Validation of a laboratory spray generation system and its use in a comparative study of hexamethylene diisocyanate (HDI) evaluation methods.

Isocyanates are well-known irritants and sensitizers, and measuring their occupational airborne exposure is challenging due to their high chemical reactivity and semi-volatile nature. This study builds on a previous publication by our team that focused on comparing evaluation methods for isocyanates. The current research aims at developing, validating, and applying a laboratory generation system designed to replicate real-world conditions for spraying clear coats in autobody shops using hexamethylene diisocyanate (HDI)-based products. The system involved a spray gun connected to two chambers in series, enabling sample collection and analysis. The system successfully generated HDI and isocyanurate concentrations ranging from 0.008 to 0.040 mg m and 0.351 to 3.45 mg m, respectively, with spatial homogeneity (RSD) of 5.8% and 16.5%. The particle-size distribution (MMAD) of 4 μm was measured using a cascade impactor and an electrical low-pressure impactor. The samples generated were used to correlate the amount of isocyanates collected with scanning electron microscope images of droplets on a filter. Three methods were compared to the reference method-an impinger with a backup glass fibre filter (GFF) and 1,2-methoxyphenylpiperazine (MP) based on ISO 16702/MDHS 25-in six generation experiments: (1) Swinnex cassette 13 mm GFF MP (MP-Swin); (2) closed-face cassette 37 mm GFF (end filter and inner walls) MP (MP-37); and (3) denuder and GFF dibutylamine (DBA) (ISO 17334-1 Asset). The analysis revealed clear trends regarding which sampler sections collected HDI (mainly in the vapor phase) or isocyanurate (exclusively in the particulate phase). The study found no significant bias between the tested methods (MP-Swin, MP-37, and Asset) and the reference method (impinger) for both HDI monomer and isocyanurate. The three tested methods showed limits of agreement beyond the acceptable range of ±30% (95% confidence interval), largely due to data variability, though MP-Swin and MP-37 exhibited lower variability than Asset. The results will be further evaluated in a real-world environment where similar clear coats are used.