CAREER: Advancing the Development of Realistic and Probabilistic Shear Wave Velocity Ground Profiles Using Advanced Inversion Strategies

University Of Arkansas, Fayetteville AR

Investigators

Abstract

This Faculty Early Career Development Program (CAREER) grant will advance our ability to image below the ground surface by utilizing surface wave methods to develop more realistic and probabilistic shear wave velocity (Vs) profiles. As advanced as the tools of daily society have become, the world of in-situ site characterization still remains mired in the past and continues to rely heavily on empirical approaches, which were developed over 100 years ago, while the medical industry has made leaps forward in the field of non-invasive imaging. As the profession moves forward, the advancement of non-invasive methods is critical to meeting the challenges of tomorrow in a cost-effective manner. As a step toward this goal, this project plans to advance our ability to develop realistic and probabilistic subsurface models through advanced inversion schemes. These schemes will harness artificial intelligence and additional wavefield information to replace a level of user skill now required to develop these subsurface models. These realistic subsurface models are critical to utilizing parameters, such as shear wave velocity, in applications including liquefaction triggering, site response analysis, bedrock rippability, and settlement analyses. In addition, the educational impacts of the project center on promoting the use of non-invasive methods by (1) inspiring future engineers to embrace new technologies through engineering summer outreach programs, (2) educating students through an international student exchange program, and (3) providing training to practicing engineers through a speakers bureau. The intellectual merit of this research lies in the development of state-of-the-art surface wave inversion algorithms. These algorithms will incorporate a Bayesian statistical framework into high-level inversion algorithms using machine learning and trans-dimensional Monte Carlo methodologies. The algorithms will incorporate expert knowledge into the inverse problem and characterize the uncertainty of the developed Vs profiles based on the experimental data. The use of Bayesian and machine learning methods will allow uncertainty in the solution to be considered and presented in a more robust way than current approaches. In addition, further understanding of the petrophysical link between multiple data types advances our knowledge of how different data types work together within joint inversion frameworks to constrain the inversion problem. Advances in the inversion framework will produce broader impacts for multiple applications including site response, liquefaction analysis, and infrastructure evaluation. Moreover, the development of more accurate, realistic, and probabilistic Vs profiles allows for the inclusion of resulting Vs profiles into performance-based designs. Lastly, advancements in inversion algorithms and knowledge of petrophysical links are transferable to other non-invasive geophysical methods, which all suffer from non-uniqueness issues. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.