Equivalency of a personal dust monitor to the current United States coal mine respirable dust sampler

Abstract

The United States National Institute for Occupational Safety and Health, through an informal partnership with industry, labor, and the United States Mine Safety and Health Administration, has developed and tested a new instrument known as the Personal Dust Monitor (PDM). The new dust monitor is an integral part of the cap lamp that coal miners normally carry to work and provides continuous information about the concentration of respirable coal mine dust within the breathing zone of that individual. Previous laboratory testing demonstrated that there is a 95% confidence that greater than 95% of individual PDM measurements fall within ± 25% of reference measurements. The work presented in this paper focuses on the relationship between the PDM and respirable dust concentrations currently measured by a coal mine dust personal sampler unit utilizing a 10 mm Dorr–Oliver nylon cyclone. The United Kingdom Mining Research Establishment instrument, used as the basis for coal mine respirable dust standards, had been designed specifically to match the United Kingdom British Medical Research Council (BMRC) criterion. The personal sampler is used with a 1.38 multiplier to convert readings to the BMRC criterion. A stratified random sampling design incorporating a proportionate allocation strategy was used to select a sample of mechanized mining units representative of all US underground coal mines. A sample of 180 mechanized mining units was chosen, representing approximately 20% of the mechanized mining units in production at the time the sample was selected. A total of 129 valid PDM/personal sampler dust sample sets were obtained. A weighted linear regression analysis of this data base shows that, in comparison with the personal sampler, the PDM requires a mass equivalency conversion multiplier of 1.05 [95% C.I. = (1.03, 1.08)] when the small intercept term is removed from the analysis. Removal of the intercept term results in a personal sampler-equivalent concentration increase of 2.9% at a PDM measurement of 2.0 mg m^–3.

---

1. Introduction

Measurement of workplace respirable dust concentration is an essential first step in eliminating lung disease caused by overexposure to dust. In the United States (US), The Federal Coal Mine Health and Safety Act of 1969, the predecessor of the Federal Mine Safety and Health Act of 1977,¹ mandates that respirable coal mine dust levels be monitored and controlled to a maximum of 2 mg m^–3 or below for a working shift, provided quartz levels remain at or below 5%. To date, this monitoring process has relied upon a coal mine dust personal sampler unit (hereafter referred to as personal sampler or CMDPSU in figures and tables) utilizing a pre-weighed 5 µm pore polyvinyl chloride (PVC) filter, preceded by a 10 mm Dorr–Oliver nylon cyclone, operated at a flow rate of 2.0 L min^–1 to collect a sample in the mine environment. The personal sampler historically includes an empirically derived² 1.38 multiplier to convert readings to the United Kingdom British Medical Research Council (BMRC) criterion.³

In consultation with labor, industry, and government, the US National Institute for Occupational Safety and Health (NIOSH) issued a contract to Rupprecht and Patashnick Co., Inc. (now Thermo Fisher Scientific Corporation), Albany, NY, USA, (CDC Contract 200-98-8004) to develop a Personal Dust Monitor (PDM). The PDM is based on existing environmental ambient air monitoring instruments, the latter of which has achieved global acceptance for use in air quality monitoring networks. The key feature of this monitor is that it directly measures the mass of dust on a filter, regardless of dust composition, size, or physical characteristics. This contract work successfully miniaturized the air quality monitor’s sensor and incorporated the sensor into a prototype, person-wearable dust monitor that provides accurate end-of-shift respirable dust concentration data to miners.⁴ Through a protocol developed by a partnership of labor, government and industry, previous work⁵ verified that the PDM laboratory accuracy met the NIOSH criterion.⁶ This report describes the underground full-shift performance of the PDM compared to the personal sampler.

2. Methods

2.1 Criteria for mine selection

A stratified random sampling design was used to select mechanized mining units that were representative of all US underground coal mines. The sampling base was developed using information extracted from the US Mine Safety and Health Administration (MSHA) Standardized Information System and reflected all producing mechanized mining units as of September 27, 2004. The selected sample of mechanized mining units was partitioned into mutually exclusive strata that reflected the type (MSHA bituminous coal district) and mining method (potentially related to size distribution) of the dust in underground coal mines. Because of the mine selection process, any mine dust size distribution effects are accounted for, on average, in the data analyses. Of the total mechanized mining unit population, only a small percentage employed longwall and other (e.g., scoop) mining methods (6% and 3.5%, respectively), with the balance being continuous mining sections. A proportionate allocation strategy with different sampling rates among the strata was used to ensure that the composition of the sampled mechanized mining units was approximately representative of the composition of the population. A sample of 180 was chosen to represent approximately 20% of all mechanized mining units in production at that time. The sample was randomly selected using the Survey Select procedure from the SAS system for the statistical analysis of data (SAS Institute, Inc., Cary, NC, USA).

2.2 Testing procedures

Planning and implementation of this segment of the testing was performed by NIOSH, in close cooperation with MSHA headquarters, district offices, and field offices. A sampling schedule was developed after the list of candidate mines and mechanized mining units was compiled. This schedule also introduced a degree of randomness in that, whenever possible, certain selected mines from different MSHA field offices and/or districts were combined into one trip. The mine sampling was conducted from October, 2004 through August, 2005.

Each sampling package contained (1) a PDM, pre-programmed to begin an 8 h sample at the selected mine’s shift start time and operating at 2.2 L min^–1; (2) two dust samplers using Dorr–Oliver cyclones, one operating at 2.0 L min^–1 and the other at 1.7 L min^–1; and (3) a Marple cascade impactor operating at 2.0 L min^–1. The Marple impactor is a miniaturized eight-stage cascade impactor intended to provide particle size distributions of sampled aerosol and sufficiently small as to be wearable as a personal sampler. The cyclones, impactor, and the PDM sampling inlet were mounted inside a specially constructed Lippmann sampling canister^7,8 with a single inlet. This procedure ensured that all samplers were exposed to the same atmospheric conditions. Flow-controlled Mine Safety Appliances Escort Elf® pumps were calibrated to 2.0 ± 0.02 L min^–1 and 1.7 ± 0.017 L min^–1 for the cyclone samplers prior to each trip using a Gilibrator (Sensidyne Inc., Clearwater, FL, USA) primary standard flow meter. An equivalent pressure restriction for the respective samplers was used during pump calibration. Flow rates for all samplers were checked prior to each trip and post-checked for proper calibration to ±1%. The post-check calibration also served as pre-calibration for the following trip. Sampling locations were chosen where miners typically worked. Data for the cascade impactor and 1.7 L min^–1 Dorr–Oliver sampler will not be included in this report because they are not pertinent to the present objective. Analysis of those samplers will be the focus of a future publication.

The PDM was pre-programmed to automatically begin sampling at the pre-determined shift start time and turn off eight hours later. Gravimetric samplers were manually started and stopped within 3 min of the PDM start/stop times. In the event that the PDM display indicated a projected end-of-shift concentration of 1.5 mg m^–3 or greater at any time during the sampling period, a HEPA filter was placed into the inlet of the Lippmann canister and all samplers finished running the 8 h period without collection of additional airborne dust into the canister. The purpose of the 1.5 mg m^–3 concentration limit was to minimize the possibility of overloading the impactor.

2.3 Samplers

2.3.1 PDM . The PDM^4,5 is a pre-commercial (not for general sale) model 3600 PDM using a tapered element oscillating microbalance (TEOM®) mass sensor. The PDM uses Teflon®-coated fiberglass (nominal 15 mm diameter) filter media TX40H120WW, (Pallflex Products Corp., Putnam, CT, USA) manufactured into special plastic holders that mount on the end of a vibrating hollow tapered element. The device is intended to be virtually “invisible” to the miner, as a replacement for the cap lamp and battery currently employed in mines. Fig. 1(a) illustrates the PDM device with accompanying charger/PC interface docking station. The docking station is used to simultaneously communicate with PC software for programming and retrieving stored data in the instrument and to recharge its batteries for the next work shift. Fig. 1(b) shows the PDM components. The enclosure of the PDM is hardened to withstand the harsh conditions found in the mine environment, with the system designed to meet MSHA drop test requirements for cap lamps,⁹ as well as intrinsic safety approval requirements.¹⁰ An illuminated data display on the top of the PDM continuously shows the dust concentration for the previous 30 min, cumulative mass concentration to that point in the shift, and a projected end-of-shift concentration. A full description of the PDM system configuration, air flow path, flow measurement and control, battery configuration, data acquisition, and mass measurement technique has been previously described^5,11 and will not be presented here. Additionally, extensive laboratory and replicate sampling data from ten underground coal mines to establish the accuracy and precision of the PDM have been previously reported.⁵

Fig. 1 (a) Pre-commercial PDM connected to docking station used for charging and communication with a PC using an RS-232 interface. (b) PDM internal components.

The PDM instrument samples, analyzes, and calculates mass-based concentrations of respirable dust. It is acknowledged that there may be some negative bias (defined as [measured concentration ÷ true concentration] –1) in the use of the PDM associated with particle losses that take place in the sampling path between the cyclone exit and the sensing zone of the instrument. This may be corrected empirically by reference to experimental data in specific situations. Prior testing⁵ of the pre-commercial PDM instrument found that the PDM consistently underestimated mass, based on comparison to reference mass samples using identical inlet configurations and Higgins–Dewell cyclones. Established scientific protocol requires measurement correction for systematic error (bias) prior to analysis. Therefore, these systematic measurement errors were used to calculate an average bias correction factor prior to regression analysis.^12,13

Additionally, prior work has documented that the PDM filter has a characteristic average positive thermally induced filter bias during an 8 h sampling period. This value was determined by random filter selection from various filter lots and is due to heating effects acting on the filter.¹⁴ The data of this work were also corrected for this bias, prior to regression analysis. While this filter-induced bias is trivial for an individual measurement, it is significant when many low mass measurements are combined in the regression analysis.

2.3.2 Cyclone gravimetric reference. Each sampler consisted of a Mine Safety Appliances 37 mm diameter, pre-weighed 5 µm pore PVC filter, preceded by a 10 mm Dorr-–Oliver nylon cyclone.

2.4 Analytical gravimetric imprecision

Gravimetric analysis was performed on a Mettler–Toledo UMT2 microbalance for the personal sampler samples. Weighing was done in the NIOSH Pittsburgh Research Laboratory at 22.8 + 0.4 °C (73 ± 0.7 °F) and 53 ± 2% RH. All samples were pre- and post-weighed, employing control filters. Two filter cassettes were used as controls for the personal sampler. Average blank control filter masses were used to correct the filter mass results for each test. Because sampling packages were deployed to numerous mines simultaneously, the control filters could not accompany the sampling packages to the mines. However, this is considered a minor factor compared to differences in weighing room conditions between pre- and post-weighings. All sample and control filters were desiccated and allowed to equilibrate to room conditions. However, small cyclical fluctuations in room conditions are not totally corrected by the control filters because: (1) the control filters cannot be weighed at the same moment as the sample filters, and (2) the pre- and post-weighings are not performed at the same time in relation to the cyclic variation of the room conditions. Since the pre- and post-weighings are not time-correlated with the cyclic variation, there is an uncorrectable, but minor, random variation. Therefore, several estimates for the weighing imprecision will be provided.

2.5 Analysis

Weighted regression is the method of choice to stabilize the variance for data analysis by estimating the relationship between the variance and the independent variable. Appendix A† describes the mathematical representation and weight variable estimation. It should be noted that, since the overall purpose is to predict contaminant concentration from an imprecise measurement in the PDM, there is no need to specifically consider random measurement error in the predictor variable. Additionally, it can be demonstrated that the random error in the PDM can be classified as one of the embodiments of the Berkson case. The Berkson case describes several classes of measurement procedure in which random error in the predictor variable has no effect on the regression analysis. Weighted regression was performed using Sigma Plot v.9.0.¹⁵

3. Results

3.1 PDM bias correction

A bias correction factor for internal dust loss within the PDM was estimated from previous empirical work.⁵ A negative bias of 6.6% for laboratory data from instrument testing was calculated. This bias correction is not applied to the tabular data of Tables B-1 and B-2 in Appendix B.† However, all subsequent data analysis will include this PDM bias correction as well as the thermal zero-drift bias correction of 25.5 µg (0.024 mg m^–3 for an 8 h sample).¹⁴

3.2 Gravimetric limit of quantification

Determination of the traditional limit of quantification (LOQ)⁶ requires determining the standard deviation (S_W) in nine consecutive weighings of a blank filter. It was determined that S_W = 1.4 µg and LOQ = 14 µg in a single weighing. However, quality control procedures involving control room filters used only in the balance room document that the total standard deviation (S_T) due to (S_W) and the cyclic fluctuations in weigh room conditions during the course of this study was S_T = 4.1 µg. Applying traditional formulae for propagation of error in the personal sampler filter dust mass gains (including an average of 2 control filters) yields S_filter = 1.8478×S_T = 7.6 µg. For an 8 h sample, this value corresponds to ₀σ = 0.011 mg m^–3, suggesting an LOQ = 0.11 mg m^–3 for the personal sampler.

3.3 In-mine testing results

Table B-1† lists all data included in the analysis, sorted in order of increasing PDM measurements. Of 180 total projected samples, 129 valid PDM/personal sampler pairs were collected. Although the number of valid samples obtained was less than the target value, the actual sample data base obtained was closely proportional to the original mine distribution by district. Table B-2† lists 9 entire data sets excluded for technical or procedural errors. It is again noted that the PDM values in Tables B-1 and B-2† do not include any bias corrections.

3.4 Data analysis

3.4.1 Weight variable estimation. Analysis of the raw mine replicate sampling data of Volkwein et al.⁵ show that the relative standard deviation (RSD) of the different samplers can be generally described by the equation

RSD = RSD₀ – A(mg^–1 m³) × X_sampler(mg m^–3) (R² < 0.1), (1)

with slopes (A) typically less than 0.05 in magnitude. The RSD is also called the coefficient of variation (CV). The low R² and the small slope of the data demonstrate general independence of the RSD on sampler concentrations above X_min ≈ 0.3 mg m^–3. Constant RSD necessarily implies variance increasing with the square of the mean. Given this relationship, a partial weighting factor to correct the regression for the increasing variance is (1/X²). It should be noted that the magnitude of the intercept constant (RSD₀) in eqn (1) is not important in establishing the weighting factor for two reasons. First, the only statistical requirement is proportionality between standard deviation and mean. Second, the intercept term incorrectly implies constant RSD as the concentration approaches zero.

The small negative slope indicates the presence of a constant variance term becoming more prominent at sampler concentrations less than 0.3 mg m^–3. Iteration of eqn (A-3) (see Appendix A),† with (₀σ)² fixed by weighing imprecision, converged quickly to a solution within 4 iterations, yielding.

σ_T² = (0.011 mg m^–3)² + 0.0155X². (2)

3.4.2 Weighted regression. Fig. 2 shows the scatter-plot relationship between the personal sampler and the PDM. The iterated weighted regression without the y-intercept term is included. In general, the data clearly indicate a linear relationship with multiplicative errors between the two instruments. Regression weight variable estimates were obtained using eqn (2).

Fig. 2 CMDPSU data plot with PDM bias corrections. CMDPSU data incorporate 1.38 multiplier historically used to estimate BMRC/MRE equivalency. Weighted regression statistics are presented in Table 1.

Table 1 shows the regression statistics for analyses with and without the intercept term. Although it is not quantifiable, the y-intercept = 0.018 mg m^–3 term is statistically significant at the 0.05 significance level (t_intercept = 4.35 > t_critical = 1.98). Removal from the regression can have a significant effect on the slope and the calculation of respirable dust concentration. For comparison, regression without the intercept yields slope = 1.05 [95% C.I. = (1.03, 1.08)]. Because t_slope = 3.82, the hypothesis [slope = 1.00] is rejected and the PDM would use a 1.05 conversion factor to be an equivalent measure of the personal sampler.

Table 1 CMDPSU (with 1.38 conversion factor) vs.PDM. Weighted regression results on bias-corrected PDM data

Statistic	Value	95% C.I.	90% C.I.
PDM designated predictor (independent) variable
Intercept y0/mg m^–3	0.018	(0.010, 0.026)	(0.011, 0.025)
S intercept	0.004
t intercept	4.35
Slope	1.01	(0.98, 1.04)	(0.98, 1.04)
S slope	0.017
t slope(wrt 1.0)	0.52
y 0 removed
Slope	1.05	(1.03, 1.08)	(1.03, 1.08)
S slope	0.014
t slope(wrt 1.0)	3.82
Regression 0σ/mg m^–3	0.011
Regression CV	12.4%
t critical(.05) =	1.98
t critical(.10) =	1.66
N =	129

---

3.4.3 Consideration of regression intercept terms. Ideally, in the absence of bias one would expect both instruments to read zero or to average zero for a very large data set when no dust is present. When this does not occur, there are several different schools of thought regarding treatment of an intercept term. One philosophy is that, in many cases, the intercept is not considered meaningful for two reasons. First, the intercept is usually beyond the range of the data and assigning any significance to an extrapolation beyond the data is suspect. Second, the intercept can be highly influenced by error inherent in the data. Therefore, a case could be made for simply ignoring an unmeasurable intercept.

Another school of thought is that in a regression, the intercept and slope are correlated. If, in particular, the intercept term is not statistically significant, it offers no useful information and should be removed from a subsequent regression. The issue is not quite so clear-cut when the intercept is statistically significant but unmeasurable. In this case, performing the regression without the intercept may have a significant effect on the slope parameter, depending on the level of significance associated with the intercept. Because consideration of the intercept term is often an end-user decision, the analysis is presented with and without the intercept term in the regression.

Additional insight regarding the intercept term can be obtained using independent data obtained by Volkwein et al.⁵ This independent data set was obtained in the same manner as the present data, with the exception that, although data was collected in every mining district, the mines were not selected in a proportional manner. Even though this proportionality is a very important consideration, the independent data is useful. The weighted linear regression result was a statistically significant intercept of –0.020 and slope of 1.05. Given that the present work has an intercept of +0.018, it could reasonably be inferred that the intercept terms are merely a result of error inherent in the data and should be removed from the regression. Removal of the intercept term in the independent data results in a slope of 1.02. Although this is slightly outside the present confidence interval (1.03, 1.08), the difference is somewhat exaggerated by rounding error (1.024 vs. 1.026). In any case, being somewhat conservative would be preferable to basing the health of miners on a statistical consideration. This would provide justification for not including the intercept term in the regression.

4. Conclusions

An accurate, direct reading dust monitor for use in coal mines has been demonstrated in underground coal mine environments. In the current, as-tested, PDM configuration and calibration, weighted regression analysis indicates that a linear relationship between the PDM and the personal sampler can be established. The conversion factors are applicable only after applying the PDM internal dust loss and filter thermal bias corrections to the raw PDM measurements for an 8 h sampling period. However, it is noted that the filter thermal bias correction is negligible at most respirable dust concentrations of interest, being only 2.4% at a concentration of 1.0 mg m^–3. It is recommended that the manufacturer document and implement the measured bias corrections to the mass concentration calculations of the PDM instrument.

The results indicate the suitability of the PDM for in-mine use to assess respirable dust concentrations defined in accordance with the currently used personal sampler. This fact, together with the advantages of a direct reading end-of-shift instrument, provides strong justification for adoption of the PDM as a primary in-mine sampling device.

In comparison to the personal sampler, the PDM would require a mass concentration equivalency conversion multiplier of 1.05 [95% C.I. = (1.03, 1.08)] when the small intercept term is removed from the analysis. Removal of the intercept term results in a personal sampler-equivalent concentration increase of 2.9% at a PDM measurement of 2.0 mg m^–3.

5. Disclaimer

The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the National Institute for Occupational Safety and Health. Mention of any company or product does not constitute endorsement by the National Institute for Occupational Safety and Health.

Acknowledgements

The authors thank and acknowledge the support and technical advice given by members of the PDM partnership, namely, Dennis O’Dell of The United Mine Workers of America, Joe Lamonica of the Bituminous Coal Operators and Bruce Watzman of the National Mining Association. Additionally, we would like to thank David Bartley, visiting scientist, NIOSH, for his technical expertise and support during the course of data analysis. We also thank MSHA for its active support and participation in the collaborative research effort by purchasing units for testing and collecting vital test data at 144 mechanized mining units around the country. We appreciate the generous engineering support provided by Erich Rupprecht, Business Development Manager and Dan Dunham, PDM project manager, from Thermo Fisher Scientific Corp. Finally, we thank the Pittsburgh Research Laboratory staff, Cal Garbowsky, Shawn Vanderslice, Jeanne Zimmer, Erica Hall, Randy Reed and Tom Mal for conscientious support in managing the logistics of the equipment for the mine tests.

References

1. http://www.msha.gov/REGS/ACT/ACT2.HTM#7 .
2. T. F. Tomb, H. N. Treaftis, R. L. Mundell and P. S. Parobeck, Comparison of Respirable Dust Concentrations Measured with MRE and Modified Personal Gravimetric Sampling Equipment, U. S. Department of Interior, Bureau of Mines, RI 7772, 1973, 28pp.
3. A. J. Orenstein, Proceedings of the Pneumoconiosis Conf., Johannesburg, 1960.
4. J. C. Volkwein, R. P. Vinson, L. J. McWilliams, D. P. Tuchman and S. E. Mischler, Performance of a New Personal Respirable Dust Monitor for Mine Use, US Department of Health and Human Services, Public Health Services, Centers for Disease Control, National Institute for Occupational Safety and Health, Report RI 9663, 2004, 20pp.
5. J. C. Volkwein, R. P. Vinson, S. J. Page, G. J. Joy, S. E. Mischler, D. P. Tuchman and L. J. McWilliams, Laboratory and Field Performance of a Personal Dust Monitor, US Department of Health and Human Services, Public Health Services, Centers for Disease Control, National Institute for Occupational Safety and Health, Report RI 9669, 2006, 47pp.
6. E. R. Kennedy, T. J. Fischbach, R. Song, P. M. Eller and S. A. Shulman, Guidelines For Air Sampling And Analytical Method Development And Evaluation, US Department of Health and Human Services, Public Health Services, Centers for Disease Control, National Institute for Occupational Safety and Health, Report 1995, pp. 95–117.
7. M. W. Blachman and N. Lippmann, Am. Ind. Hyg. Assoc. J., 1974, 35, 311–326.
8. J. C. Volkwein and E. D. Thimons, Proceedings of VIIth International Ventilation Symposium, Kracow, 2001.
9. US Code of Federal Regulations 30 CFR Part 19.
10. US Code of Federal Regulations 30 CFR Part 18.
11. Rupprecht & Patashnick Co., Inc., Operating Manual TEOM Series 3600/3700 Personal Dust Monitor, Revision A, R&P Part Number 42-009904, September 2004.
12. National Institute of Standards and Technology, http://www.itl.nist.gov/div898/handbook/mpc/section6/mpc632.htm, 2006.
13. Intergovernmental Panel on Climate Change, http://www.grida.no/climate/ipcc_tar/wg1/192.htm, 2006.
14. S. J. Page, D. P. Tuchman and R. P. Vinson, J. Environ. Monit., 2007, 9, 760–767.
15. Systat Software, Inc. SigmaPlot v.9.0, Richmond, CA, USA.

---

Footnote

† Electronic supplementary information (ESI) available: Appendices A and B, including Tables B-1 and B-2. See DOI: 10.1039/b714381h

---