Document aJr9p8ZqXBLBRrzQ9J5NQJNv9

DownloadSearchable PDFRandom document
SAFETY IN MINES RESEARCH ADVISORY COMMITTEE Final Report Engineering Gravimetric Monitoring Methods to Cope With Excessively High Dust Levels - 2 B K Belle and J J L du Plessis Research agency Project number Date CSIR Miningtek COL 515 November 1999 0 Executive Summary In the first-phase of the project, Oberholzer et al., (1998) developed the rationale for adapting an addition of a cowl (shield) to the existing South African cyclone for measuring dust at high concentrations. In this report, background to the reasons for sampling, rationale for adapting the sampling head, the evidence of various problems with discrepancies between instruments, the same instruments of different manufacturers, and the difficulties in evaluating these samplers are discussed. This report, a second-phase of COL 515 project titled "Engineering Gravimetric Sampling Methods for Monitoring the Conditions of Excessively High Dust Levels-2." The objective of the second phase of the work was to assess the results of the field trials on modified cyclones for dust measurements. Field experience of dust measurement in South African coal mines has suggested that in very high dust concentrations, cyclones for sampling respirable dust can become overloaded (Oberholzer, 1998). In the light of studies conducted by Kenny, Baldwin and Maynand (1998) in UK, the use of a cowl (shield) attachment was favoured. The study reported to have shown that there was no significant difference in the concentration level measured between the cyclone with and without the shield. The general view of the several international opinions is that the use of the cowl will help to increase the accuracy of dust concentration measurements without increasing the bias of samples. It is further recommended that the CEN/ISO respirable curve with a d50 of 4 pm should be introduced as the standard. In overall, following conclusions were made from the current study: The work carried out in the field demonstrates equivalent dust concentration results from the two sampler types in different positions within the mine. The extent of the agreement between the two sets of results seems to be excellent and much better than that for results obtained in the laboratory results or in non-mining workplaces. From the statistical analysis (f-test and F-test), the mean dust concentration measured by pairs of cyclones is not significantly affected at 95 percent level of confidence. In other words, there is no difference between the measured mean dust concentration levels between various sampler pairs. 1 From the general size distribution plots from the sample dust, we notice that the mineface respirable dust has two relative maximums, called modes and can be referred to as a bi-modal distribution. From the size analysis on dust samples (with and without cowl samples) of the measured dust concentrations in the range 2,37 mg/m3 to 10,59 mg/m3, it is seen that the sampled respirable dust contained particles greater than 10 pm, except for the BGI sampler dust (3,89 mg/m3 concentration). At very high measured dust concentrations, the cowl sampler lowered the loading of large particles. From the size distribution analysis data, we can conclude that the cowl samplers are the preferred samplers for taking engineering samples at high dust concentrations. However, the results showed that at very high dust concentrations, there is a huge variation in both the measured dust concentration and the size distribution for a paired sample. For a pair of cowl and without-cowl samplers, for the same dust concentration levels, the evidence showed the presence of wide size distributions. It is thus evident that, although the dust mass collection of the samplers conforms to the requirements, the size distribution of the dust is not concomitantly accurate. The field study failed to conclusively determine whether the modifications made to the South African cyclone either improve or degrade its performance for the conditions tested, based on the results of the size analyses. From the field observations, it can be stated that the cowl sampler can be used for engineering sampling to avoid the dependence of the cyclone inlet on orientation in the collection of particles. The data analysis and field experience have shown that in comparison with other samplers, BGI samplers greatly reduce the amount of oversized particles deposited onto the filter from the inversion of cyclones. Also, due to the design of its cyclone inlet, the BGI sampler is not affected by orientation dependence. The field observations did indicate that when collecting samples closer to the face, where high concentrations of dust are present, sample handling is extremely difficult. At high dust concentrations, the probability of non-respirable particles depositing onto the filter from the grit pot is very high when the pump is switched off. However, the BGI sampler, due to its design, alleviates this problem to some extent. In both the laboratory and field tests, a sufficient mass of sample dust was required for analysis with the Fritsch analyser to determine the presence of non-respirable particles. This led to the conclusion that there may have been non-respirable particles in the concentrations below 5 mg/m3 (as found from this study). 2 The size analysis of the collected dust mass for sample concentrations greater than 5 mg/m3 did not follow the D50 of either BMRC or ISO/ACGIH/CEN respirable curves. The cowl sampler could not be recommended for personal sampling because the results of the size analysis are inconclusive and wide-ranging. Finally, question of cowl under-sampling (under-estimation of dust concentration) or over-sampling (over-estimation of dust concentration) could not be resolved with confidence from size distribution data at concentrations below 5 mg/m3 despite the insignificant differences obtained from comparison of mass concentrations alone. From the conclusions drawn, the following recommendations for future research are proposed, although it is suggested that there should be a paradigm shift in the existing sampling procedure and instruments in the future project: The results were inconclusive with regard to the recommendations on the usage of adding a cowl to the cyclone in the personal sampling method. However, the use of a cowl for taking engineering samples can be recommended on the grounds that it lowers sample-handling errors, the probable reason being that in practice it reduces factors such as the wind and orientation-dependence of the cyclone inlet. At high dust concentrations, it was difficult to determine which sampler measured the "true concentration." Therefore, future field tests should incorporate the aspect of "impactor sampling" from which both the mass concentration and the size distribution of ambient particles can be accurately determined. Despite some of the cyclones being accepted for sampling in the industry, concerns were expressed about the cyclone performance with respect to the BMRC and ISO/ACGIH/CEN respirable curves. Fears were also expressed about whether the cyclones do follow the BMRC respirable curve under concentrations below 5 mg/m3. In addition to the investigation, further work on recommendations for the use of the cowl sampler for personal sampling could be considered. Similar studies should be carried out to compare these samplers with CIP10 samplers and to investigate their variances and size distribution. The different makes of samplers should also be investigated to determine the relationship between them, not only with regard to the mass of dust collected but also in terms of the size characteristics of the dust. Tests should be carried out to determine the size characteristics of personal dust samples and engineering dust samples to compare with the existing DME requirements for engineering sampling. 3 Immediate attention needs to be given to analysing both the compliance and non compliance samples from the mines, which have difficulty in maintaining the 1997 DME directive. On the basis of this outcome, there needs to be clarification on the matter of performance evaluation of all the cyclones currently used in the mines. An extensive facility, including a chamber and a particle size analyser for analysing sample mass on compliance samples of less than 5 mg/m3, needs to be established for this serious issue of sampling. A programme needs to be established for dust sampling, measurement and data analysis for exposure assessment. 4 Table of Contents Page Executive Summary .......2 List of Figures ... ....8 List of Tables ... 11 Glossary of Abbreviations and Symbols . . . . . 12 Acknowledgements . . . . . . . . 13 1 Introduction 14 2 Background.........................................................14 3 Test Samplers.........................................................16 3.1 BGI Stainless Steel Cyclone. .... 16 3.2 GME-G05 Cyclone........................................................16 3.3 GME-G05 Cyclone with Cowl. . . . . 17 4 Sampler Positions.................................................18 5 Data Analysis.........................................................19 6 Results and Analysis 6.1 Statistical Analysis . 6.1.1 F-Test analysis . . 6.1.2 f-Test analysis . . 6.2 Particle Size Distribution 20 31 31 33 37 5 7 Discussions 42 8 Conclusions.........................................................45 9 Recommendations................................................ 46 10 References.........................................................48 Appendix A......................................................... 51 6 List of Figures Figure 3 Figure 4 Figure 6a Figure 6b Figure 6c Figure 6d Figure 6e Figure 6f Figure A1 Figure A2 Figure A3 Figure A4 Figure A5 Figure A6 Figure A7 Figure A8 Pictorial view of the BGI stainless steel sampler (left), the cowl sampler (centre) and the without-cowl sampler (right). 17 Layout of the test section . . . . . 19 Plot of combined dust concentrations obtained from two BGI samplers ...... 22 Plot of combined dust concentrations obtained from two cowl samplers ...... 23 Plot of combined dust concentrations obtained from two without-cowl samplers ..... 24 Plot of combined dust concentrations obtained from BGI and cowl samplers....... 25 Plot of combined dust concentrations obtained from BGI and without-cowl samplers ..... 26 Plot of combined dust concentrations obtained from cowl and without-cowl samplers ..... 27 Plot of the dust concentrations obtained from two BGI samplers at the section intake .... 54 Plot of the dust concentrations obtained from two cowl samplers at the section intake .... 55 Plot of the dust concentrations obtained from two without-cowl samplers at the section intake ... 56 Plot of the dust concentrations obtained from BGI and cowl samplers at the section intake. . . 57 Plot of the dust concentrations obtained from BGI and without-cowl samplers at the section intake . 58 Plot of the dust concentrations obtained from two cowl and without-cowl samplers at the section intake . 59 Plot of the dust concentrations obtained from two BGI samplers at the operator. .... 60 Plot of the dust concentrations obtained from two cowl samplers at the operator .... 61 7 Figure A9 Figure A10 Figure A11 Figure A12 Figure A13 Figure A14 Figure A15 Figure A16 Figure A17 Figure A18 Figure A19 Figure A20 Figure A21 Figure A22 Figure A23 Figure A24 Figure A25 Plot of the dust concentrations obtained from two without-cowl samplers at the operator ... Plot of the dust concentrations obtained from BGI and cowl samplers at the operator ... Plot of the dust concentrations obtained from BGI and without-cowl samplers at the operator . . Plot of the dust concentrations obtained from two cowl and without-cowl samplers at the operator .. Plot of the dust concentrations obtained from two BGI samplers at the section return .... Plot of the dust concentrations obtained from two cowl samplers at the section return .... Plot of the dust concentrations obtained from two without-cowl samplers at the section return ... Plot of the dust concentrations obtained from BGI and cowl samplers at the section return... Plot of the dust concentrations obtained from BGI and without-cowl samplers at the section return . Plot of the dust concentrations obtained from two cowl and without-cowl samplers at the section return . Particle size distribution of sampled dust from without-cowl cyclone .. ... Particle size distribution of sampled dust from BGI cyclone ...... Particle size distribution of sampled dust from without-cowl cyclone pair . . ... Particle size distribution of sampled dust from without-cowl cyclone pair ..... Particle size distribution of sampled dust from without-cowl cyclone pair ..... Particle size distribution of sampled dust from cowl and without-cowl cyclone pair ..... Particle size distribution of sampled dust from without-cowl cyclone pair ..... 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 Figure A26 Particle size distribution of sampled dust from BGI and 8 Figure A27 Figure A28 Figure A29 without-cowl cyclone pair ..... Particle size distribution of sampled dust from cowl and without-cowl cyclone pair ..... Particle size distribution of sampled dust from cowl and without-cowl cyclone pair ..... Particle size distribution of sampled dust from BGI and without-cowl cyclone pair ..... 79 80 81 82 9 List of Tables Table 6a Summary statistics of the concentration levels from the various sampler pairs ..... 21 Table 6b Summary statistics of the concentration levels from the various sampler pairs ..... 21 Table 6c Results of the sampler biases at the intake, operator and return ...... 28 Table 6d Overall results of the sampler biases. ... 30 Table 6.1.1a Results of the F-test hypothesis (on transformed values) . 33 Table 6.1.1b Results of the F-test hypothesis (on transformed values) . 33 Table 6.1.2a Results of the paired t-test hypothesis (on transformed values) 35 Table 6.1.2b Results of the paired t-test hypothesis (on transformed values). 36 Table 6.2a Measured respirable dust fraction and mass concentrations using the cowl and without-cowl samplers (Kenny et a/., 1998) ...... 37 Table 6.2b Measured particle size distributions (by volume) using the cowl and without-cowl samplers (Kenny et a/., 1998) ...... 38 Table 6.2c Measured particle size distributions from the BGI, cowl and without-cowl samplers (using a Fritsch size analyser) .. .40 Table A1 Linear regression details between pairs of samplers at the intake ....... 52 Table A2 Linear regression details between pairs of samplers at the operator ...... 52 Table A3 Linear regression details between pairs of samplers at the return ....... 53 Table A4 Combined linear regression details between pairs of samplers ...... 53 10 Glossary of Abbreviations and Symbols Abbreviations ACGIH American Congress of Governmental Industrial Hygienists BMRC British Medical Research Council CEN European Standards Organization CM Continuous Miner DME Department of Minerals and Energy, South Africa HSL Health and Safety Laboratories, UK ISO International Standards Organization MRE Medical Research Establishment, UK NIOSH National Institute for Occupational Safety and Health, USA SC Sample Concentration SIMRAC Safety in Mines Research Advisory Committee TWA-CONC Time-Weighted Average Dust Concentration USBM United States Bureau of Mines Symbols mg/m3 m3/s L/min pm mm m milligrams per cubic metre cubic metres per second litres per minute micron millimetre metre 11 Acknowledgments The authors wish to express their sincere gratitude and appreciation to the following people and organisations whose help and support made the completion of this project possible: Sarel W Pretorius, an independent contractor, for his assistance with the field tests. T Bleeker, Manager, D du Plooy, Production Superintendent, A Cronje, Environmental Supervisor, and the section workers of Matla Colliery, No. 3 Mine, Section 25, for their assistance with the test section facilities. Colleagues at CSIR Miningtek: Don Bryden, Lynn Milns and all other individuals for their contribution towards the underground phase of the project. Dr Fred Kissell (NIOSH), John Volkwein (NIOSH), Dr Paul Baron (NIOSH), Dr Thomas F Tomb (MSHA) and Dr Lee Kenny (HSL) for their input during various informal discussions on the area of dust measurement. 12 1 Introduction In the first phase of the project, Oberholzer et al., (1998) developed the rationale for adding a cowl (shield) to the existing South African cyclone for measuring at high dust concentrations. In the report on the first phase, they discussed the background to the reasons for sampling, the rationale for adapting the sampling head, the evidence of various problems leading to discrepancies between instruments, the same instruments made by different manufacturers and the difficulties in evaluating these samplers. The present report deals with the second phase of project COL 515, entitled "Engineering gravimetric monitoring methods to cope with excessively high dust levels2." The objective of the second phase of the work was to assess the results of the field trials on modified samplers (cyclones) for dust measurements. 2 Background Studies in the USA and Europe resulted in a new particle size selection criterion, namely the Soderholm respirable dust curve in which D50 = 4 pm, for the measurement of dusts in the workplace (ACGIH, 1991; 1992). This criterion was agreed upon by both the International Standards Organization (ISO) and the Comite European de Normalisation (CEN) (ISO, 1991; CEN 1991). Several laboratory evaluation studies were carried out on the performance of the various cyclones by determining the size characteristics of the sample dust collected by the cyclone. Determining the size distribution of the actual face dust underground is, however, a cumbersome and time-consuming process. Nevertheless, the size distribution of the sampled dust is one of the major parameters in determining sampler performance efficiencies. Most of the cyclone tests are carried out in the laboratory under steady state air conditions, quite unlike the turbulent air conditions underground. Establishing the presence of non-respirable particles carried over onto the filter requires the ability to count and analyse the particles in the collected sample. Due to the dynamic nature of mining and the harsh environmental conditions underground, the behaviour of particles in this situation is not very clearly understood. It is not yet known whether the size distribution of the particles generated will be uniform throughout a particular mining operation. From the sampling point of view, it is a well-observed phenomenon that some of the respirable dust 13 particles deposit on the inner walls of cyclones. This observed behaviour is unavoidable and persisted during most of the sampling done underground. It can therefore be speculated that the sample concentration results obtained are an underestimation of the sample mass concentration to which workers are exposed underground. Growing concern about dust-related diseases and the concomitant emphasis on dust exposure levels make the precise and accurate assessment of exposure to harmful coal dust of the utmost importance. Currently, South African coal mines are assessing workers' dust exposure using various gravimetric dust samplers (cyclones), viz. Casella 10 mm cyclones, Gilian cyclones and MSA cyclones. These samplers are designed for monitoring coal dust and are approved by the South African Department of Minerals and Energy (DME) for use at an airflow rate of 1,9 L/min, in agreement with the BMRC respirable convention (BMRC, 1952). The CIP 10 sampler is also approved at an airflow rate of 10 L/min. The mines are also obliged to submit "engineering samples" to the DME, consisting of samples collected at the operator's position on the continuous miner (CM) (irrespective of whether the machine is remotely operated or not). In addition, the DME separately collects two monthly "personal samples" from each section of each South African coal mine for assessing personal dust exposure levels. Field experience of dust measurement in South African coal mines has suggested that in very high dust concentrations, cyclones for sampling respirable dust can become overloaded (Oberholzer, 1997, 1998). This study further indicated that there was a need to test an additional fitting (cowl) to the cyclones most commonly used in South African collieries, aimed at reducing the potential for large particle carryover. These tests were undertaken at the Health and Safety Laboratories (HSL) in the UK. In the light of the findings by Kenny et al. (1998) at the HSL, the use of a cowl (shield) attachment was favoured. The HSL study reported that there was no significant difference in the dust concentration levels measured between the cyclone with and without the cowl. Several international expert opinions about the use of a cowl head were canvassed. The general view was that the cowl would help to increase the accuracy of dust concentration measurements, without increasing the bias of samples. It was further recommended that the CEN/ISO respirable dust curve in which D50 = 4 pm should be introduced as the standard. 14 In South Africa, fears were expressed that the additional cowl (shield) surrounding the cyclone would inevitably result in lower respirable dust concentration measurements. Studies had previously been carried out in the USA by Cecala et al. (1983) to evaluate the use of cowls on personal samplers to help reduce bias when sampling in high-velocity air streams and to reduce bias caused by sampler inlet orientation. The study indicated that the shield placed around the cyclone diminished the effects of wind velocity and sampler orientation on sampler performance. However, the study was not structured to produce definitive results, but rather to discern likely trends. 3 Test Samplers (Cyclones) In this study the cyclones were operated at the proposed airflow rate of 2,2 L/min in accordance with the new ISO/CEN/ACGIH respirable dust curve. It was hoped that the new flow rate would confer an immediate advantage in sensitivity since existing South African cyclones sample 16 percent less air per minute. The cyclones used in this study are described below. 3.1 BGI Stainless Steel Cyclone The BGI stainless steel cyclone used in the tests is of all-metal construction (Figure 3). The cyclone body is fabricated from stainless steel. The dust cup (grit pot) is fabricated from aluminum and is threaded to the cyclone body with an "O" ring seal. Filtration is accomplished with a 37-mm disposable cassette, which is pressed on over an "O" ring seal. This instrument has a recommended flow rate of 2,2 L/min to match the new 4 pm, 50 percent cut-off, respirable curve (ACGIH, 1991) and is not sensitive to charge effects (Bartley et al., 1994). The cyclone is also available with a plastic body and has been the subject of a careful study by the HSL (UK) (Maynard, and Kenny, 1995). 3.2 GME-G05 Cyclone This locally manufactured and DME-approved cyclone (GME-G05, 10 mm cyclones), which was used in the tests and is hereafter referred to as "cyclone without cowl or Cwcowl", is shown in Figure 3. The cyclone body is fabricated from plastic. The dust cup (grit pot) is also fabricated from plastic and is fitted to the cyclone body. Filtration is accomplished with a 37-mm disposable cassette, which is pressed on over an "O" ring seal. Measurement of the size-selection characteristics of this cyclone, as tested in the UK, confirmed that they are similar to those of the Higgins-Dewell designs commonly used in the UK and elsewhere in Europe. 15 3.3 GME-05 Cyclone with Cowl This locally manufactured and DME-approved cyclone (GME-G05, 10 mm cyclones) was fitted with an additional round steel shield, which is attached to surround the cyclone. It is hereafter referred to as "cyclone with cowl or Ccowl" (Figure 3) and was also used in the tests. The cowl surrounds the entire unit and is the same length as the cyclone. It operates as a vertical elutriator, preventing larger particles from entering the cyclone as a result of sedimentation within the cowl. It has the additional benefit of protecting the cyclone inlet slit from strong external winds, which are known to affect the sampling characteristics of an open-slit entry. Figure 3: Pictorial view of the BGi stainless steel sampler (left), the cowl sampler (centre) and the without-cowl sampler (right) Each sampling train consists of an air pump which draws air through the cyclone, and the cyclone itself which in turn selectively collects the fraction of airborne respirable dust less than 7 pm particles on a pre-weighed filter disc. Filters from the samplers were weighed 16 on an analytical electronic balance calibrated in readable units of 0,0001 mg. The procedure for determining the particulate mass as given in the DME guidelines (DME, 1997) was followed. Well-maintained pumps were used to avoid the effect of pump pulsations and fluctuations in the flow rate, as reported by Berry (1990). 4 Sampler Positions All the field measurements were carried out in a bord-and-pillar CM production section. Figure 4 shows the typical deployment of the dust-monitoring instruments in the test section. The samplers were positioned in the section intake, in the operator's cabin and in the section return airway, with the following combinations of cyclone pairs: BGI sampler (CBgi) and BGI sampler (CBgi) Cowl sampler (CCOWL) and cowl sampler (CCOWL) Without-cowl sampler (CWCOWL) and without-cowl sampler (CWCOWL) BGI sampler (CBgi) and cowl sampler (CCOWL) BGI sampler (CBgi) and without-cowl sampler (CWCOWL) Cowl sampler (CCOWL) and without-cowl sampler (CWCOWL). Samplers were kept together (approximately 12 inches apart) at the same position to avoid the possibility of greater spatial variations in the aerosol concentrations. 17 Figure 4: Layout of the test section 5 Data analysis Typically, the inaccuracy maps and penetration curves for the cyclones used in the samplers are available from the instrument manufacturer or at laboratories for performance evaluation. However, for the calculation of inaccuracies, one needs to have information on the aerosol size distribution and this is not generally known at the time of sampling. Also, most cyclone tests are carried out in the laboratory under steady state air conditions, unlike the turbulent air conditions that prevail underground during tests. Carrying out successful sampler efficiency tests requires the ability to count and analyse both the collected sample and the ambient aerosol. Collecting a representative ambient air sample underground is difficult in itself, and determining the size distribution of the ambient air as well would be expecting too much. Therefore, in phase two of this project, a comparative evaluation of the various samplers was carried out. The comparison was based on the mass concentrations of the respirable dust sampled and on the results of the size distribution analyses of selected samples which contained enough dust mass to enable size analysis to be carried out. The dust concentrations presented throughout this report reflect gravimetric dust measurements taken over a specific sampling period with the sampler pair. The gravimetric concentration was calculated using the mass of the dust collected over the 18 sampling period and the flow rate. Using the mass of dust collected on the filters, the sample dust concentration (SC) in mg/m3 is obtained as follows: where: sc = fi (Fl x T) (1) Ci = corrected initial filter mass in mg (correction factor is for moisture content change) Cf = corrected final filter mass in mg (correction factor is for moisture content change) Fl = sample flow rate in m3/min T = sampling time in min. Since the "true value" of the concentration measured at the predetermined position is not known, measurement bias for a pair of samplers was defined as follows: MBab = (CA -C>x 100 CA where: MBaB =measured mean concentration bias between the sampler pair type-AB CA =measured mean concentration by a sampler A (first sample in the pair) CB =measured mean concentration by a sampler B (second sample in the pair) The sampler pair data at various positions were combined for analysis procedure, due to the small number of sampler pairs collected at the intake, operator and return. However, individual sampler pair data at various locations were plotted and shown in Appendix A. 6 Results and Analysis The relationship between the concentration values obtained at the section intake, the CM operator's position and the section return from the various pairs of samplers used in the underground tests is shown in Appendix A (Figures A1 to A18). The measured concentration levels in the section (intake, operator and return) were combined in order to make a statistical comparison of six pairs of samplers and are shown in Figures 6a to 6f. The correlation coefficient (r) and the linear regression line between the various sampler pairs at the section intake, operator and return are summarised in Tables A1 to A4 in Appendix A. Summary statistics of the dust concentration values obtained with the various samplers at the different sampling positions are shown in Tables 6a and 6b. 19 In Table 6a, the comparison of similar sampler types is shown. In Table 6b, the dust concentration data for various combinations of samplers are shown. Table 6a: Summary statistics of the concentration levels from the various sampler pairs Statistics Mean Variance Median Minimum Maximum Count Sampler Pair CBGI 2,69 5,32 CBGI 2,54 3,94 2,10 1,97 0,79 0,84 10,48 8,35 23 23 Sampler Pair CCOWL 2,63 4,76 CCOWL 2,74 5,86 1,80 1,87 0,45 0,43 7,96 10,06 40 40 Sampler Pair CWCOWL 3,27 4,34 CWCOWL 3,12 4,14 2,88 2,55 1,04 0,93 9,63 8,52 22 22 Table 6b: Summary statistics of the concentration levels from the various sampler pairs Statistics Mean Variance Median Minimum Maximum Count Sampler Pair CBGI 1,86 CCOWL 1,73 4,57 4,06 1,15 1,17 0,10 0,11 13,62 15,87 123 123 Sampler Pair CBGI 1,25 CWCOWL 1,45 0,75 1,00 1,10 1,62 0,27 0,25 2,81 2,95 15 15 Sampler Pair CCOWL 2,80 CWCOWL 2,85 9,95 10,29 1,07 1,43 0,37 0,39 11,38 11,79 26 26 20 12 10 a> o_ E 8 c<o/) OCQ n aa) 6 3<(a0/>) caocoo) 4 at3o 2 0 y = 0.8483x + 0.2475 R2 = 0.9717 4 6 8 10 Dust concentration measured by BGI sampler in mg/m3 12 Figure 6a: Plot of combined dust concentrations obtained from two BGI samplers 21 Dust concentration measured by Cowl sam pler in m g/m3 0123456789 Dust concentration measured by Cowl sampler in mg/m3 6b,: Plot of combined dust concentrations obtained from two cowl samplers 22 Figure 6c: Plot of combined dust concentrations obtained from two without-cowl samplers 23 Figure 6d: Plot of combined dust concentrations obtained from BGI and cowl samplers 24 4 "e E c 0) 3 Q_ Ec</o> o1o 3O n2 ~0i3<(0_a0/))) E caocoo> i1 to Q3 0 0 y = 1.103x + 0.068 R2 = 0.9073 23 Dust concentration measured by BGI sampler in mg/m3 4 Figure 6e: Plot of combined dust concentrations obtained from BGI and withoutcowl samplers Figure 6f: Plot of combined dust concentrations obtained from cowl and without cowl samplers 26 The "bias" between samplers is calculated according to the formula discussed in Section 5. Sampler biases during the measurement by various sampler types at the section intake, operator and section return are shown in Table 6c. Table 6c: Results of the sampler biases at the intake, operator and return Intake (% bias) Operator (% bias) Return (% bias) CBGI CCOWL CWCOWL CBGI CCOWL CWCOWL CBGI CCOWL CWCOWL CBGI 5,8 -1,96 1,37 8,6 10,25 -12,21 2,3 1,1 -28,63 CCOWL -3,97 2,71 -8,33 2,21 8,08 -19,31 cwcowl 4,34 5,62 3,44 From a comparison of the results of the mean dust concentrations measured at the intake with the various sampler pairs, the following conclusions were drawn (Appendix A Figures A1 to A6, and Table 6c): On average, with the same sampler types, there was a bias of 4,7 percent in the concentration measurement (i.e., 5,78 percent for CBGI-CBGI; 3,97 percent for CCOWLCcowl; and 4,34 percent for Cwcowl-Cwcowl). A comparison of the mean concentration levels from CBGI and CCOWL data shows that the cowl sampler oversamples by approximately 1,96 percent when compared with the BGI sampler. A comparison of the mean concentration levels from CBGI and CWCOWL data shows that, on average, the sampler without cowl undersamples by approximately 1,37 percent when compared with the BGI sampler. However, a comparison of the mean concentration levels from CCOWL and CWCOWL data shows that, on average, the without-cowl sampler undersamples by 2,71 percent when compared with the sampler with cowl. From a comparison of the results of the mean dust concentrations measured at the operator with the various sampler pairs, the following conclusions were drawn (Appendix A - Figures A7 to A12, and Table 6c): On average, with the same sampler types, there was a bias of 7,5 percent in the concentration measurement (i.e. 8,6 percent for CBGI-CBGI; 8,33 percent for CCOWLCcowl; and 5,68 percent for Cwcowl-Cwcowl). 27 A comparison of the mean concentration levels from CBgi and Ccowl data shows that the cowl sampler undersamples by approximately 10,25 percent when compared with the BGI sampler. A comparison of the mean concentration levels from CBgi and CwCowl data shows that, on average, the without-cowl sampler, i.e. a normal 10 mm cyclone without a shield, oversamples by approximately 12,21 percent when compared with the BGI sampler. However, a comparison of the mean concentration levels from CCowl and CwCowl data shows that, on average, the sampler without-cowl under-samples by 2,21 percent when compared with the cowl sampler. From a comparison of the results of the mean dust concentrations measured at the return with the various sampler pairs, the following conclusions were drawn (Appendix A Figures A13 to A18, and Table 6c): On average, with the same sampler types, there was a bias of 4,6 percent in the concentration measurement (i.e. 2,30 percent for CBgi-CBgi; 8,08 percent for CCowlCcowl; and 3,44 percent for Cwcowl-Cwcowl). A comparison of the mean concentration levels from CBgi and CCowl data shows that the cowl sampler undersamples by approximately 1,1 percent when compared with the BGI sampler. A comparison of the individual mean concentration levels derived from CBgi and CwCowl data shows that, on average, the without-cowl sampler, i.e. a normal 10 mm cyclone without a shield, oversamples by approximately 28,63 percent when compared with the BGI sampler. However, a comparison of the mean concentration levels derived from CCowl and CwCowl data shows that, on average, the sampler without cowl over-samples by 19,31 percent when compared with the cowl sampler. Sampler biases on the combined data (intake, operator and return) using measured concentrations by various sampler types are shown in Table 6d. Table 6d: Overall results of the sampler biases 28 CBGI CCOWL cwcowl CBGI 6,04 Overall (% bias) CCOWL 7,10 (CCowl under-samples) -4,20 CWCOWL -15,71 (Cwcowl over-samples) -1,60 (Cwcowl over-samples) 4,76 By combining and comparing the results of the mean dust concentrations measured at the intake, operator and return, the following conclusions were drawn (Appendix A Figures 6a to 6f, and Table 6d): On average, with the same sampler types, there was a bias of 5 percent in the concentration measurement (i.e. 6 percent for CBGI-CBGI; 4,2 percent for CCoWi_-CCoWi_; and 4,76 percent for Cwcowl-Cwcowl). A comparison of the mean concentration levels derived from Cbgi and CCowl data from the different respective positions shows that the cowl sampler undersamples by approximately 7,1 percent when compared with the BGI sampler. A comparison of the individual mean concentration levels derived from Cbgi and CwCowl data from the different respective positions shows that, on average, the sampler without-cowl, i.e. a normal 10 mm cyclone without a shield, oversamples by approximately 15,1 percent when compared with the BGI sampler. It is interesting to note that the highest concentration measured in this pair was 2,95 mg/m3, which is low in comparison with the other pairs (see Table 6b). However, a comparison of the mean concentration levels derived from CCowl and CwCowl data from the different respective positions shows that, on average, the without-cowl sampler, i.e. a normal 10 mm cyclone without a shield, oversamples by a mere 1,6 percent when compared with the cowl sampler. In order to conclusively determine the significant differences in measured concentration levels (variance and mean), using various sampler pairs, further statistical analyses were carried out between the sampler pairs at 95 percent level of significance. 6.1 Statistical Analyses 29 An analysis of the frequency distribution of the concentration values obtained by the samplers yielded a set of histograms. Comparing the sample distributions with a normal distribution leads to rejection of the hypothesis that the sample distribution was normal. Therefore, plotting the histogram of the loge transform of the dust concentration data led to the conclusion on the hypothesis that the measurements were loge normally distributed. The improved fit of the normal distribution to the loge transform of the concentration data was obvious. 6.1.1 F-Test Hypothesis First, F-test statistics were calculated to determine whether there was a statistical difference in the variance obtained between two samplers. A test of hypotheses was carried out to determine the difference in the variances of the measured concentration levels between sampler pairs. The test of hypothesis was developed to compare the sample variance measured with two sampling instruments (cA and cB). The null and alternative hypotheses for the tests for the various sampler pairs were: Hypothesis (a) H0: cBGI = cBGI, and any observed variability is due to chance H1: cBGI * cBGI, and there is some variability between the samplers. Hypothesis (b) H0: cCowl = cCOWL, and any observed variability is due to chance H1: cCOWL * cCOWL, and there is some variability between the samplers. Hypothesis (c) H0: cWCOWL = cWCOWL, and any observed variability is due to chance H1: cWCOWL * cWCOWL, and there is some variability between the samplers. Hypothesis (d) H0: cBGI = cCOWL, and any observed variability is due to chance H1: cBGI * cCOWL, and there is some variability between the samplers. 30 Hypothesis (e) H0: cbgi = owcowl, and any observed variability is due to chance Hi: cBGI * cWCOWL, and there is some variability between the samplers. Hypothesis (f) H0: cCOWL = cWCOWL, and any observed variability is due to chance H1: cCOWL * cWCOWL, and there is some variability between the samplers. In the above F-test, hypothesis H0 states that the sample variances from both samples (cA and cB) are equal. On the other hand, the alternative hypothesis states that there is some difference in sample variances between two samplers. The hypotheses stated were cA = cB and cA * cB. All analyses were two-tailed (95 percent confidence level) to account for both conditions cA < cB and cA > cB. Hypothesis tests were carried out for sampler pairs, and the results of the F-test statistical analyses are given in Tables 6.1.1a and 6.1.1b. The tables show that all hypothesis test pairs, H0, were accepted. In other words, the level of confidence that the dust measurement readings from the two samplers are not affected is 95 percent. For example, for the test hypothesis (d), the number of degrees of freedom associated with the numerator yCowl = 124 - 1 = 123; for the denominator yBGI = 124 - 1 = 123. At the 0,95 confidence level for 123 (123 degrees of freedom, two-tailed test), we would accept that H0 as the F-value (1,127) was within the critical F-value range (0,74<F<1,348). From the F-test analysis, following conclusions were made: F-test results indicate that the sample variances from measured dust concentrations using various sampler pairs are equal. In other words, dust readings measured by the two samplers are not affected at 95 percent level of confidence on the basis of differences in dust concentration variances. Table 6.1.1a: Results of the F-test hypothesis (on transformed values) 31 Mean Variance (S2) Size F-statistic Critical F-value Hypothesis Hypothesis (a) CBGI 0,744 CBGI 0,704 0,462 0,432 23 23 1.069 0,488<F<2,048 Accept H0 Hypothesis (b) CCOWL 0,620 CCOWL 0,629 0,752 0,814 40 40 0.924 0,586<F<1,704 Accept H0 Hypothesis (c) CWOWL 1,001 CWCOWL 0,935 0,394 0,433 22 22 0.911 0,479<F<2,084 Accept H0 Table 6.1.1b: Results of the F-test hypothesis (on transformed values) Mean Variance (S2) Size F-statistic Critical F-value Hypothesis Hypothesis (d) CBGI 0,123 1,059 CCOWL 0,094 0,939 123 123 1,127 0,741<F<1,348 Accept H0 Hypothesis (e) CBGI -0,085 0,799 CWCOWL 0,018 0,958 15 15 0,832 0,403<F<2,484 Accept H0 Hypothesis (f) CCOWL 0,470 1,156 CWCOWL 0,521 1,088 26 26 1,062 0,512<F<1,955 Accept H0 6.1.2 t-Test Hypothesis F-test of hypothesis results showed that there is no significant difference between the sample variances. Therefore an additional statistical test was performed to identify if there is any significant difference between sample mean concentrations. A paired t-test was performed on the set of all the sampler-pair data to determine whether there was a statistical difference in the loge-transformed (normally distributed) concentration levels between the sampler pairs. A paired f-test of hypotheses was developed to compare the mean concentration levels measured with two sampling instruments (pA and pB). A paired t-test analysis procedure would probably have a smaller error term than the corresponding unpaired procedure because it removes the variability caused by differences between the pairs. The null and alternative hypotheses for the tested sampler pairs were: 32 Hypothesis (a) Ho: Mbgi = Mbgi Hi : Mbgi * Mbgi Hypothesis (b) Ho: Mcowl = Mcowl Hi: M-cowl * Mcowl Hypothesis (c) Ho: Mwcowl = Mwcowl Hi: Mwcowl * Mwcowl Hypothesis (d) Ho: Mbgi = Mcowi Hi: Mbgi * MCowl Hypothesis (e) Ho: Mbgi = Mwcowl Hi: Mbgi * Mwcowl Hypothesis (f) Ho: Mcowl = Mwcowl Hi: Mcowl * Mwcowl In the paired t-test, hypothesis Ho states that the mean dust concentration levels from both samples (ma and mb) are equal. On the other hand, the alternative hypothesis states that the two samplers in fact measure different concentration levels. It is therefore necessary to use hypothesis testing to accept or reject Ho. For this work, a standard 95 percent confidence level was chosen. As the hypotheses stated were ma = MB and ma * MB, all analyses were two-tailed to account for both conditions MA < MB and ma > MB. Therefore, the critical t-values were determined by to,o25 rather than to,o5. The results of the t-test statistical analyses are given in Tables 6.i.2a and 6.i.2b. 33 P-values are often used in hypothesis tests in which one either rejects or fails to reject a null hypothesis. The p-value represents the probability of making a Type 1 error, that is rejecting the null hypothesis when it is true. The smaller the p-value, the smaller is the probability that one would be making a mistake by rejecting the null hypothesis. In this study, a cut-off p-value of 0,05 was used (95 percent confidence level). From the analysis tables, we observe that with various degrees of freedom, the large p-value is >0,05, suggesting that the measured mean concentration levels are consistent with the null hypothesis, H0 = pA = pB, i.e. the dust concentration measured by pairs of cyclones is not affected at the 95 percent level of confidence. Further, the confidence interval for the difference between the pairs of cyclones includes zero, which suggests that there is no difference between the measured concentration levels between the various sampler pairs. Table 6.1.2a: Results of the paired t-test hypothesis (on transformed values) Statistics Mean Variance Size t-statistic Critical t-value P-value 95 % CI Hypothesis Hypothesis (a) CBGI 0,744 0,462 CBGI 0,704 0,432 23 23 1,70 -2,01 < t < 2,01 0,103 (-0,009, 0,089) Accept H0 Hypothesis (b) CCOWL 0,620 0,752 CCOWL 0,629 0,814 40 40 -0,32 -1,99< t < 1,99 Hypothesis (c) CWCOWL 1,001 0,394 CWCOWL 0,935 0,433 22 22 1,89 -2,02 < t < 2,02 0,750 (-0,063, 0,046) Accept H0 0,072 (-0,007, 0,138) Accept H0 From Table 6.1.2b, we observe that the t-statistic CBG|-Ccowl was 1,78. This indicates that the mean dust concentration level from the BGI sampler (CBGI) was generally higher than the mean dust concentration level from the cowl sampler (CCOWL) at all positions. Similarly, the t-statistic CBGI-CWCOWL was -1,91 and the t-statistic CCOWL-CWCOWL was - 1,09. This indicates that the mean dust concentration level from the BGI sampler (CBGI) was generally less than the mean dust concentration level from the without-cowl sampler (CWCOWL) at all positions. Similarly, the mean dust concentration level from the cowl sampler (CCOWL) was generally lower than the mean dust concentration level from the without-cowl sampler (CWCOWL) at all positions. 34 From the f-test analysis, following conclusions were made: The mean dust concentration measured by pairs of cyclones is not significantly affected at 95 percent level of confidence. In other words, there is no difference between the measured mean dust concentration levels between various sampler pairs. It is also important to note that, f-statistic, indicates that mean dust concentration level from the BGI sampler (CBGI) was greater than the mean dust concentration level from the Cowl sampler (CCOWL) at all locations. Similarly, f-statistic, indicates that mean dust concentration level from the BGI sampler (CBGI) was less than the mean dust concentration level from the without Cowl sampler (Cwcowl) at all locations. Table 6.1.2b: Results of the paired t-test hypothesis (on transformed values) Statistics Mean Variance Size t- statistic Critical t- value P-value 95 % CI Hypothesis Hypothesis (d) CBGI 0,123 CCOWL 0,094 1,059 0,939 123 123 1,78 -1,97 < t < 1,97 0,078 (-0,003, 0,062) Accept H0 Hypothesis (e) CBGI -0,085 CWCOWL 0,018 0,799 0,958 15 15 -1,91 -2,05 < t < 2,05 Hypothesis (f) CCOWL 0,470 CWCOWL 0,521 1,156 1,088 26 26 -1,09 -2,01 < t < 2,01 0,076 (-0,219, 0,013) Accept H0 0,284 (-0,147, 0,045) Accept H0 6.2 Particle Size Distribution Table 6.2a summarises the respirable fractions and the dust concentrations measured in the UK using the South African cyclone with cowl and without cowl. The respirable fractions were determined as the ratio of the cyclone concentration to the concentration measured using a reference probe. It can be seen that the average measured dust concentrations (from six tests) were 34,63 mg/m3 and 42,09 mg/m3 respectively. From the 35 measured concentration alone, we can conclude that the addition of cowl results in a "measured concentration bias" of approximately 18 percent. However, the study could not conclusively report if the addition of cowl results in under-estimation or over-estimation of dust concentration. Table 6.2a: Measured respirable dust fraction and mass concentrations using the cowl and without-cowl samplers (Kenny et al., 1998) Test No. 1 2 3 4 5 6 Av. Test Concentration (mg/m3) 35 68 140 122 106 165 106 Respirable Fraction (%) Without Cowl Cowl 40 22 39 36 42 34 27 26 59 41 35 32 40,33 31,83 Mass Concentration in (mg/m3) Without Cowl With Cowl 14,00 26,52 58,80 32,94 62,54 57,75 42,09 7,70 24,48 47,60 31,72 43,46 52,80 34,63 Table 6.2b shows the average size distribution of the sample dust measured from six laboratory tests in the UK using the South African cyclones with cowl and without cowl (Figure 6.2). One observes from the plot that the cumulative percentage of particles finer than 10 pm measured using the cyclones with and without cowl were 84 percent and 90 percent respectively. Similarly, the cumulative percentage of particles finer than 10 pm measured using the reference test probe was 66 percent. Generally, more than 90 percent of the sampled mass is contained particles with volume diameters less than 10 pm. Using cyclones; the actual numbers of particles counted having diameters exceeding 10 pm were in all cases very small and given the level of coincidence counts, not significant (Kenny et al., 1998). The CEN respirable convention requires 1,3 percent of particles having an aerodynamic diameter of 10 pm to be sampled; however, no particles with aerodynamic diameters greater than 16 pm should be sampled. 36 Table 6.2b: Measured particle size distributions (by volume) using the cowl and without-cowl samplers (Kenny et a!., 1998) Test No. 1 2 3 4 5 6 Reference Probe Without Cow! Cow! % mass finer % mass finer % mass finer 5 gm 10 gm 38 gm 5 gm 10 gm 38 gm 5 gm 10 gm 38 gm 52 85 100 66 78 100 60 91 100 55 83 99 83 97 100 57 68 100 57 84 98 68 88 100 80 91 100 43 74 98 73 90 100 58 75 100 35 70 100 69 94 100 66 87 100 45 79 100 71 92 100 76 91 100 Finally, the laboratory tests carried out in the UK (at the HSL) deduced the following (Kenny et al, 1998): The modifications made to the South African cyclones did not either improve or degrade their performance in the laboratory experiment. However, consideration should be given to adding the cowl to the South African design because this provides a simple means of reducing the wind and orientation dependence of the cyclone inlet in practical use. The laboratory study concluded that the inadequate field performance of South African cyclones was attributable to the particular conditions of use in mines, and may be related to the sample-handling procedures. In this work, laboratory study of Kenny, Baldwin and Maynand (198) was re-visited to clarify the evaluation structure on its definitive results. The following critical comments were made on the laboratory evaluation of cowl samplers that was carried out in UK: The laboratory test aerosol for sampler evaluation contained too few large particles compared with the bulk of the coal dust supplied by the CSIR Miningtek, Kloppersbos. The dust supplied was specifically sized to ensure that although there were sufficient of the smaller-size particles available, it also contained a significant amount of larger dust particles (more than 50 percent of the bulk coal contained particle sizes (volume diameter) greater than 13 gm). Due to the losses of larger 37 particles within the aerosol generation and dispersal system in the laboratory, the conclusive inference on the influence of a large amount of non-respirable particles on individual cyclone collection performance is questioned. The tests may have failed to ensure that the cyclone has the ability to separate the larger dust particles from the respirable range. Although there was a 17 percent difference in the mean measured dust concentration between the samplers with cowl and without cowl, at very high test-dust concentration levels (35 mg/m3 to 165 mg/m3), the conclusion was reached that the modifications made to the South African cyclones did not either improve or degrade their performance in the laboratory experiment. In this second phase of the project, particle size analysis on the sample dust collected from the field experiments was carried out using a Fritsch size analyser at the CSIR Miningtek laboratory. The volume diameter of the particle data was converted to aerodynamic diameter by a factor of 1.183 using the density of the particle. Table 6.2c summarises the particle size distributions from the BGI, cowl and without-cowl sampler dust. 38 Table 6.2c: Measured particle size distributions from the BGi, cowl and without-cowl samplers (using a Fritsch size analyzer) Sampler Pair0 Cwcowl (2,37 mg/rn3) Cbgi (3,89 mg/m3) Cwcowl (5,21 mg/m) Cwcowl (6,13 mg/m) Ccowl (5,62 mg/m3) Cwcowl (5,56 mg/m3) Cwcowl (10,59 mg/m3) Cwcowl (9, 37 mg/m) Ccowl (39,92 mg/m3) Cwcowl (36,60 mg/m3) Cwcowl (53,26 mg/m3) Cwcowl (38,17 mg/m) Cbgi (17,11 mg/m3) Cwcowl (1261,72 mg/m) Ccowl (91,81 mg/m3) Cwcowl (2991 mg/m) Ccowl (1156 mg/m3) Cwcowl (42,89 mg/m) Cbgi (1201 mg/m3) Cwcowl (2471 mg/m) Particulate % Finer, pm < 16 % < 50 % < 84 % 0,75 2,20 6,50 0,70 2,00 3,50 0,85 3,00 24,0 1,30 4,50 27,0 1,50 4,50 27,0 1,30 3,10 27,0 1,30 3,50 17,0 0,95 2,40 7,00 1,30 3,50 23,0 1,30 4,00 26,0 1,30 4,30 25,0 1,40 4,30 26,0 1,30 3,50 22,0 2,90 17,0 31,0 1,50 6,00 24,0 5,00 20,0 35,0 2,00 10,0 27,0 1,30 4,20 24,0 1,40 4,00 26,0 2,00 12,0 29,0 Dust sample pairs with enough weight for size analysis (greater than 2 mg) were collected for analysis. However, only a few samples with low concentrations were analyzed due to the sample mass limitations of FSA for size characterization. Therefore, in order to obtain a higher mass of dust on the filter, samples collected at a distance of greater than 2,5 m from the operator's position , whose concentration for the sampling period was in the range of 36 mg/m3 to 2 991 mg/m3 were analysed. 0 Numbers in parentheses are the respective measured dust concentrations for the sampling period. 1 This concentration should not be interpreted as the operator position concentration. 39 The plots of the size analysis of the sampled dust are shown in Appendix A (Figures A19 to A29) using the BGI, cowl and without-cowl samplers. From the results, the following conclusions were drawn: From the size distribution plots (Figures A19, A20 to A29), we notice that the mineface atmospheric dust has two relative maximums, called modes (term adopted from statistics). Thus, the size distribution depicted can be referred to as a bi-modal distribution of mine-face atmospheric dust. The two identified modes are in the respirable (< 10 pm) and non-respirable (> 10 pm) size range of the dust size distribution. From the size analysis data of all the measured dust concentrations in the range 2,37 mg/m3 to 10,59 mg/m3, it is seen that the sampled respirable dust contained particles greater than 10 pm, except for the BGI sampler dust (3,89 mg/m3 concentration). At very high measured dust concentrations, the cowl sampler lowered the loading of large particles (Figures A26). From the size distribution analysis data, we can conclude that the cowl samplers are the preferred samplers for taking engineering samples at high dust concentrations. However, the results showed that at very high dust concentrations, there is a huge variation in both the measured dust concentration and the size distribution for a paired sample (Figures A25 to A29). For a pair of cowl and without-cowl samplers, at a concentration of 5,6 mg/m3, the D50 for the cowl and without-cowl samplers was 4,5 pm and 3,1 pm respectively (Figure A9). At a concentration of 38 mg/m3, the D50 for cowl and without-cowl samplers was 3,5 pm and 4,0 pm respectively (Figures A24). At concentrations ranging from 5 mg/m3 to 1 200 mg/m3, the BGI performed well on the criteria of D50 and the presence of non-respirable particles. The analysis of the various samplers indicated the presence of wide size distributions for the same mass concentration of dust. Efforts were made to carry out a size analysis of dust mass concentrations of less than 5 mg/m3. The analysis on the dust sample concentrations in the range of 5 to 6 mg/m3 for an approximate sampling period of 300 minutes did indicate the presence of non-respirable particles (both with cowl and without cowl samplers). The D50 value of 4,0 pm with the use of a sample flow rate of 2,2 L/min, regardless of sampler type, could not be obtained for the dust samples analysed. The size analysis on the BGI sampler did indicate that the amount of non-respirable particles present is lower than that with the cowl and without-cowl samplers. 40 It is thus evident that, although the dust mass collection of the samplers conforms to the requirements, the size distribution of the dust is not concomitantly accurate. 7 Discussions Extensive work has been done to identify the factors associated with the variability of measured dust concentrations overseas. In connection with the current research work, the following studies reflect the extent of and the possible reasons identified for the variability in the measured sample concentrations: A Government Accounting Office (1975) report to the US Congress indicated that, under certain conditions, the error associated with respirable mine dust samples could be as great as 50 percent. A US National Bureau of Standards (1975) study focused specifically on gravimetric sampling and analysis, examining each step of the sampling process such as dust weighing and pump flow variation. It was concluded that under tightly controlled conditions, with a "well-trained" technician, the average standard deviation associated with the process was 0,39 mg/m3, or 19 percent (at the 2 mg/m3 standard dust concentration). NIOSH (1976) found that in high-risk mine sections, which failed to comply with the 2 mg/m3 standard, the coefficient of variation in dust measurements was 91,6 percent. The National Resource Council in the USA (1980) concluded that the uncertainties associated with spatial and temporal variation in dust estimates from machinemounted samplers precluded this method for estimating personal exposures. A study by Page and Jankowski (1984) comparing the dust measurements from paired RAM gravimetric samplers in a longwall mining operation gave results, expressed as dust ratios, of 0,41 to 1,63. The authors attributed this variation to differences in the aerosol cloud being sampled, to airflow velocity at the face and to cyclone orientation. Comparative studies of personal and fixed-point (area) samplers by Breslin et al. (1983) reported that the coefficient of variation of measured mine dust concentrations was typically less than 20 percent. In 1986, a study by Kissell et al. reviewed several factors contributing to the variability in measured dust concentration. They concluded that sampler position, geological variation in the composition of the coal (for variability in measured free silica), production factors such as deep or continuous cutting and failure to control known 41 sources such as shuttle-car loading, play an crucial role in dust concentration sample results. Sampling of and laboratory variability in respirable mine dust was studied by Hall et al. (1997) using 23 and 20 pairs of dust samples from coal and non-coal mines. In coal mines, mine dust ratios (larger to smaller values) exceeded 1,5 in half of the paired samples and 2,5 in 10 percent of the pairs. The variability of mine dust was somewhat lower in non-coal mines, with 50 percent of the samples having ratios greater than 1,13. Ten per cent of the samples demonstrated ratios of 6,19. Discussions indicated that in coal mines, the position of the sampler was an important contributor to the variability. Machine-mounted samplers showed a 40 percent improvement in variability for all measured parameters compared with personal samplers. In spite of the problems facing the mining industry with regard to the accuracy of dust measurements, the industry's goal should be to find a standard measurement procedure using well-accepted sampling instruments. In an underground environment, the ideal sampler would collect a sample that represents the mine-ambient aerosol. In this study, the cyclone efficiencies of samplers were not evaluated, as a particle size analyser is needed to determine the size fraction characteristics of both the ambient and sample dust. Until now, it is not known whether the ambient coal dust size characteristics remain in a closely defined size range during the underground production period. Had data been available on such size characteristics, it would have been possible to determine the collection efficiency of samplers and the penetration contours to arrive at a conclusive judgment. In an underground face, there is no clear air direction in terms of upwards or sideways as one finds in a test tunnel. Due to the harsh conditions that prevail underground, the effect of wind (speed and direction) on sampler performance at the operator could not be determined. However, the use of correlation coefficients and regression lines at individual positions does allow one to speculate as to the probable reasons. Although the two sampling instruments differ with regard to their design, cost and ease of use, the statistical analysis showed that there were no significant differences between the mean dust concentrations measured by the various sampler pairs tested in the underground section. From all the measured sampling positions, the coefficient of correlation (r) determined was better than 0,95 for various regression lines (combinations 42 of CBgi, CCowl, and CwCowl). Unlike in the laboratory experiments, the reference concentration could not be measured underground. Overall, the following inferences can be drawn from the analysis of the concentration data from field tests: On average, for the same sampler types (Cbgi-Cbgi, Ccowl-Ccowl and CwcowlCwCowl), there was a bias of 5,0 percent in the mean concentration measurement at the intake, operator and return positions. A comparison of the mean concentration data for CBgi and CCowl from the different positions shows that, on average, the cowl sampler under-samples by approximately 7,1 percent compared with the BGI sampler. A comparison of the combined mean concentration data from CBgi and CwCowl from the different respective positions shows that, on average, the sampler without cowl over-samples by approximately 15,1 percent compared with the BGI sampler. However, it can be stated with confidence that this relation will hold for concentrations above 5,0 mg/m3. A comparison of the combined CCowl and CwCowl data from the different positions shows that, on average, the sampler without cowl over-samples by a mere 1,6 percent compared with the cowl sampler. From the statistical analysis (/-test and F-test), the mean dust concentration measured by pairs of cyclones is not significantly affected at 95 percent level of confidence. in other words, there is no difference between the measured mean dust concentration levels between various sampler pairs. Overall, the following inferences can be drawn from the size analysis of the sample dust from various cyclones from field tests: From the size distribution plots, we notice that the mine-face atmospheric dust has two relative maximums, called modes. Thus, the size distribution depicted can be referred to as a bi-modal distribution of mine-face atmospheric dust. The two identified modes are in the respirable (< 10 pm) and non-respirable (> 10 pm) size range of the dust size distribution. From the size analysis data of all the measured dust concentrations in the range 2,37 mg/m3 to 10,59 mg/m3, it is seen that the sampled respirable dust contained particles greater than 10 pm, except for the BGI sampler dust (3,89 mg/m3 concentration). 43 At very high measured dust concentrations, the cowl sampler lowered the loading of large particles. From the size distribution analysis data, we can conclude that the cowl samplers are the preferred samplers for taking engineering samples at high dust concentrations. However, the results showed that at very high dust concentrations, there is a huge variation in both the measured dust concentration and the size distribution for a paired sample. For a pair of cowl and without-cowl samplers, for the same dust concentration levels, the D50 of both samples were different. The analysis of the various samplers indicated the presence of wide size distributions for the same mass concentration of dust. At concentrations ranging from 5 mg/m3 to 1 200 mg/m3, the BGI performed well on the criteria of D50 and the presence of non-respirable particles. The analysis on the dust sample concentrations in the range of 5 to 6 mg/m3 for an approximate sampling period of 300 minutes did indicate the presence of nonrespirable particles (both with cowl and without cowl samplers). The D50 value of 4,0 pm with the use of a sample flow rate of 2,2 L/min, regardless of sampler type, could not be obtained for the dust samples analysed. The size analysis on the BGI sampler did indicate that the amount of non-respirable particles present is lower than that with the cowl and without-cowl samplers. It is thus evident that, although the dust mass collection of the samplers conforms to the requirements, the size distribution of the dust is not concomitantly accurate. 8 Conclusions In summary, following conclusions were made from the current study: The work carried out in the field demonstrates equivalent dust concentration results from the two sampler types in different positions within the mine. The extent of the agreement between the two sets of results seems to be excellent and much better than that for results obtained in the laboratory results or in non-mining workplaces. The field study failed to conclusively determine whether the modifications made to the South African cyclone either improve or degrade its performance for the conditions tested, based on the results of the size analyses. From the field observations, it can be stated that the cowl sampler can be used for engineering sampling to avoid the dependence of the cyclone inlet on orientation in the collection of particles. 44 The data analysis and field experience have shown that in comparison with other samplers, BGI samplers greatly reduce the amount of oversized particles deposited onto the filter from the inversion of cyclones. Also, due to the design of its cyclone inlet, the BGI sampler is not affected by orientation dependence. The field observations did indicate that when collecting samples closer to the face, where high concentrations of dust are present, sample handling is extremely difficult. At high dust concentrations, the probability of non-respirable particles depositing onto the filter from the grit pot is very high when the sampling pump is switched off. However, the BGI sampler, due to its design, alleviates this problem to some extent. In both the laboratory and field tests, a sufficient mass of sample dust was required for analysis with the Fritsch analyser to determine the presence of non-respirable particles. This led to the conclusion that there may have been non-respirable particles in the concentrations below 5 mg/m3 (as found from this study). The size analysis of the collected dust mass for sample concentrations greater than 5 mg/m3 did not follow either BMRC or ISO/ACGIH/CEN respirable curves. The evaluation of the same samplers in the UK laboratory produced conflicting results. The cowl sampler could not be recommended for personal sampling because the results of the size analysis are inconclusive and wide-ranging. Finally, question of cowl under-sampling or over-sampling could not be resolved confidently from size distribution data at concentrations below 5 mg/m3 despite the insignificant differences obtained from comparison of mass concentrations alone. 9 Recommendations From the conclusions drawn, the following recommendations for future research are proposed, although it is suggested that there should be a paradigm shift in the existing sampling procedure and instruments in the future project: The results were inconclusive with regard to the recommendations on the usage of adding a cowl to the cyclone in the personal sampling method. However, the use of a cowl for taking engineering samples can be recommended on the grounds that it lowers sample-handling errors, the probable reason being that in practice it reduces factors such as the wind and orientation-dependence of the cyclone inlet. Further work on proving the use of the cowl sampler for personal sampling could be considered. 45 At high dust concentrations, it was difficult to determine which sampler measured the "true concentration." Therefore, future field tests should incorporate the aspect of "impactor sampling" from which both the mass concentration and the size distribution of ambient particles can be accurately determined. Despite some of the cyclones being accepted for sampling in the industry, concerns were expressed about the cyclone performance with respect to the BMRC and ISO/ACGIH/CEN respirable curves. Fears were also expressed about whether the cyclones do follow the BMRC respirable curve under concentrations below 5 mg/m3. Further work for recommendations on the use of cowl sampler for personal sampling could be considered. Similar studies should be carried out to compare these samplers with CIP10 samplers and to investigate their variances and size distribution. The different makes of samplers should also be investigated to determine the relationship between them, not only with regard to the amount of dust collected but also in terms of the size characteristics of the dust. Immediate attention needs to be given to analysing both the compliance and non compliance samples from the mines which have difficulty in maintaining the 1997 DME directive. On the basis of this outcome, there needs to be clarification on the matter of evaluating all the cyclones currently being used in the mines. An extensive facility, including a chamber and a particle size analyser for analysing sample mass on compliance samples of less than 5 mg/m3, needs to be established for this serious issue of sampling. A respirable dust programme needs to be established covering the area of dust sampling, measurement and data analysis that will enable the mines to resolve the technical aspects of exposure assessment. 46 10 References ACGIH (American Conference of Governmental Industrial Hygienists), 1991-1992. Threshold limit values," ACGIH, USA, p 42. ACGIH (American Conference of Governmental Industrial Hygienists), 1991. Notice of intended change -Appendix D - Particle size-selective sampling criteria of airborne particulate matter, Applied Occup. Environ Hygiene, Vol. 6, No. 9, USA. Bartley et al., Bartley, D. L., Chen, C. C., Song, R., Fischbach, T. J., 1994. Respirable Aerosol Sampler Performance Testing, American Industrial Hygiene Association Journal (No. 11, Vol. 55), USA, pp. 1036-1046, USA. Berry, R.D., 1990. Effect of flow pulsations on cyclone penetration, AAAR 1990 Annual Meeting, Abstract Book, UK. BMRC (British Medical Research Council), 1952. Recommendations of the BMRC Panels relating to selective sampling. Taken from the Minutes of a Joint Meeting of Panels 1, 2 and 3 held on 4 March. Breslin, J.A., Page, S.J. and Jankowski, R.A., 1993. Precision of personal sampling of respirable dust in coal mines, USBM RI 8740. Cecala, A.B., Volkwein, J.C., Timko, R.J. and Williams, K.L., 1983. Velocity and orientation effects on the 10-mm Dorr-Oliver cyclone, BOM RI 8764. CEN (Comite European de Normalization), 1991. Size fraction definitions for measurement of airborne particles in the workplace, Pr EN 481, CEN, Luxembourg. DME (South African Department of Minerals and Energy), 1997. Guidelines for the gravimetric sampling of airborne particulates for risk assessment in terms of the Occupational Diseases in Mines and Works Act, No. 78 of 1978, Parent Document, 3rd Edition. General Accounting Office, US Government, 1975. Improvements still needed in coal mine dust sampling program and penalty assessment and collection, Report RED-76-56, 47 1975. Hall, T.A., Corn, M. and Zeger, S., 1997. Respirable dust and free silica variation in mine environments: Strategies for mine dust measurement, USBM, USA ISO (International Standards Organization), 1991. Air quality: Size fraction definitions for measurement of airborne particles, CD 7708. ISO, Geneva. Kenny, L., Baldwin, P.E.J. and Maynard, A.D., 1998. Respirable dust sampling at very high concentrations, HSL, UK. Kissell, F.N., Ruggier, S.K. and Jankowski, R.A., 1986. How to improve the accuracy of coal mine dust sampling, Am. Ind. Hyg. Assoc. J, Vol. 47, No. 10, pp 602-606. Maynard, A.D. and Kenny, L.C., 1995. Performance assessment of three personal cyclone models, using an Aerodynamic Particle Sizer, Journal of Aerosol Science, Vol. 26, No. 4, pp 671-684. National Bureau of Standards, US, 1975. An evaluation of the accuracy of the coal mine dust sampling program administrated by the Department of Interior, Final Report to the Senate Committee on Labor and Public Welfare, Washington, DC. NIOSH, CDC, 1976. Statistical analysis of (Mining Enforcement Safety Administration) respirable dust data. National Resource Council, US, 1980. Measurement and control of respirable dust in mines. Oberholzer, J.W., Du Plessis, J.J.L., Belle, B.K. and Eroglu, N., 1998. Gravimetric monitoring of excessively high dust levels, SIMRAC Symposium, South Africa. Oberholzer, J.W., 1997. Discrepancies in the measuring of dust using gravimetric samplers, CSIR Miningtek, South Africa. Oberholzer, J.W., 1998. Dust instrumentation, STEP Project Y3691, CSIR Miningtek, South Africa. 48 Page, S. and Jankowski, R., 1984. Correlations between measurements with RAM-1 and gravimetric samplers on longwall shearer faces, Am. Ind. Hyg. J, Vol 45, No. 9, pp 610 616. Parent Document. 1994. Guidelines for the gravimetric sampling of airborne particulates for risk assessment in terms of the Occupational Diseases in Mines and Works Act No 78 of 1973, Third Edition, South Africa. 49 APPENDIX A 50 Table A1: Linear regression details between pairs of samplers at the intake Sample Pair @ Intake Cbgi - Cbgi CcowL - CcowL Cwcowl - CwCOWL Cbgi - Ccowl Cbgi - Cwcowl Ccowl - Cwcowl Regression Line y = 0,6202x + 0,3839 y = 1,1587x - 0,1134 y = 0,7505 x + 0,3279 y = 0,9783x + 0,0211 y = 0,151x + 0,2437 y = 0,8345x + 0,0715 Correlation Coefficient (r) 0,881 0,975 0,687 0,987 0,090 0,850 Table A2: Linear regression details between pairs of samplers at the operator Sample Pair @ Operator Cbgi - Cbgi Ccowl - Ccowl Cwcowl - Cwcowl Cbgi - Ccowl Cbgi - Cwcowl Ccowl - Cwcowl Regression Line y = 0,8201x + 0,4258 y = 1,0746x + 0,0384 y = 0,928x + 0,0751 y = 0,9492x - 0,1984 y = 0,6017x + 1,0441 y = 1,1193x - 0,883 Correlation Coefficient (r) 0,978 0,967 0,912 0,952 0,823 0,975 51 Table A3: Linear regression details between pairs of samplers at the return Sample Pair @ Return Cbgi - Cbgi CcowL - CcowL Cwcowl - CwCOWL Cbgi - Ccowl Cbgi - Cwcowl Ccowl - Cwcowl Regression Line y = 1,0251x - 0,0041 y = 0,9032x + 0,0373 y = 0,9288x + 0,1008 y = 0,9409x + 0,0589 y = 1,5487x - 0,3487 y = 0,8734x + 0,4167 Correlation Coefficient (r) 0,989 0,859 0,942 0,910 0,938 0,927 Table A4: Combined linear regression details between pairs of samplers Sample Pair (all positions) Cbgi - Cbgi Ccowl - Ccowl Cwcowl - Cwcowl Cbgi - Ccowl Cbgi - Cwcowl Ccowl - Cwcowl Regression Line y = 0,848x + 0,247 y = 1,085x - 0,1122 y = 0,929x + 0,076 y = 0,914x + 0,0267 y = 1,103x + 0,068 y = 1,002x + 0,0396 Correlation Coefficient (r) 0,986 0,978 0,951 0,969 0,953 0,985 52 Figure A1: Plot of combined dust concentrations obtained from two BGI samplers at the section intake 53 3 ,, 2.5 E o> Ec 0) Q. Etmo 2 o3o 1 0.5 0 --------------------- 1--------------------- 1--------------------- 1--------------------- 1--------------------- 1--------------------0 0.5 1 1.5 2 2.5 3 Dust concentration measured by Cowl sampler in mg/m3 Figure A2: Plot of combined dust concentrations obtained from two cowl samplers at the section intake 54 Dust concentration measured by W ithout Cowl sam pler in m g/m 3 O Figure A3: Plot of combined dust concentrations obtained from two without-cowl samplers at the section intake 55 Dust concentration measured by Cowl sam pler in m g/m 3 O Figure A4: Plot of combined dust concentrations obtained from BGI and cowl samplers at the section intake 56 Figure A5: Plot of combined dust concentrations obtained from BGI and withoutcowl samplers at the section intake 57 Figure A6: Plot of combined dust concentrations obtained from cowl and withoutcowl samplers at the section intake 58 Figure A7: Plot of combined dust concentrations obtained from two BGI samplers at the operator 59 n Dust concentration measured by Cowl sampler in mg/m3 Figure A8: Plot of combined dust concentrations obtained from two cowl samplers at the operator 60 Figure A9: Plot of combined dust concentrations obtained from two without-cowl samplers at the operator 61 Figure A10: Plot of combined dust concentrations obtained from BGI and cowl samplers at the operator 62 Dust concentration measured by BGI sampler in mg/m3 Figure A11: Plot of combined dust concentrations obtained from BGI and withoutcowl samplers at the operator 63 Figure A12: Plot of combined dust concentrations obtained from cowl and withoutcowl samplers at the operator 64 Figure A13: Plot of combined dust concentrations obtained from two BGI samplers at the section return 65 Dust concentration measured by Cowl sam pler in m g/m3 5 4.5 y = 0.9032x + 0.0373 R2 = 0.7382 4 3.5 3 2.5 2 1.5 1 0.5 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Dust concentration measured by Cowl sampler in mg/m3 Figure A14: Plot of combined dust concentrations obtained from two cowl samplers at the section return 66 Dust concentration m easured by W ithout Cowl sam pler in m g /m 3 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Dust concentration measured by Without Cowl sampler in mg/m3 Figure A15: Plot of combined dust concentrations obtained from two without-cowl samplers at the section return 67 Dust concentration m easured by Cowl sam pler in m g /m 3 3 y = 0.9409x + 0.0589 R2 = 0.8285 2.5 2 1.5 1 0.5 0 --------------------- 1--------------------- 1--------------------- 1--------------------- 1--------------------- 1--------------------- 0 0.5 1 1.5 2 2.5 3 Dust concentration measured by BGI sampler in mg/m3 Figure A16: Plot of combined dust concentrations obtained from BGI and cowl samplers at the section return 68 Dust concentration measured by BGI sampler in mg/m3 Figure A17: Plot of combined dust concentrations obtained from BGI and withoutcowl samplers at the section return 69 Figure A18: Plot of combined dust concentrations obtained from cowl and withoutcowl samplers at the section return 70 Figure A19: Particle size distribution of sampled dust from without cowl cyclone 71 Figure A20: Particle size distribution of sampled dust from BGI cyclone 72 Figure A21: Particle size distribution of sampled dust from cowl and without cowl cyclone 73 Figure A22: Particle size distribution of sampled dust from without cowl cyclone 74 Figure A23: Particle size distribution of sampled dust from without cowl cyclone 75 Figure A24: Particle size distribution of sampled dust using cowl and without cowl cyclone 76 Figure A25: Particle size distribution of sampled dust from without cowl cyclone 77 Figure A26: Particle size distribution of sampled dust from BGI and without cowl cyclone 78 Figure A27: Particle size distribution of sampled dust from cowl and without cowl cyclone 79 Figure A28: Particle size distribution of sampled dust from cowl and without cowl cyclone 80 Figure A29: Particle size distribution of sampled dust from BGI and without cowl cyclone 81