviations (average 8.6%, including problem periods) and clearer differences between dusty and less dusty samples than the Perra unit. We have also tried to use the fluid-bed unit to study alumina attrition, but the operating conditions have not been rough enough to give significant breakdown.

Figure 1 shows Perra Dustiness Indices (D.I.) plotted against fines content (-20 Ám and –45 Ám). The upper part contains shipments samples from various sources. There is a weak trend of increasing D.I. with increasing fines, but due to many deviations the correlation coefficients are as low as 0.25 and 0.20. These deviations seem mainly to be related to source, especially sources "D" and "H". The lower part of Figure 1 shows similar results for one single source, taken from a plant test at the Alpart refinery in Jamaica. Here we find an unambiguous correlation between fines content and D.I., with correlation coefficients at 0.95 for –20 Ám and 0.80 for –45 Ám. These findings are in accordance with previous work (Authier-Martin 1989, Roach and Cornell 1990). Our data does not show any correlation between Perra D.I. and any other physical or chemical alumina quality parameters.

 


Figure 1

Perra D.I. versus fines content for multiple and single alumina sources

2.3 Electrostatics

Alumina in bulk acts as a conductor and does not retain electrical charges, but alumina particles in an aerosol can do so. Initial tests in a modified Faraday Cup indicated a strong negative correlation between the absolute charge-to-mass ratio of the dust, and the dustiness of the alumina as defined by the Perra Index. Further tests have not verified this as a general trend, but confirmed that one of the most dusty sources ("B") hardly builds up any charge at all, while source "D" which is much less dusty than what could be expected from the high fines content, builds the highest charge-to-mass ratio of all samples tested. One explanation may be that movement between particles builds up electrical charges due to friction (tribocharging) which creates interparticular forces. These forces make the particles stick together and thereby reduce the dustiness of the alumina.

We do not know the reason for this difference in electrostatic properties, but it is assumed to be a surface phenomenon. Our data shows no correlation between bulk alumina properties, for instance Fe2O3, and the tendency to build up electrical charge in the dust.

3.0 DUST MEASUREMENTS AT HYDRO ALUMINIUM SMELTERS

3.1 Measurements At Different Locations And Operations

Measurements of dustiness during unloading and pot operation have been done at the smelters in Ardal and Sunndal. Dustiness during unloading has been measured in the channel above the transport conveyor. Manual samples of dust and ore have been drawn for selected shipments, while several additional shipments have been measured by an on-line instrument detecting the degree of extinction of a laser beam. Dust emission from the potlines (roof emission) is measured regularly as a part of environmental regulations. Potroom dustiness has also been measured by laser instruments. However, since these instruments cover a very limited area of each potline and therefore are biased towards the conditions at a few cells close by, they may not be representative of the whole potline, and their data is not given much weight in this paper. All our measurements in connection with the smelter process itself have been done on Soederberg lines operating on primary alumina. The number of shipments having been followed up by various methods per 1 September 1998 is given in Table 2.

Table 2

Number of shipments followed up by various measurements

 

Ardal

Sunndal

Port, manual sampling

6

8

Port, continuous measurements

26

6

Port, dust samples

3

9

Potroom, continuous measurements

15

11

Potroom, roof emission

21

17

Perra Dustiness Index

14

13

During unloading at Ardal, it is always alumina from one single source that is most dusty, and another source that is least dusty. The remaining sources are grouped within these two extremes, without any apparent trends. At the other measuring points our data does not show any unambiguous link between alumina source and dustiness, although some sources are generally regarded as more dusty than others.

3.2 Initial Correlations With Bulk Alumina Quality Parameters

The content of fines is commonly thought to be the most important property with respect to dustiness. A selection of plots of observed dust levels versus contents of fines at various operations at the two plants are summarised in Figure 2. At the Ardal port we find no significant effect of fines, the dustiness seems to be mostly related to source. Furthermore, we find no obvious trend within each single source. At the remaining locations we find more or less clear trends of increased dustiness with more fines. However, there are many outliers, not all of them related to source. One should also be aware that due to source related characteristics, there is an apparent negative correlation between A.I. and fines content at arrival to port. This may disturb the picture because the fines content when the alumina reaches the pots may be different from the composite sample drawn during unloading.

Efforts to correlate the dustiness of each increment of 1000 mt during unloading with bulk alumina properties failed due to very small internal variations in the shipments compared to the magnitude of the effects and the quality of the measurements. Only one shipment had significant internal variations in the granulometry of the 1000 mt segments. In that case we found a clear correlation between the content of –10 Ám material and the dust level.

Attrition strength is the other parameter often related to alumina dustiness. Our data shows no relationship between dustiness and the Attrition Index during unloading at either of the two ports. If we look at the dust emission across the potline roofs, we find that the A.I. has a significant effect in Ardal, but not in Sunndal, see Figure 3. This is in accordance with local experience and can be explained by differences in the handling systems at the two plants. We have not found any significant correlation between any other physical or chemical quality parameter and the dustiness that can not be explained by the correlation of that single parameter with the fines content.


Figure 2

Dustiness vs. fines content

Figure 3

Dust emission vs. A.I.

3.3 Effects of Meteorological and Operating Parameters

Variations in meteorological and operational conditions may influence the dust level and hide the effect of alumina quality parameters. To investigate the magnitude of these disturbances we included key meteorological and operational data in the statistical analysis. The results indicate that meteorological and operational conditions can have an equally important influence on the dust level as the alumina quality. Some of the effects of weather conditions seem to be different at the two smelters, but this is probably due to very specific local conditions. The data can not support any model or estimate of these effects, but shows that they can be so strong that we should not expect to see very clear correlations between dust levels and alumina quality parameters alone.

During unloading at Ardal we find no significant correlation with meteorological data, the strongest is a negative correlation with atmospheric pressure at 0.39. The most influential parameter during smelter operation seems to be the number of anode effects with correlation coefficients against roof dust emission at 0.50 at Ardal and 0.61 at Sunndal. The number of anode effects can be influenced by several factors, also alumina quality. At Ardal we also find a significant positive correlation with ambient temperature (0.67) which is in agreement with local experience. Other apparent correlations with meteorological data are probably functions of their internal correlation with the temperature. There are also strong indications that the force and direction of wind have a big impact on the dustiness level in the potrooms. We are currently working on data to try to verify this.

3.4 Properties Of The Dust Fraction

Dust samples from 12 shipments have been drawn in the conveyor channel during unloading using so-called Sochslet nozzles. The samples have been analysed for physical, chemical and mineralogical properties and by SEM. Some key properties are listed in Table 3.

Table 3

Average properties of dust samples by source

Parameter\Source

Unit

"B"

"C"

"D"

"E"

"F"

"H"

No. of shipments  

3

3

1

1

2

2

Mean particle size

Ám

7

10

13

10

13

13

BET

m2/g

47

31

37

29

17

29

LOI

%

0.87

1.79

3.06

1.01

0.93

0.93

Alpha-alumina

%

24

45

20

N.A.

72

46

Average mean particle size for the dust is typically in the 10-14 Ám range. This is in accordance with previously published data (Roach and Cornell, 1990). However, source "B" which is the most dusty alumina during unloading at Ardal, and is generally considered to be a dusty alumina in our system, has a mean size of 6-9 Ám. SEM and thin section microscopy reveal that this alumina consists of very small agglomerates, loosely bound together without much cementation by growth. A.I. is around 18-20%. Our theory is that this alumina when breaking down forms very small particles with a high potential for dusting.

In Figure 4 the roof dust emission from the SuI Soederberg line at Sunndal is plotted against the mean size of the dust samples collected during unloading of the corresponding shipments. Apart from one single point ("D"), there is a clear negative trend between mean size of dust and emission. If we exclude this point as an outlier, the resulting correlation coefficient is 0.70. The resulting regression line is drawn in the graph.

Dust from source "B" has significantly higher BET than any of the other samples. It also has the lowest alpha content, except for the single sample from source "D". Gibbsite has only been detected in two dust samples, one from source "D" and one from source "H", at 14 and 5% respectively. SEM pictures of selected samples indicate that dust particles from sources "B" and "C" are relatively round, while dust particles from "D" and "F" have sharper edges.

Figure 4

Sunndal roof emission vs. median size of dust

4.0 PERRA INDICES VERSUS OBSERVED DUST LEVELS AT SMELTERS

In addition to accuracy and reproducibility, the real test of any dustiness testing equipment lies in its ability to provide reasonably correct predictions of the dustiness of a given alumina sample during handling and/or smelter operations. Figure 5 shows measured dust levels at the three previously mentioned locations at Ardal and the port in Sunndal plotted against Perra D.I. for the respective shipment samples. We observe general trends of increased measured dust levels with increasing D.I., but there are also several outliers. It is interesting to observe that the Perra instrument seems to correlate reasonably well also with dustiness during pot operation, in spite that it is mostly regarded as an instrument for predicting handling dustiness. However, one should remember that our data is limited to Soederberg cells operating on primary alumina.


Figure 5

Plant data versus Perra D.I.

5.0 SUMMARY OF PLANT OBSERVATIONS

Table 4 summarises correlation coefficients between measured dust levels at the plants and selected alumina quality parameters, operational parameters, meteorological data and Perra D.I.

Table 4

Summary of correlation coefficients

Parameter\Location

Ardal Port

Ardal Roof

Sunndal Port

Sunndal Roof

-45 Ám

-0.33

0.41

0.38

0.37

-20 Ám

0.11

-0.12

0.19

0.51

A.I.

0.16

0.73

0.10

-0.07

No. of anode effects

N.A.

0.50

N.A.

0.61

Ambient temperature

-0.29

0.67

N.A.

0.06

Atmospheric pressure

-0.39

0.17

N.A.

-0.20

Perra D.I.

0.64

0.55

0.71

0.47

Even if we do not find any clear, general correlation between fines content and alumina dustiness, our data still shows that fines and superfines are the two most important alumina quality parameters with respect to dustiness. The negative correlation with –20 Ám at Ardal Roof is probably due to a combination of the previously mentioned apparent negative correlation between A.I. and fines content at arrival, and the strong influence of A.I. at Ardal.

The influence of operational and meteorological parameters can be more than strong enough to hide the effects of alumina quality on dustiness.

The Perra D.I. generally gives a better prediction of alumina dustiness than any of the traditional alumina quality parameters, also in our Soederberg potlines. However, the strong sensitivity due to individual instruments and operators is a severe disadvantage with the Perra Pulvimeter.

6.0 IMPORTANT FACTORS IN THE ALUMINA REFINERY

As part of a joint study by Alpart, Kaiser and Hydro, plant tests were run at the Alpart plant in 1997 to study the effect of precipitation mode on hydrate and alumina quality. The tests gave important background and documentation for the modification of Alpart’s Precipitation circuit that is described in a paper by Alpart at this workshop. The results show that the dominating factor determining the alumina dustiness as defined by the Perra Index was the strength of the hydrate, in our case defined by the A.I. Figure 6 shows fines content in alumina plotted against the hydrate A.I. and indicates an almost perfect correlation. The corresponding effect on alumina dustiness is shown in the single source part of Figure 1. The differences in hydrate strength between the various test modes can be explained by well-known theories. New tests will be run when the plant has stabilised after the modifications. Alpart considers reporting the entire study at a later event.

The internal structure of the hydrate has a major influence on the strength of the hydrate, the fines content of the alumina, and also what kinds of particles are formed during breakdown. It is our firm belief that this is the primary key to alumina dustiness.

Figure 6

Alumina fines versus hydrate A.I. (Alpart study)

7.0 CONCLUSIONS

Observed dustiness during various operations at our smelters is influenced by numerous parameters such as alumina source, alumina quality parameters, handling systems, smelter technology, operational conditions and weather. Several of these may mask the effects of the alumina quality.

At most of our measuring points at the smelter plants, we find a trend of increased dustiness with increased fines content. Even if we can not find any clear, general correlation between fines and alumina dustiness, it is still the most important alumina quality parameter with respect to dustiness.

Data from Perra tests on alumina samples from one single source, covering a wide range of fines content, shows a clear correlation between fines content and dustiness index.

The effect of alumina A.I. depends strongly on the handling system at each smelter.

We have observed a negative correlation between the mean size of the dust particles sampled during unloading and the dust emission from the potroom.

There are indications that electrostatic properties of the dust fraction affect the dustiness.

Our data and experience so far shows that the Perra D.I. gives a better prediction of alumina dustiness than any of the traditional alumina quality parameters.

Plant tests at Alpart show an almost perfect correlation between hydrate strength and alumina fines content.

8.0 FURTHER WORK

Based on our data, experience and discussions we strongly believe that the strength and structure of the hydrate is the key to alumina dustiness. We need to learn more about the link between White Side conditions, hydrate properties and further on to alumina properties and dustiness.

It seems like the dustiness is closer correlated to properties of the dust fraction itself (granulometry, shape, electrostatics, etc.) than to the bulk alumina properties. We need a better understanding of the link between bulk alumina and dust properties. Our working hypothesis is that the structure of the hydrate can explain a major part of this relationship, because it determines the way hydrate and alumina is broken down during and after calcination. The A.I. alone does not tell the whole story, one has to consider the entire attrition curve, not only the –45 Ám fraction, and look at what kind of particles are formed during breakdown.

We will continue to build up experience regarding the correlation between various dustiness testers and measured dustiness at our smelters.

In addition to our internal work, Hydro Aluminium has signaled its support to the planned AMIRA project "Gibbsite to Alumina Quality" which hopefully will be up and running at the time of this workshop.

 

ACKNOWLEDGMENTS

The author wants to thank the entire project team; A. Bruusgaard, A.K. Espeseth, K. Hamberg, N. Hegna, M.S. Konanur, H.D. Soerensen, J. Toft and E. Tveten for their dedicated work. I would also like to thank our client J.A. Larsen and my supervisor B. Lillebuen for their invaluable interest and support. Our contacts at the smelters should also be thanked for all their help and assistance. Finally I will thank ALPART and Kaiser Aluminum for the permission to quote data from our joint study.

REFERENCES

Authier-Martin, M. (1989). Alumina Handling Dustiness. TMS Light Metals 1989, pp.103-111

Hsieh, H.P. (1987). Measurement Of Flowability And Dustiness Of Alumina. TMS Light Metals, pp. 139-149

Perra, S. (1984). Measurement of sandy alumina dustiness. TMS Light Metals 1984, pp. 269-286

Roach, G.I.D. and Cornell, J.B. (1990). Dust And Dustiness Testing Of Smelter Grade Alumina. Proc. Second International Alumina Quality Workshop, pp.246-262