Exposure Assessment of Nanoparticles at the Workplaces - Characterization, Statistical Analysis and Instrument Comparison

Abstract

Since the 1990s, nanotechnology (NT) has rapidly grown in Korea, which has become one of the worlds leading nations for the technology. Despite the bright outlook for the future of NT, concern remains human exposure to engineered nanoparticles may lead to significant adverse health effects. Workers could be intentionally and unintentionally exposed to engineered nanoparticles during routine work or during research and development. In 2008, the number of researchers and workers involved in NT was estimated to be about 400,000 worldwide. This number is expected to grow to six million by 2020. Therefore, nanoparticle exposure assessments are important for the protection of workers. The purposes of the study were as follows: (1) to compare the exposure characteristics of nanoparticles produced in laboratory (LAB), engineered nanoparticles (ENP), and unintended nanoparticle emissions (UNP) in workplaces

(2) to develop an exposure assessment method using a task-based approach for ENP and UNP in workplaces

(3) to suggest possible data interpretation methods for real-time nanoparticle monitoring data by comparing statistical models, such as classical regression and first-order autoregressive (AR(1)) and autoregressive integrated moving average (ARIMA) models, and then investigate the effect of different averaging times on autocorrelation using field data

and (4) to determine the relationships among three nanoparticle monitoring devices, i.e., scanning mobility particle sizer (SMPS), condensation particle counter (CPC), and surface area monitor (SAM), and to compare two widely used SMPSs for harmonization. In Chapter 1, the concentrations and characteristics of nanoparticle exposure by size and type of nanoparticles for ENP and UNP are reported. The concentration and characteristics of nanoparticles at nine workplaces where the three types of workplace (LAB, ENP, and UNP) are produced were compared using real-time monitoring instruments (SMPS, CPC, and SAM) and a gravimetric method. The concentrations of UNP were higher than those of LAB and ENP for all of the metrics measured. Geometric means and geometric standard deviations of LAB, ENP, and UNP for the total number concentration measured using a SMPS were 8,458 (1.41), 19,612 (2.18), and 84,172 (2.80) particles/cm3, respectively. The concentrations of LAB, ENP, and UNP measured by a CPC were 6,143 (1.45), 11,955 (2.42), and 38,886 (2.61) particles/cm3, respectively. The surface area concentrations of LAB, ENP, and UNP were 32.79 (1.46), 93.68 (2.60), and 358.41 (2.74) μm2/cm3, respectively. The exposure characteristics and size distributions differed among workplaces. Some tasks or processes producing LAB, such as sonication and reaction (LAB-B), produced higher concentrations than those found at workplaces producing ENP or UNP. Local exhaust ventilation (LEV) could be an effective control measure for ENP. Therefore, different exposure characteristics for LAB, ENP, and UNP were observed, and different risk management strategies were required. In Chapter 2, an exposure assessment method based on a task-based approach is presented. Unlike time-weighted average (TWA) concentrations collected during shift sampling, measurement of activities of workers who perform a task that is believed to cause concentration fluctuations can precisely reflect actual exposure variations. Most of the NT industry has used batch processing rather than continuous production, making it difficult to generalize. Task-based exposure assessment is potentially appropriate for the ENP manufacturing industry because of the use of irregular processes in the workplace. Two ENP and two UNP workplaces were selected for exposure assessments. Real-time devices were used to characterize the concentration profiles and size distributions of airborne nanoparticles. Filter-based sampling was performed to measure time-weighted average (TWA) concentrations, and using an electron microscope. Workplace tasks were recorded by researchers to determine the concentration profiles associated with particular tasks. This study demonstrated that exposure profiles differed greatly in terms of concentrations and size distributions according to the tasks. The size distributions of emissions produced during tasks were different from those during periods with no activity and from the background. The airborne concentration profiles of the nanoparticles varied according to both the type of workplace and the concentration metrics. The results of this study suggest that a task-based exposure assessment could provide useful information regarding the exposure profiles of nanoparticles and can therefore be used as an exposure assessment tool. In Chapter 3, appropriate statistical models for autocorrelation were compared. Real-time monitoring is necessary for nanoparticle exposure assessment to characterize the exposure profile, but the data are autocorrelated due to the short measurement intervals. Two methods have been proposed to deal with autocorrelation in data analysis. One is a statistical approach, and the other entails changing the averaging time. This study identified possible data interpretation methods for nanoparticle monitoring data by comparing the results of statistical methods with the effect of averaging time on the reduction of the autocorrelation using field data. The classical regression model was compared with AR(1) and ARIMA. The AR(1) and ARIMA models are alternative statistical methods that remove autocorrelation effects in real-time monitoring data. Three real-time monitoring data sets were used. The first data set was for engineered nanoparticles (ENP

Fe2O3, Ti) measured at a LAB workplace, and the second data set was from an ENP (Cu, Ni) manufacturing facility. The third data set, for welding fumes, was for UNP. An SMPS with a 1-minute sampling interval was used to obtain the data. The results of a classical regression, the AR(1) model, and the ARIMA model with averaging times of 1, 5, and 10 minutes were compared. The classical regression model overestimated all of the tasks or processes due to autocorrelation. Of the three statistical methods, the AR(1) and ARIMA models had a similar capacity to adjust the autocorrelation of real-time nanoparticle data. Because of the non-stationary characteristics of real-time monitoring data in the field, the ARIMA model, which incorporates a differencing term (I) is more appropriate. When using the AR(1) model, transformation into a stationary form is necessary. Changing the averaging time did not influence the autocorrelation effect. The results of this study suggest that an ARIMA model could be used to process real-time monitoring data, especially for non-stationary data, and the use of different averaging times (within 10 minutes) had no effect on the autocorrelation with a specific statistical model. Therefore an averaging time could be used based on the instrument measuring time of one cycle in the workplace, and this was flexible depending on the data interval required to capture the effects of a series of processes for occupational and environmental nanoparticle measurements. In Chapter 4, the relationships of a portable (P)-SMPS with the CPC and SAM instruments, which are measurement devices with different metrics, were investigated, and two widely used SMPSs were compared to harmonize the measurement protocols. Nanoparticles were measured by several sampling devices. It is necessary to understand the relationship among these measurements because devices produced by different manufacturers are based on different techniques and principles. For LAB and ENP, there was a good correlation among the (P)-SMPS, CPC, and SAM. However, the correlation among (P)-SMPS, CPC, and SAM was only fair in UNP workplaces. This was partly explained by the fact that the particles were not spherical, although the calibration of the sampling instruments was performed using spherical particle

additionally, the concentration of UNP in workplaces was very high, allowing aggregates to form easily. A chain-like particle morphology was identified using the scanning electron microscope (SEM) for workplaces with UNP. The CPC or SAM instrument could be used as an alternative to an SMPS in workplaces handling ENP. In workplaces producing UNP where the concentration is high, real-time instruments should be used with caution. Electron microscope images can be used to confirm the morphology of nanoparticles. There were significant differences between the two SMPSs. A TSI SMPS indicated a concentration about 20% higher than that produced by a Grimm SMPS in all workplaces. Several factors are responsible for these differences between the two SMPSs: 1) instabilities in the aerosol, 2) the scanning sequences, and 3) the sampling time interval. Therefore, caution is required when comparing data from different SMPSs. In summary, the applying of different control strategies is required because the concentration profiles and characteristics were different among LAB, ENP and UNP workplaces. The nanoparticle concentrations varied depending on the tasks performed and working status (working or off-duty). Task-based exposure assessments could provide useful information regarding nanoparticle exposure profiles and could be used as nanoparticle exposure assessment tools in place of traditional full-shift measurements. For real-time monitored data, the ARIMA model is the most suitable for the analysis because it accounts for autocorrelation and the stationary nature of the data. Using the ARIMA model, averaging times of 1, 5, and 10 minutes gave almost identical results. Flexible time averaging is suggested because it can be used with a research-grade sampling device, which in this study was a full-size SMPS that had a sampling interval longer than one minute. The CPC or SAM with electron microscope imaging are suitable alternative instruments for nanoparticle exposure assessment, rather than the SMPS, which is a relatively expensive device. Caution is required when interpreting the results and comparing the exposure monitoring field data using different SMPSs.