Probability Models for Live-Load Survey Results

Structural floor load data from five major live-load surveys are analyzed probabilistically in order to determine the accuracy of fitted normal, lognormal, and gamma distributions. Histograms and scaled cumulative graphs are made for every use type from each survey and statistical goodness-of-fit tests made on the assumed distributions. Data from the surveys are combined by occupancy type to compute load levels with high probabilities of nonexceedance, and comparisons are made among conservative levels of observed loads, present design values, and the computed loads. Results indicate that the gamma distribution provides a realistic approximation to the observed data, and in particular yields excellent prediction of observed office load data up to the 99.9% cumulative level. Present design loads for offices, domestic dwellings, and mercantile occupancies appear to approximate the 99.99% cumulative level.