The stress concentration and damage evolution of ultra-high performance concrete (UHPC) under long-term dynamic loading are difficult to monitor in real time, as conventional sensors suffer from poor durability, high cost, and incompatibility with matrix deformation. Recently, a team from the University of Shanghai for Science and Technology developed a machine learning framework that significantly improves dynamic compressive stress prediction in high-sensitivity ultra-high performance concrete (HS-UHPC) by incorporating electrical resistivity as a key input parameter. Using three machine learning algorithms—double-layer neural network, boosting tree, and squared exponential Gaussian process regression (SE-GPR)—the team demonstrated that adding resistivity measurements alongside traditional displacement data enhances predictive accuracy, with the SE-GPR model achieving an R² of 0.85 and reducing mean absolute error by 41.1% compared to displacement-only models. The core innovation is using electrical resistivity to directly capture load‑induced microstructural changes, overcoming the damage‑detection limitations of traditional strain or displacement measurements. This provides a new theoretical basis for intelligent monitoring of self‑sensing concrete.

A research team investigated how high-volume fly ash replacement affects concrete’s early-age characteristics and hardening properties. Tests with 0%–60% fly ash replacing cement show that fly ash improves flowability but delays early hydration. While early strength drops with higher fly ash, 10%–40% replacement delivers superior long-term mechanical performance. The team recommends 40% fly ash as the optimal cement replacement, balancing environmental sustainability, cost efficiency, and structural performance for green concrete engineering.