Emerging frontiers in wind engineering: Computing, stochastics, machine learning and beyond

Abstract Over the last several decades, wind engineering a multi-disciplinary subject involving engineering meteorology, fluid dynamics, structural dynamics, structural engineering, probabilistic methods, and design has addressed the challenges posed by winds of synoptic and non-synoptic origins. Combined computational approaches and laboratory to full-scale experiments have enhanced our ability to design and construct wind-resistant structures that range from low-rise to supertall buildings, footbridges to super long-span bridges and, wind turbines on the ground and floating foundations and floating offshore drilling and production systems. During this period, we have seen extraordinary advances in experimental facilities, instrumentation and data acquisition and management. At the laboratory scale, new wind tunnels have emerged with added features like extra-wide cross-sections, from passive to active driving systems, from boundary layer to flow simulators with vortical flows mimicking non-synoptic winds features. While at full-scale, we have been able to use deployable sensing networks in the path of landfalling hurricanes/typhoons to monitoring in real-time the performance of tall buildings and long-span bridges during extreme wind events. Advanced technologies like aerial surveying using drones and satellite imagery have been employed to enhance the post-storm surveillance capabilities. These advances have enabled us to build a cadre of civil infrastructure that meets some of the challenges posed by the extreme winds. Yet there remain several frontiers that still need to be addressed for example the three “Nons,” the triple emerging fronts, i.e., non-stationarity, non-Gaussianity, non-linearity prevalent in the changing dynamic of winds prevailing in gust fronts, vortical and convective systems, rolls, meso-scale features and intermittent turbulence. In the face of these challenges, increasing heights, spans, and depths of structures exposed to these winds pose additional challenges as their performance becomes more sensitive to their dynamics, thus necessitating new tools and perspectives that go beyond customary analysis and modeling norms. Fortunately, amidst these challenges, there are new opportunities to complement our existing capabilities as the burgeoning growth in computational resources and parallel computational advances coupled with data analytics and AI-based schemes, e.g., machine learning hold the promise of expanding our modeling and simulation capacity far beyond our current conventional schemes offer. All these advances can be couched in a Generalized Wind Loading Chain to capture the three “Nons” by building upon the wind loading chain proposed by Davenport based on linear and stationary conditions. An example of the Gust Front Factor in this framework is an effective means for designing under non-synoptic winds. This paper expands on these new computational opportunities and ways to take advantage of their added capabilities to address emerging challenges in building a resilient and sustainable civil infrastructure and beyond to stability and safety of high-speed trains.