Mark G. Vangel
Statistical Engineering Division, ITL
Walter J. Rossiter
Building Materials Division, BFRL
Lead paint is known to be potentially dangerous to health, particularly for children; consequently it cannot be used for new structures in this country. However, much lead paint still remains in older housing. Since properly removing and disposing of lead paint is an expensive process, it is natural to consider the feasibility of covering lead-painted surfaces with a coating designed to prevent the lead from escaping into the environment. NIST is in the process of evaluating various lead encapsulants which are commercially available. In particular, the experiment reported on here was designed to evaluate the impact resistance of these coatings, and to compare them with ordinary paints.
Steel, plywood, and drywall panels were painted first with a pink undercoat, and then with one of 12 white coatings: coatings 1-6 are unreinforced lead encapsulants, coatings 7-10 are reinforced encapsulants, coating 11 is a latex paint, and coating 12 is an alkyd paint. Two replicate panels were made for each substrate/coating combination, and 100 squares were ruled on each specimen. According to a statistically-designed experiment, randomly-chosen squares on the various panels were impacted with weights of various impact energies (dropped from various heights, under laboratory conditions). Because of the pink undercoat, cracked coatings were usually obvious, and they were regarded as failures. The impact energies were varied so as to attempt to obtain both failures and non-failures for each panel; however this was not always possible. The dataset, not including controls, consists of binary outcomes from 2970 impacts.
A logistic regression model was fit to these
data, and this model assumes that the probability
of failure is linearly related to impact energy,
with an intercept which depends on coating,
substrate, and their interaction. One summary
of the results of this analysis, shown in the
figure, consists of the estimated energy corresponding
to 50% probability of penetration, as a function
of coating and substrate. The intervals in the
figure are approximate 95% confidence intervals.
The very wide confidence intervals usually correspond
to substrate/coating combinations for which either
all impacts were failures, or else for which none
were failures. Perhaps the most obvious conclusion
to be reached from this summary is that reinforced
coatings are more impact resistant than paint, but it
is not clear whether the same can be said for
unreinforced encapsulants. Also, it appears that
steel, perhaps because it is a hard substrate,
provides data which are more sensitive to differences
among the coatings than the other substrates.
Figure 19: Impact energy corresponding to 50% probability of failure for various coatings and substrates. The plot symbols indicate estimated median failure energies, the smallest (F) energy at which a failure occurred, and the largest (S) energy at which a failure did not occur.
Date created: 7/20/2001