Alignment of Wind Reliability Analysis with Code-Based Equivalent Static Wind Load Methods

Abstract

This project aimed to first identify the underlying reasons why wind demands estimated through wind time history analysis have been seen on occasion to exceed demand envelopes estimated from equivalent static wind loads (ESWLs) and secondly to uncover the reasons why initial reliability studies have suggested that there is a discrepancy between the reliabilities obtained from systems designed to comply with current load and resistance factor design (LRFD) procedures for steel and reinforced concrete main wind force resisting systems (MWFRS) and the target reliabilities indicated in Table 1.3-1 of ASCE 7-22.

In the first phase of the project, record-to-record variability was investigated as a possible cause for wind demands estimated from time history analysis exceeding those estimated from ESWLs. Record-to-record variability is a fundamental aspect of any system exposed to stochastic excitation, including structural systems subject to dynamic wind loads. In the past, wind engineering methodologies have typically disregarded this variability by focusing on the expected peak responses, i.e., the average peak response obtained from analyzing the MWFRS for multiple realizations of the dynamic wind loads for each relevant wind speed and direction. Various historical factors have contributed to this state of affairs, including the efficient estimation of expected peaks in the frequency domain for linear elastic systems. However, the growing interest in wind analysis schemes that employ time history analysis, for example, performance-based wind design, can lead to important inconsistencies. Indeed, wind time history analysis is often carried out (for a multitude of reasons) using a single wind record for each critical wind direction. Findings from this phase of the project demonstrate that record-to-record variability can significantly alter the peak responses of the MWFRS. The observed coefficient of variation on peak responses, particularly the demand-to-capacity ratios of critical components, generally falls between 0.1 and 0.3. A formal sensitivity analysis conducted using the Sobol method revealed that record-to-record variability can represent an important source of system uncertainty. This variability in peak responses is accounted for in traditional wind design approaches, which are based on ESWLs, through the use of expected peak values. As would be expected, an analysis of archetype structures revealed that variability around the expected peak can lead to wind responses surpassing the wind demand envelopes estimated from ESWLs. If practical, it is therefore proposed that a suite of wind records be considered for each wind speed and direction involved in time history analysis. This would allow for the direct estimation of expected peak responses and the variability around this value due to record-to-record variability. If running multiple wind records is computationally prohibitive, using a single record that is appropriately scaled to produce the expected peak of a critical response parameter is recommended. Such considerations are particularly important when conducting nonlinear time history analysis due to its path-dependent nature.

During the project’s second phase, the causes behind the apparent discrepancies in the reliability of the MWFRS, designed to comply with current LRFD requirements using ESWL derived from building-specific wind tunnel tests and calibrated to the wind intensities suggested in ASCE 7, were examined. To this end, the reliability of two Risk Category II archetype MWFRS was investigated with the “true” reliabilities of these archetypes estimated through the application of the wind reliability modeling environment, WiRA. To ensure the WiRA models were compliant with the finite element models of the designers of the archetypes, a rigorous QA/QC was carried out. This led to baseline component reliability estimates that appeared to be deficient compared to the target reliabilities suggested in ASCE 7-22 for code compliant building systems. The influence of modeling choices related to the wind hazard, such as the choice of the wind hazard curve, on the results was examined and was not found to be the root cause of the deficit in reliability. Consequently, the theory supporting the wind reliability estimates used in calibrating LRFD was revisited. It was shown that if wind loads are calibrated using wind effects with mean recurrence intervals (MRIs) that are consistent with those suggested in ASCE 7 for Risk Category II structures, the target reliabilities of Table 1.3-1 will not, in general, be achieved. Consequently, it could be argued that the MRIs for the design winds speeds of ASCE 7 should be multiplied by a coefficient of 1.095 before use in design. As noted in the literature, for Risk Category II buildings in extratropical regions and limit states where “Failure that is not sudden or does not lead to widespread progression of damage”, the lack of such a factor can lead to an apparent reliability reduction from 3.0 to around 2.5. Currently, these nuances are lacking in the presentation of Table 1.3-1 of ASCE 7.

In summary, record-to-record variability, i.e., the natural variability in the time history wind load traces given an identical average wind speed and wind direction, was seen to be a contributing factor why wind time history analysis can exceed demand envelopes estimated from ESWLs while the apparent discrepancy between the reliability of structural systems designed using advanced wind analysis procedures and the targets suggested in Table 1.3-1 of ASCE 7 was revealed to be contributed by the loss of an implicit factor that can be interpreted as unduly reducing the MRIs of the design wind speeds (or load effects used to define loading scenarios for use in LRFD or performance-based wind design).