A neural network based spatial light scattering instrument for hazardous airborne fiber detection

Abstract

A laser light scattering instrument has been designed to facilitate the real-time detection of potentially hazardous respirable fibers, such as asbestos, within an ambient environment. The instrument captures data relating to the spatial distribution of light scattered by individual particles in flow using a dedicated multi-element photodiode detector array. These data are subsequently processed using an artificial neural network which has previously been trained to recognise those features or patterns within the light scattering distribution which may be characteristic of the specific particle types being sought, such as for example, crocidolite or chrysotile asbestos fibers. Each particle is thus classified into one of a limited set of classes based upon its light scattering properties, and from the accumulated data a particle concentration figure for each class may be produced and updated at regular intervals. Particle analysis rates in excess of 103 per second within a sample volume flow-rate of 1 litre per minute are achievable, offering the possibility of detecting fiber concentrations at the recommended maximum exposure limit of 0.1 fibers/ml within a sampling period of a few seconds.