Summary

Every year, Quebec workers suffer the health effects of exposure to chemicals, and particularly solvents. The modelling of occupational exposure plays a major role in risk prevention. The goal of this research project was to study various aspects of modelling for the purpose of estimating occupational exposure to solvent vapours and improving prediction accuracy. The study was carried out in two stages. It involved both small-scale (Stage I) and human-scale (Stage II) testing that focused on determining emission rates for pure solvents and mixtures, as well as on investigating the behaviour of solvent vapours released into the air and subject to different experimental ventilation conditions.

In Stage I, exponentially decreasing emission rates (α) were determined experimentally in a controlled environment (temperature, humidity, air speed) for five pure solvents under different conditions. Multiple linear regression analyses were conducted to assess the influence of the various tested parameters on α values. Emission rates were also calculated for pure solvents and mixtures, while vapour concentrations in a 0.085 m³ box were estimated using the uniformly mixed room model with exponentially decreasing emissions. All in all, 18 scenarios were carried out with different solvents: 4 pure solvents, 12 aqueous mixtures (10%, 5% and 1% solvent in water) and 2 organic solvent mixtures. The concentration estimates assumed both ideality (use of non-corrected emission rates) and non-ideality (use of emission rates corrected by activity coefficients). These estimated figures were compared with concentrations measured using a gas phase chromatography system coupled with a thermal conductivity detector (TCD). The concentrations were compared graphically with the values predicted by the model, and ratios of the maximum measured to estimated concentrations were calculated.

In Stage II, 19 different experimental situations, created in a human-scale room with a volume of 53.4 m³, were tested three times. The room was ventilated by means of two ventilation strategies (floor/ceiling) and at different airflow rates (low rate [L] at 0.8 air changes per hour [ACH] [12 l/s], high rate [H] at 2.3 ACH [32 l/s] and very high rate (VH) at 4.5 ACH [64 l/s]). Four scenarios were tested: 1) evaporation on a table, 2) spill on the floor, 3) application of solvent to a rag followed by manual cleaning and 4) spraying of solvent followed by manual cleaning. The evaporation and spill tests were conducted using a watch glass containing 20 mL of acetone placed on an analytical balance. The cleaning tests were conducted by an operator who simulated cleaning a piece of aluminum with a rag. The solvent vapour concentrations were measured using direct-reading instruments placed in the near field (NF, 30 cm from source) and far field (FF, rest of the room). For the near field, two photoionization detectors (PIDs) were used to measure the concentrations. For the far field, two Varian Micro GC chromatographs coupled to TCDs were used. For each test, concentrations in the near and far fields were modelled using the two-zone model, and these concentrations were compared with the measured values. The near field radius values were optimized so that the measured and estimated concentrations corresponded. Statistical analyses were conducted to determine whether significant differences existed between the near and far concentrations. Variance and multiple linear regression analyses were also performed to assess the influence of different variables in the models. Computational fluid dynamics (CFD) modelling of air movements and gas contaminant dispersion was carried out for some scenarios. Solvent evaporation was modelled in the code as a boundary condition at the surface of the watch glass whereby a predetermined mass of solvent was injected into the ambient air. Comparisons between estimated and measured values were therefore performed.

For Stage I, the variations observed in coefficient α values were primarily due to the variables vapour pressure, surface/volume ratio and air speed above the spill. Estimates of concentrations in the box that considered non-ideality in the case of mixtures, i.e., corrected estimates, were higher than the non-corrected estimates and closer to the measured values. In addition, the times required to reach the concentration peaks of the corrected estimates provided a means of adequately estimating emission kinetics.

For Stage II, the analysis of variance showed that all of the variables had an effect on the near field concentrations, whereas only airflow rate and air intake position had an effect on the far field. The increase in ventilation rates caused a significant drop in concentrations in both fields. The radiuses obtained through optimization of the measured and estimated concentrations for the evaporation and spill scenarios were very homogeneous, with a mean radius of 0.72 m (geometric standard deviation, GSD, of 1.3) and the corresponding mean estimated coefficient β (interfield airflow rate) was 3.9 m³/min (0.92–16.9). Using this radius for near field geometry allows adequate estimation of the concentration of solvent vapours at a distance of 30 cm from the source. In contrast, for the rag application and spray scenarios, the optimized radiuses were larger and varied more broadly, with respective mean radiuses of 1.1 m (GSD of 1.6) and 1.2 m (GSD of 1.9). The CFD modelling provided a means of studying the concentration gradient around the source for the evaporation and spill scenarios. The concentration gradient diminished rapidly, with concentrations dropping from 1,757 mg/m³ to 83 mg/m³ for a cube-shaped near field of 14 cm a side and 64 cm a side, respectively. The CFD modelling also highlighted contaminant displacements due to vapour density relative to the air, without the need to involve advection. This transport mechanism is especially significant when the air delivery rate in the room is low.

This study has demonstrated the importance of the different variables used to estimate emission rates in the event of small spills, the importance of considering non-ideality in cases of the use of non-ideal mixtures and the importance of various concentration determinants in near and far fields. These data improve our general understanding of solvent vapour dispersion and the models used in occupational health and safety to estimate worker exposure to such emissions.