Data-Driven Models: From Data to Knowledge for Extreme Events

In the area of dynamic load effects, advanced analysis, modeling, and simulation tools are becoming increasingly important in order to address the non-stationarity, non-Gaussianity, and nonlinearity inherent in hazard related events and their associated structural response. Previous assumptions of stationarity, Gaussianity, and linearity, while convenient, have proven to over-simplify the complexity of the problems associated with dynamic effects associated with wind, wave, and earthquake related events. In order to address these issues, this research introduces a suite of tools that can be used to analyze, model, and simulate complex environmental hazard-induced loads and resulting structural response. Univariate and multivariate processes are both considered. Specific areas of focus include:

• Develop improved methods for the detection of key features in datasets, such as non-stationarity, using the method of surrogates;
• Assess current analysis procedures utilized for both stationary and non-stationary wind datasets and introduce more flexible averaging interval schemes for the proper determination of turbulence characteristics in non-stationary wind;
• Develop an efficient multivariate non-Gaussian simulation approach that considers non-Gaussianity of the joint probability density function in addition to considering non-Gaussianity of the marginal distributions;
• Improve upon existing non-stationary simulation approaches by also allowing for consideration of non-Gaussianity; and
• Develop a more flexible and robust approach for the detection and analysis of nonlinear coupling between two processes related to hazard loading/response or within the same process between two different modes.

These tools allow for the consideration of extreme events, which is often not addressed due to lack of data and complexity of interactions. The assurance of the safety and reliability of structures subjected to environmental loads, however, requires the consideration of extreme loading scenarios and response.

Besides consideration of non-stationarity, non-Gaussianity, and nonlinearity in events, it is also important to consider the multiple hazards with which a structure may be faced over its lifetime. Design and construction in a coastal environment would benefit from the consideration of the effects of concurrent hazards as well as hazards in isolation. Multi-hazard engineering strategies can be incorporated into design to enhance the inherent resilience and improve the robustness of structures. A performance-based engineering (PBE) approach would also allow for a structure to be designed for specified levels of performance under particular hazard scenarios. A possible design methodology that integrates multi-hazard engineering with PBE is introduced and seeks to improve structural reliability for a wide range of loading scenarios.

The success of the proposed research tools will provide engineers with a better understanding of the loads to which a structure will be subjected, improving the ability to more accurately predict structural response in the design stages. This knowledge will help designers to determine the most efficient and effective designs for structures to best withstand a range of natural hazards.