A comparison of methods to estimate peak wind loads on buildings

• A review of the observed peak and translation methods for peak pressure coefficient estimation is presented.
• Two new translation-based methods for estimating the peak distribution of pressure coefficients are presented.
• 30 h (full scale) wind tunnel dataset from two low-rise building models provide a benchmark to quantify the precision and accuracy of reviewed and proposed methods.
• A comparative study of the accuracy and precision of the reviewed and proposed methods is conducted.

The most appropriate statistical technique to estimate a peak pressure coefficient from wind tunnel data is not a settled issue. The lack of a standard acceptable method can lead to inconsistent definitions and interpretations of peak pressure coefficients, particularly since time constraints associated with wind tunnel tests necessitate relatively short test durations. A Gumbel model is commonly used to represent the peak distribution, where parameters are determined using observed peaks. Recent papers have proposed several variations of a peak estimation procedure using the entire time history and a translation from a Gaussian peak distribution model to non-Gaussian. It is shown that, in the case of mildly non-Gaussian data, translation methods achieve accuracy comparable to the Gumbel method. It is also shown that translation methods lose accuracy when the record deviates significantly from Gaussian, while the Gumbel model maintains stable accuracy and precision. This paper presents two new translation-based peak pressure coefficient estimation schemes that offer accurate and stable performance for strongly non-Gaussian data. Very long duration wind tunnel data provide empirical peak distributions with which to compare the relative performance of the Gumbel, existing translation and proposed new translation methods. One of the new methods slightly outperforms the Gumbel method.