28/05/2026
A study just published in Buildings (MDPI, May 12, 2026) presents a faster way to assess how a building would collapse if fire, wind, and an earthquake hit in sequence. The authors built a physics constrained neural network that generates collapse fragility curves for these cascading scenarios without running thousands of hours of non-linear analysis.
The clever piece is a thermodynamic rule embedded in the model itself: it cannot predict that a fire-damaged structure becomes stronger. That guardrail is what separates this work from typical machine learning shortcuts in structural reliability.
For practicing engineers working on hospitals, schools, and tall residential towers, this is the direction the profession is moving. The 2026 NEHRP Provisions and ASCE 7-22 already signal that multi-hazard thinking is expected for higher Risk Categories.
Should multi-hazard fragility checks become a standard part of design for Risk Category III and IV buildings, or stay as a specialty study?