Determinants of Exposure to Dust and Dust Constituents in the Norwegian Silicon Carbide Industry

ABSTRACT

Introduction: The aim of this study was to identify important determinants of dust exposure in the Norwegian silicon carbide (SiC) industry and to suggest possible control measures.

Methods: Exposure to total dust, respirable dust, quartz, cristobalite, SiC, and fiber was assessed in three Norwegian SiC plants together with information on potential determinants of exposure. Mixed-effect models were constructed with natural log-transformed exposure as the dependent variable.

Results: The exposure assessment resulted in about 700 measurements of each of the sampled agents. Geometric mean (GM) exposure for total dust, respirable dust, fibers, and SiC for all workers was 1.6 mg m−3 [geometric standard deviation (GSD) = 3.2], 0.30 mg m−3 (GSD = 2.5), 0.033 fibers cm−3 (GSD = 5.2), and 0.069 mg m−3 (gSD = 3.1), respectively. Due to a large portion of quartz and cristobalite measurements below the limit of detection in the processing and maintenance departments (>58%), GM for all workers was not calculated. Work in the furnace department was associated with the highest exposure to fibers, quartz, and cristobalite, while work in the processing department was associated with the highest total dust, respirable dust, and SiC exposure. Job group was a strong determinant of exposure for all agents, explaining 43–82% of the between-worker variance. Determinants associated with increased exposure in the furnace department were location of the sorting area inside the furnace hall, cleaning tasks, building and filling furnaces, and manual sorting. Filling and changing pallet boxes were important tasks related to increased exposure to total dust, respirable dust, and SiC in the processing department. For maintenance workers, increased exposure to fibers was associated with maintenance work in the furnace department and increased exposure to SiC was related to maintenance work in the processing department.

Conclusion: Job group was a strong determinant of exposure for all agents. Several tasks were associated with increased exposure, indicating possibilities for exposure control measures. recommendations for exposure reduction based on this study are (i) to separate the sorting area from the furnace hall, (ii) minimize manual work on furnaces and in the sorting process, (iii) use remote controlled sanders/grinders with ventilated cabins, (iv) use closed systems for filling pallet boxes, and (v) improve cleaning procedures by using methods that minimize dust generation.

METHODS

Sampling strategy and job groups

Exposure to total dust, respirable dust, fibers, quartz, cristobalite, and SiC was assessed in three Norwegian SiC plants during 2002–2003, resulting in around 700 measurements of each of the agents. Each plant was sampled twice in different seasons to evaluate seasonal and process-related changes in exposure. Each worker was, if possible, sampled twice in both periods to assess between-worker (BW) and within-worker (WW) variability. The workers were divided into job groups according to department (furnace, processing, or maintenance department) and tasks performed. The groups consisted of workers in the furnace department who are involved in the production of SiC (mix operator, charger, charger/mix operator, pay loader operator, crane operator, control room operator, cleaner, and sorting operator), workers processing SiC (crusher operator, other refinery operator, and fines operator), and maintenance workers (mechanics and electricians), see Table 1. A random sample of workers from each job group was measured. A more detailed description of the sampling strategy and job groups has been reported previously (Føreland et al., 2008).

Determinants

Information on potential determinants within each plant, department, and job group was obtained from several sources: (i) the workers themselves, (ii) the industrial hygienist responsible for the sampling, and (iii) foremen in the furnace and processing departments. The workers provided information about type and duration of tasks performed during sampling by filling out a plant- and department-specific form. The industrial hygienist did a walk through survey of the premises and recorded information such as type of equipment used and organization of work. The foreman registered information on department-specific production parameters on each shift when sampling was performed.

Air sampling and analysis

Full shift total dust samples were collected on 37-mm cellulose acetate filters with a 5.0-µm pore size (Millipore Corporation, Billerica, MA, USA), fitted in 37-mm closed-faced aerosol filter cassettes (Millipore Corporation) and a sampling flow rate of 2 l min −1 was applied. Respirable dust was collected on 37-mm cellulose acetate filters with a pore size of 5.0 µm (Millipore Corporation) using cyclones (Casella T13026/2, London, UK) at a sampling flow rate of 2.2 l min−1. The masses of all air filters were measured gravimetrically using a Sartorius AG, MC 210p laboratory balance (Göttingen, Germany), limit of detection (LOD) 0.06 mg.

The quartz, cristobalite, and SiC contents of the respirable dust were determined by X-ray powder diffractometry [Philips PW1729 X-ray generator, Philips PW 1710 diffractometer control, and Philips APD software (PANalytical, Almelo, The Netherlands)]. Crystalline silica was determined by standard methods modified for quartz due to the presence of graphite in the furnace hall (Bye, 1983; NIOSH, 1998). SiC was determined by a corresponding method (Bye et al., 2009). LOD for quartz was 5 µg, for cristobalite 10 µg, and for SiC 12 µg. Respirable dust samples were combined prior to analysis of crystalline constituents if the amount of dust in each sample was not sufficient for analysis (<0.7 mg). Samples within plant and job groups were combined, if possible, from the same persons. A total of 272 combined samples were analyzed from the 680 respirable dust samples. Of the combined samples, 64% were from different persons, but from the same job group. The airborne concentrations of the crystalline constituents in the individual samples were then obtained by multiplying the concentration of respirable dust with the percentage of crystalline constituents detected in the combined sample.

Fibers were collected on 25-mm cellulose acetate filters (Millipore Corporation) with 1.2-µm pore size using open face aerosol filter cassettes of conducting polypropylene (Gelman Sciences, Ann Arbor, MI, USA) and 1 l min −1 flow rate. Fibers were counted by light microscopy according to World Health Organization counting criteria (WHO, 1997). The LOD was four fibers for the analyzed area of the filter.

Data analysis

A total of 40% of the fiber, 55% of the quartz, 73% of the cristobalite, and 7.6% of the SiC samples were below LOD. Due to the large proportion of samples below the LOD for quartz and cristobalite, only the samples from the furnace department were used for regression modeling (29 and 36% below LOD, respectively). Measurements below LOD were imputed using a maximum-likelihood estimation method (Lubin et al., 2004; Vermeulen et al., 2010). The imputation procedure was repeated nine times giving nine data sets consisting of different (imputed) values for measurements below LOD. Each of these data sets was used in the modeling procedure for fiber, SiC, quartz, and cristobalite. Proc Mianalyze (SAS institute) was used to obtain valid estimates and standard errors for the estimated parameters. BW and WW variance was summarized by computing median values.

Cumulative probability plots of total dust and respirable dust showed that the exposure data were best described by lognormal distributions and the exposure data were log e transformed before statistical analysis.

Standard measures of central tendency and distributions [arithmetic mean (AM), geometric mean (GM), geometric standard deviation (GSD), and 95th percentiles] were calculated.

Workers were grouped by job group, department, and plant. Thirty-three workers changed job group between samplings, and the worker was then treated as a new person.

Separate models were constructed for total dust, respirable dust, fiber, quartz, cristobalite, and SiC. To evaluate determinants of exposure, mixed-effect regression models were constructed using Proc Mixed (SAS institute), with exposure as the dependent variable. Determinants were treated as fixed effects and worker as a random effect. The restricted maximum-likelihood algorithm was used to estimate variance components due to the unbalanced nature of the data. Univariate models were first performed after which multivariate models were built by forward stepwise selection starting with the variable with the lowest P-value in the univariate model. Variables with P-values >0.2 in univariate models were excluded from further analysis. Akaike’s information criterion was used to select the optimal combination of exposure determinants in the multivariate model. Determinants were modeled on a general level where all measurements were included, on a department level with separate models for each department, and on a job group level with separate models for each job group. Plant, department, and job group were the possible determinants for all measurements on the general level. Job group, plant, season, and the process-related parameters given in Table 1 were the determinants explored on a department level. The determinants explored on a job group level were task and location of the sorting area (inside furnace building or separate building) (Table 1).

Tasks were modeled as a dichotomous variable (task performed yes/no). Tasks with a frequency of n < 4 were excluded from the analyses.

In order to quantify the contribution of the fixed effects to the BW and WW variance components, values of the variance components obtained using the mixed-effect model were compared with those from a mixed-effect model without fixed effects and the reduction expressed by percentage change.

SAS version 9.2 (SAS institute Inc., Cary, USA) and SPSS version 18.0 for Windows (SPSS Inc., Chicago, IL, USA) were used for statistical analyses.
RESULTS

Of the 293 workers participating in the exposure study, 75–78% (depending on exposure agent) was monitored on more than one occasion for the same agent. The GM exposure for total dust, respirable dust, fibers, and SiC for all workers was 1.6 mg m−3, 0.30 mg m−3, 0.033 fibers cm−3, and 0.069 mg m−3, respectively (Tables 2 and 3). The GM exposure for quartz and cristobalite in the furnace department was 2.2 and 3.7 µg m−3 , respectively (Table 3). The crystalline silica exposures for workers in the processing and maintenance departments were generally low. More than 90% of the cristobalite samples from each of these departments were below the LOD, whereas the corresponding proportions for quartz exposure levels were 65 and 58%, respectively. Respirators were used by 74% of the workers some or all of the time (Føreland et al., 2008). The actual exposure levels were therefore lower than indicated by the measurements.

Determinants of exposure for all measurements

Plant as a determinant of dusts exposure explained only 0.33–4.4% of the BW variance. Department was a major predictor of fiber and SiC exposure, explaining 70 and 38% of the BW variance, respectively, whereas the BW variances for total dust and respirable dust were reduced only by less than 10%. The furnace department was associated with higher exposure to fibers and lower exposure to SiC. The processing department was associated with higher exposure to total dust, respirable dust, and SiC, and lower exposure to fibers. Job group was a major predictor of exposure for all agents explaining 43% (total dust), 44% (respirable dust), 82% (fiber), and 78% (SiC) of the BW variance. Combining plant and job group in the model gave the best fit for all agents explaining 45% (total dust), 48% (respirable dust), 82% (fiber), and 79% (SiC) of the BW variance. The explained WW variance was 0.5% or less for all the models.

Determinants of exposure at department level

Job group was the most important determinant of exposure for all agents in the furnace and processing departments. The final models for the furnace and processing departments are presented in Tables 4 and 5, respectively. The only department-specific determinant in the furnace and processing departments that was a significant predictor of exposure in the final models was process disturbances, i.e. unwanted events leading to a stop or delay in the production process.

The shift schedules could differ between plants and job groups. Workers within some job groups worked only daytime (e.g. all job groups in the maintenance department), while other worked two, three, or five shift schedules. The effect of shift was seen most clearly in the furnace department where working night shift was a determinant of lower exposure to most agents. The same effect was seen for total dust in the processing department. Job group was a less important predictor of exposure in the maintenance department compared to other departments and was only retained in the final models for total dust, respirable dust, and fiber (Table 6). The models explained 14–89% of the BW variance and 0–11% of the WW variance.

Determinants of exposure at job group level

Several tasks were identified as predictors of exposure. Assisting with assembling and filling of the furnaces had a geometric mean ratio (GMR) of 3.3–8.0, meaning a 3.3–8.0-fold increased exposure for the charger and mix operator. Manually sorting and having the sorting area inside the furnace hall was associated with a 1.5–6.7-fold increased exposure for the sorter operator. Cleaning lead to a 1.3–6.2-fold increase in exposure for operators in the furnace department, but a decreased exposure for other refinery workers. Filling of pallet boxes with SiC resulted in a 1.6–2.7-fold increased exposure for operators in the refinery and changing of pallet boxes resulted in 1.5–2.4-fold increased exposure for fines operators. Maintenance in the furnace hall resulted in a 3.9–4.8-fold increase in fiber exposure, and maintenance in the processing department resulted in a 1.7–2.1-fold increase in exposure to SiC. Work in control rooms, laboratories, fresh air ventilated crane cabins, offices and maintenance outside the furnace hall, and processing department were predictors of decreased dusts exposure with a GMR of 0.14–0.74. The final models for determinants on job group level are presented in Table 7 (furnace department), Table 8 (processing department), and Table 9 (maintenance department).
