N047
Prediction of Filter Performances against Organic Vapours:
Statistical Analysis of Breakthrough Times
as a Basis for a Simple
Prediction Model
Andre Lavanchy
Individual NBC Protection
SPIEZ LABORATORY
Spiez,
Tel +41-33-228-17-17 Fax
+41-33-228-14-02
e-mail andre.lavanchy@babs.admin.ch
ABSTRACT
Over 75 breakthrough times of organic vapours on a
typical respirator filter have been measured under EN 141 similar conditions. The
paper describes the analysis of the results in an attempt to develop a simple, but
reliable, prediction model for breakthrough times. The 46 organic compounds involved, including alkanes, cyclo-alkanes,
alkenes, alcohols, ketones and carbonic acids have been clustered based on their
structure and functional groups, leading to a matrix
like structured dataset. Afterwards,
this structural information has been coded and complemented with two physical
properties; the saturation pressure and the affinity coefficient. This final dataset has been analysed by means
of a general linear regression model. We found out,
that a simple linear model with only five regressors yields to satisfactory breakthrough
times. Furthermore, the calculated breakthrough times
of five control substances which were initially retrieved from the data set lie
within a ten percent deviation of the experimental values. Therefore, this
study shows that for a given filter type and well defined testing conditions,
satisfactory predictions of breakthrough times of organic vapours are feasible.