Scaling in size, time and risk—The problem of huge extrapolations and remedy by asymptotic matching

The scaling of structural response of concrete structures to large structure sizes, to long service lives and to tolerable failure probabilities is a problem of order-of-magnitude extrapolations, which are intractable by AI and machine learning and require significant theoretical advances. The present review, based on a lecture at Yonggang Huang's 60th birthday symposium in Houston, summarizes the recent advances, with a focus on those achieved at Northwestern University. Reliable extrapolation requires two-sided asymptotic matching. Most existing databases provide support on only one extreme of the range of size, time or failure probability, but theoretical support can be obtained for the asymptotic behaviors on both sides of the range. The advantage is that the asymptotics are much simpler than the behavior in the central, transitional, range. In closing it is explained that realistic extrapolations are required to mitigate the calamitous CO₂ emissions from cement and concrete industry.