The rotating drum dustiness tester: Variability in dustiness in relation to sample mass, testing time, and surface adhesion

The rotating drum dustiness tester: Variability in dustiness in relation to sample mass, testing time, and surface adhesion

PII: S0003-4878(99)00049-6 Ann. occup. Hyg., Vol. 43, No. 8, pp. 557±566, 1999 # 1999 British Occupational Hygiene Society Published by Elsevier Scie...

186KB Sizes 0 Downloads 10 Views

PII: S0003-4878(99)00049-6

Ann. occup. Hyg., Vol. 43, No. 8, pp. 557±566, 1999 # 1999 British Occupational Hygiene Society Published by Elsevier Science Ltd. All rights reserved Printed in Great Britain. 0003±4878/99/$20.00

The Rotating Drum Dustiness Tester: Variability in Dustiness in Relation to Sample Mass, Testing Time, and Surface Adhesion N. O. BREUM* National Institute of Occupational Health, Lersoe Parkalle 105, DK 2100, Copenhagen 0, Denmark A rotating drum dustiness tester was used to characterize variability of dustiness in dependence of type and mass of test material, testing time, and surface adhesion. Powders of six common materials entered the study: bentonite, barium sulphate, talc, Aloxite, carbon black, and coal. Except for coal, dustiness was in general positively correlated to the mass of powder under testing. Surface adhesion tended to be a€ected from the type of test material. For a ®xed dust dispersion time (180 sec) the dependence of dustiness on time was characterized in terms of the time required to arrive at the median of the cumulative distribution of mass delivered at the outlet of the drum. In general the time required was positively correlated to the mass of material under testing. A three-parameter multiplicative model for dustiness potential was developed for two of the test materials (bentonite and barium sulphate). The model included surface adhesion, time, and mass of material under testing as predictors. The model was highly signi®cant ( p < 0.001) and accounted for more than 80% of the observed variation in dustiness. It is concluded that for dustiness testing to become useful a careful control of all the operating parameters are required in order to have reproducible tests. Therefore standardization of the method is essential. # 1999 British Occupational Hygiene Society. Published by Elsevier Science Ltd. All rights reserved. Keywords: dustiness; surface adhesion; rotating drum

INTRODUCTION

does not characterise the generated dust in the inhalable, thoracic and respirable fractions as de®ned by European Standard EN 481 (`Size fraction de®nitions for the measurement of airborne particles in the workplace'). The present study focuses on a rotating drum dust generator, collecting the dust on a ®lter. The purpose of the study was to characterize variability of dustiness and the dependence on type and mass of test material, testing time, and surface adhesion.

Dustiness of a material is de®ned as the tendency of dry materials to liberate dust into the air when handled under speci®c conditions (BOHS, 1985). Dustiness testing, therefore, is empirical and the results are method-dependent (Chung and Burdett, 1994). A dustiness tester consists of two partsÐa dust generator and a dust sampler. Three di€erent principles are common in the design of a dust generator for dustiness testing: gravity (drop methods), mechanical dispersion (rotating drum methods) and gas dispersion (¯uidized bed methods). The dust sampler, which is the measurement stage, can be a variety of systems dependent mainly of the ease of use and cost. A simple method for quantifying the released dust is sampling on a ®lter, although this

DESCRIPTION OF TEST EQUIPMENT

The aerosol generator A rotating drum was used for generating dust from the selected test materials (see below: study design). In principal the design of the drum was similar to the WSL drum dustiness apparatus (BOHS, 1985), see Fig. 1. The all stainless steel drum (0.30 m internal dia) had removable conical

Received 3 March 1999; in ®nal form 16 April 1999. *Tel.: +45 3916 5299. 557

558

N. O. Breum

Fig. 1. Rotating drum dustiness tester (not drawn to scale).

ends with six lifter bars inside the cylindrical part. The length of the drum was 0.46 m for the cylindrical part and 0.11 m for a conical end (V = 40 L volume in total). Air was exhausted at a controlled ¯ow rate of Q = 50 L minÿ1 from one end of the drum (0.02 m dia outlet opening) and excess of ®ltered air was delivered at an inlet opening (0.03 m dia) at the opposite end of the drum. Note that the inlet was kept at atmospheric pressure. A humidity generator kept the relative humidity of the supplied air at 50%. For this study the rotation speed of the drum was set at 40 rpm and the test period was set at 180 sec. A material under testing was loaded onto the bottom of the drum (at center of the drum). Dust from the material was generated upon rotation of the drum, and a sampler collected the dust at the outlet.

The dust sampler The dust sampler at outlet of the drum had two parts in parallel: a preweighed Whatman GF/C ®lter (90 mm dia) and a dust monitor to give the dust concentration against time. An isokinetic probe in front of the ®lter delivered a constant air¯ow (1.9 L minÿ1) to the dust monitor (see below: dustiness as a function of time) and the remaining air¯ow (48.1 L minÿ1) passed the ®lter for collection of the aerosols. The mass of dust collected onto the ®lter was determined by re-weighing the ®lter at a detection limit of 0.45 mg. The dustiness (as a percentage) of a test material was estimated as the mass of dust collected in proportion to the mass of material under testing. Note that the dustiness was adjusted by a factor of 1.04 (=50 L minÿ1/48.1 L minÿ1) to

take into account the sub-sample of air delivered to the dust monitor. Dustiness as a function of time The TEOM monitor (Tapered Element Oscillating Microbalance) was used for recording the dust concentration versus time at the outlet of the drum. The monitor is an instrument that draws air through a ®lter at constant ¯ow rate. The gain in mass of the ®lter is derived from a change in oscillating frequency and approximately real-time (5 sec time resolution) mass concentrations are obtained. It is noted that the data obtained were not corrected for surface deposition of dust in the sampling line leading from the outlet of the rotating drum to the monitor. For a given test material the cumulative distribution of mass collected was plotted against time. The data obtained were summarized in terms of the time (median time) required to arrive at the median of the cumulative distribution. Surface adhesion In dustiness testing part of the released dust may stick to surfaces in the drum due to electrostatics and cohesion. To quantify the surface deposition, gelatine foils were used to lift dust from surfaces according to the sampling method described by Schneider et al. (1996). Three positions were chosen, with position 1 at the conical inlet of the drum, position 2 at center of the top of the cylindrical part after it has come to rest, and position 3 at the conical outlet of the drum. One imprint was taken at each position and overlap was corrected for as

Fig. 2. Dustiness versus type and mass of di€erent test materials. Vertical bars represents 95% con®dence. Pearson's correlation coecient (r ) is given in the ®gure.

The rotating drum dustiness tester 559

560

N. O. Breum

outlined by Schneider et al. (1996). At the outset of an experiment a set of foil imprints were taken from the clean drum. The surface deposition was given as the percentage of area covered with dust, and data of an experiment were summarized in terms of the mean of the samples from the three positions.

STUDY DESIGN

A balanced randomized factorial design was used for the study. Six di€erent test materials were included: bentonite, barium sulphate, talc, Aloxite, carbon black, and coal. A given material was tested (in triplicate) at four di€erent levels of mass in the rotating drum: 25, 50, 100, and 200 g. Samples of material for testing were produced (spinning ri‚er method) at the Institute of Occupational Health, University of Birmingham. The test materials were delivered to the laboratory in bottles with 200 g of material, and the sub-samples required were produced with a spinning ri‚er method (Retsch Divider model PT with a vibratory feeder model DR 40). The data obtained were log-transformed and tested (Anderson±Darling test) for normality at a 5% level of statistical signi®cance. Bartlett's test was performed for the homogeneity of variance between groups of data. To achieve homogeneity of variance some groups (see below) were to be considered outliers leaving an unbalanced design, and a general linear model (GLM) was used to study the e€ect of the type and mass of test material on the dustiness, the surface adhesion, and the time dependence. The standardardized residuals were tested for normality (Anderson±Darling test) at a 5% level of statistical signi®cance. In case of signi®cant evidence of interaction between the type and mass of test material Pearson's correlation coecient was used to characterize the e€ect of mass on the dustiness, the surface adhesion, and the time dependence. Multiple regression was used to characterize variability of dustiness in dependence of the time, the surface adhesion, and the type and mass of test material. The standardardized residuals were tested for normality (Anderson±Darling test) at a 5% level of statistical signi®cance. Minitab software was used for the statistical analysis (Minitab release 10Xtra).

RESULTS

Dustiness versus type and mass of test material For an given mass and test material normal plotting and testing (Anderson±Darling test) indicated that the log-transformed dustiness data were normally distributed within the block of data, but Bartlett's test indicated heterogeneous variances

( p = 0.001) between the blocks of data. Visual inspection of the data (Fig. 2) indicated abnormal variances for three blocks of data: talc (200 g), Aloxite (25 g), and carbon black (25 g). Considering these three blocks as outliers Bartlett's test indicated ( p = 0.21) homogeneity of variance between all blocks of data. A general linear model was used to study the e€ect of the type and mass of material on the dustiness. The model indicated that the type and mass of test material had a signi®cant e€ect on the dustiness ( p < 0.005). The type and mass of test material interacted ( p < 0.005) and for each type of material Pearson's correlation coecient was used to characterize the e€ect of mass on the dustiness. The data obtained are plotted in Fig. 2, and in general dustiness was positively correlated to the mass of material under testing. Dustiness and surface adhesion For an given mass and test material normal plotting and testing (Anderson±Darling test) indicated that the log-transformed surface adhesion data were normally distributed within the block of data, but Bartlett's test indicated heterogeneous variances ( p = 0.001) between the blocks of data. Visual inspection of the data (Fig. 3) indicated abnormal low variances for three blocks of data: talc (50 g), talc (100 g), and coal (25 g). Considering these three blocks as outliers Bartlett's test indicated ( p = 0.12) homogeneity of variance between all blocks of data. A general linear model was used to study the e€ect of the type and mass of material on the dustiness. The model indicated a tendency ( p = 0.07) of surface adhesion to be a€ected from the type of test material. Data on the surface are plotted in Fig. 3 against the type and mass of test material and Pearson's correlation coecient are given in the ®gure. Dustiness as a function of time For a given type and mass of a test material the concentration of dust versus time was measured at the outlet of the rotating drum. The data obtained were plotted and abnormal pro®les (e.g., low concentrations) were observed for some cases (N = 11). These cases were considered outliers leaving an unbalanced design of the experiments. For each of the remaining experiments the cumulative distribution of mass collected was plotted against time. A plot was normalized with the total mass of dust collected and typical results are given in Fig. 4. As observed from Fig. 4 there was a delay (approx. 20 sec or more) in the measured mass of dust at outlet of the drum (see discussion). The cumulative distributions were summarized in terms of the time required to arrive at the median of the distributions. For a given mass and test material normal plotting and testing (Anderson±Darling test) indicated that

Fig. 3. Surface deposition of dust in the rotating drum tester versus type and mass of di€erent test materials. Vertical bars represents 95% con®dence. Pearson's correlation coecient (r ) is given in the ®gure.

The rotating drum dustiness tester 561

Fig. 4. Typical cumulative distributions of mass of dust arriving at outlet of the rotating drum tester. The distribution is for a test period of 180 sec and a sample mass of 25 g.

562 N. O. Breum

Fig. 5. Time required to arrive at the median of the cumulative distribution of the mass of dust arriving at outlet of the rotating drum tester (180 sec test period). Vertical bars represents 95% con®dence or the range in data. Pearson's correlation coecient (r ) is given in the ®gure.

The rotating drum dustiness tester 563

564

N. O. Breum

the log-transformed median times were normally distributed within the block of data, but Bartlett's test indicated heterogeneous variances ( p = 0.001) between the blocks of data. Visual inspection of the data (Fig. 5) indicated an abnormal variance for one block of data (Aloxite, 200 g). Considering this block as an outlier Bartlett's test indicated ( p = 0.075) homogeneity of variance between all blocks of data. A general linear model (GLM) was used for the statistical analysis of the data and the type and mass of test material proved to have a signi®cant ( p < 0.05) in¯uence on the time (median time) required to arrive at the median of the cumulative distribution of the mass collected. The type and mass of test material interacted ( p < 0.005) and for each type of material Pearson's correlation coecient was used to characterize the e€ect of mass on the dustiness. The data obtained are plotted in Fig. 5, and in general the median time was positively correlated to the mass of material under testing. Dustiness as a function of type and mass of test material, time, and surface adhesion Let the dustiness for a given material (No. N) be denoted DN. The actual mass of test material normalized to the maximum of mass under testing (200 g) is denoted MN. The median time normalized to the test period (180 sec) is denoted a MTN and the surface adhesion (%) is denoted SAN. An analysis of multiple linear regression was used to explain the variation in DN from the three predictors (MN, MTN and SAN). To keep homogeneity of variance some blocks of data were to be considered outliers (see above), and for some materials few cases were left for the analysis. A full set of data (N = 12) was left for bentonite, and for barium sulphate one case was missing due to a missing observation on the median time. The data for bentonite and barium sulphate were ®tted to a multiplicative model: DN ˆ a  MbN  MTdN  SAlN The model was highly signi®cant ( p < 0.001) for both test materials, and the obtained coecients for the model are summarized in Table 1. DISCUSSION

Dustiness of powders is of importance for the air

quality in the workplace. A risk assessment of personal dust exposure should be based on the biologically relevant size fractions; the inhalable, the thoracic and the respirable fraction. Several methods are available for dustiness testing of powders (Chung and Burdett, 1994) but except for one (HSE, 1996) none of the methods include size separation into the three relevant aerosol fractions. The HSE-method is a modi®ed version of the existing WSL rotating drum and the size selectors are cylindrical plugs of porous polyurethane foams. Unfortunately the foams may have a low dust holding capacity and the obtained thoracic and respirable fractions of dust may not be valid for dusty materials (Breum, 1999). Therefore the WSL rotating drum was used for the present study. This drum was suitable for testing both dusty and less dusty materials but it has to be noted that the results were not characterized in terms of the size of particles arriving at the dust sampler. The detection limit of the method depends on the sample mass and the limit was 0.0009% for a sample mass of 50 g. The membrane ®lter used to collect the dust was in a vertical position and there is a risk to loose some dust in removing the ®lter from the drum. Therefore the upper limit of the method depends on the procedure in handling the ®lter. For the test materials of the present study no losses were observed in handling the ®lter and it seems safe to assume that the method can be used for a dustiness at least up to a level of 5%. Thus the method has a dynamic range of at least a factor of 5500. The rotating drum is a versatile approach in dustiness testing but for a given test material the data obtained are in¯uenced from several parameters including air ¯ow rate in the drum, rotation speed of the drum, the selected test period, and the mass of material under testing (Hjemsted and Schneider, 1996). This study focused on the type and mass of a test material as governing parameters for the dustiness, the time dependence in dustiness, and the surface adhesion. Within a material the dustiness was in general positively correlated to the mass of test material. Recently Lyons and Mark (1994) did a comprehensive study on some of the main operating parameters of a modi®ed WSL drum. They used Aloxite as a test material and consistent with the present study dustiness was positively correlated to the mass of material under testing. Hjemsted and

Table 1. Estimated coecients for a multiplicative model of dustiness Test material Bentonite Barium sulphate *p < 0.05;. **0.05 < p < 0.10.

No. 1 2

a

b

d

l

R2

1.46* 0.093

0.37* 1.25*

1.50** ÿ0.39

ÿ0.079 ÿ0.20

0.80 0.87

The rotating drum dustiness tester

Schneider (1996) did a similar study on the WSL drum and observed that dustiness was negatively correlated to the mass of material under testing. Heitbrink (1990) characterized some of the main operating parameters for another rotating drum dustiness tester (Heubach) and observed that dustiness was positively correlated to the mass for a mass of test material (limestone) at or below 100 g. The data available on the mass of test material as a main parameter for dustiness suggest the need to carefully control the mass to have reproducible dustiness tests. For a modi®ed WSL drum Lyons and Mark (1994) observed less surface adhesion of coarse Aloxite (grade F360) as compared to a ®ner grade (F1200). Consistent with this observation, data of the present study indicated that surface adhesion was a€ected from the type of test material (Fig. 3). For some test materials (e.g., bentonite) surface adhesion was low compared to other materials (e.g., coal). The statistical analysis did not indicate the mass of test material as a governing parameter for the surface adhesion, but for some test materials (barium sulphate, aloxite or carbon black) surface adhesion was positively correlated to the mass loading while the surface of the drum was saturated with dust even for a low mass loading (25 g) of other materials (bentonite or coal). It is noted that Hjemsted and Schneider (1996) did not observe an in¯uence from the mass of test material (alumina powder) on the surface adhesion in the WSL drum. In dustiness testing surface adhesion is of importance for the mass of dust delivered to the dust sampler. Therefore the data available on the surface adhesion suggest the need to carefully control the surface of a dustiness tester to have reproducible results. For a testing time of more than approx. 20 sec Lyons and Mark (1994) observed no in¯uence of the dust dispersion time on the dustiness of a test material (Aloxite, grade F360). They ®tted a transparent window to the non-sampling end of a modi®ed WSL drum tester and observed ``that a dust cloud was produced as the drum rotated, but after a few seconds the walls of the drum became coated with a thin layer of particles. On further rotation of the drum the remaining airborne particles stuck to those already deposited on the walls and no further dust was dispersed. This did not happen when a coarser grade of Aloxite (F230) was used, as the weight of the particles was suciently large to overcome the adhering forces''. Recently Hjemsted and Schneider (1996) observed that dustiness of aluminia was positively correlated to the dust dispersion time. Consistent with the data reported by Lyons and Mark (1994) and Hjemsted and Schneider (1996) the present study indicated (Fig. 4) that dust generated from some materials (e.g., bentonite) arrived almost continuously at the dust sampler

565

while the dust from other materials (e.g., talc) arrived almost like a burst. From the start of an experiment there was a delay in the mass of dust arriving at the outlet of the drum (Fig. 4). The delay might involve time constants of the drum and the TEOM. The time constant of the drum was 48 sec (=V/Q ). A short sampling line (volume approx. 5 cm3) was used for the TEOM leading to a time constant of 0.1 sec. Therefore the drum dominated the time constants but the time constant on its own does not explain the observed delay in the measured mass and this indicate that other mechanisms than the air ¯ow remove airborne dust generated in the drum. For aluminia as a test material Hjemsted and Schneider (1996) observed almost no interaction between sample mass and the dust dispersion time, but from the present study (Fig. 5) the interaction can be substantial for some types of material (e.g., talc). Therefore the data available on the dust dispersion time suggest the need to carefully control the dispersion time of a dustiness tester to have reproducible results. As reviewed by Hjemsted and Schneider (1996) dustiness as derived from a rotating drum tester is in¯uenced from several parameters including the dust dispersion time, the air ¯ow rate in the drum, the mass and moisture content of the test material, the size distribution of the test material, and the surface adhesion. For this study an empirical model was developed for dustiness of two speci®c test materials in dependence of three variables: the mass of test material (M ), the surface adhesion (SA ), and the time required to arrive at the median of the cumulative mass distribution (MT ). In terms of R 2 the model proved useful to predict the dustiness (Table 1), but the model was sensitive to the type of test material. For one material (barium sulphate) only one of the variables (mass) turned out to be statistically signi®cant while two variables (mass and median time) were signi®cant predictors for the other material (bentonite). Although the surface adhesion was a non-signi®cant predictor for both test materials it has to be expected that surface adhesion is of importance for the mass of dust delivered to the dust sampler and surface properties should not be ignored in the design of a dustiness tester. For general use an essential requirement for a standard test method for dustiness is that the results should be reproducible in di€erent laboratories and for this to be possible the equipment and procedure needs to be described in sucient detail. From the present study it is apparent that standardization of the drum method is required in terms of the surface of the drum, the dust dispersion time, and the mass of material under testing. This study was restricted to include three main operating parameters of the drum but it is emphasized that standardization also is required for other parameters of the drum includ-

566

N. O. Breum

ing rotation speed, air ¯ow rate in the drum, and humidity of the air delivered to the drum.

CONCLUSION

A three-parameter multiplicative model for dustiness potential was developed for two speci®c test materials. The model included surface adhesion, time, and mass of material under testing as predictors. The model was highly signi®cant and accounted for more than 80% of the observed variation in dustiness. For dustiness testing to become useful a careful control of all the operating parameters are required in order to have reproducible tests. Therefore standardization of the method is essential.

AcknowledgementsÐThe study is part of a project supported by the European Commission (contract SMT4CT96-2074 ``Development of methods for dustiness testing'').

REFERENCES Breum, N. O. (1999) The dust holding capacity of porous plastic foam used in particle size-selective sampling. J. Aerosol Sci. (in press). BOHS (1985) Dustiness Estimation Tests for Dry Materials. In British Occupational Hygiene Society Technical Guide No. 4. Science Reviews Ltd, Northwood, ISBN 0 905927 71 0. Chung, K. Y. K. and Burdett, G. J. (1994) Dustiness testing and moving towards a biologically relevant dustiness index. Ann. Occup. Hyg. 38, 945±949. Heitbrink, W. A. (1990) Factors a€ecting the Heubach and MRI dustiness tests. Am. Ind. Hyg. Assoc. J. 51, 210±216. Hjemsted, K. and Schneider, T. (1996) Documentation of a dustiness drum test. Ann. Occup. Hyg. 40, 627±643. HSE (1996) Dustiness of powders and materials. In Methods for the Determination of Hazardous Substances, MDHS 81. Health and Safety Executive, UK. Lyons, C. P. and Mark, D. (1994) Development and testing of a procedure to evaluate the dustiness of powders and dusts in industrial use. HSE Contract Research Report No. 62/1994. HSE Books (ISBN 0 7176 0727 5, HMSO). Schneider, T., Petersen, O. H., Kildesù, J., Kloch, N. P. and Lùbner, T. (1996) Design and calibration of a simple instrument for measuring dust on surfaces in the indoor environment. Indoor Air 6, 204±210.