Mechanical behaviors and internal pressure bearing capacity of nuclear containment using UHPC and ECC: From numerical simulation, machine learning prediction to fragility analysis

Abstract

This paper examines the use of Ultra-High Performance Concrete (UHPC) and Engineered Cementitious Composites (ECC) for containment structures subjected to internal pressure. It presents finite element models for these structures using UHPC, ECC, and conventional C60 concrete, validated through existing experimental data. The research compares the failure modes and mechanical behaviors of structures built from these materials. Findings show that UHPC and ECC have similar capacities for bearing internal pressure and demonstrate reduced deformation. Both materials notably diminish deformation, curtail cracking, and boost bearing capacity by approximately 30.6% compared to C60. To further analyze these properties, thirty parameter sets for C60 were created using the Latin hypercube sampling method and incorporated into the validated models to evaluate internal pressure capacity. Additionally, the study employed three machine learning algorithms (Bayesian networks, decision trees, and random forests) to predict bearing capacities effectively. With an additional thirty parameter sets, the random forest method emerged as the most precise. Parameter sets for UHPC and ECC were similarly generated and used to develop prediction models for the internal pressure capacities. In a broader scope, one hundred parameter sets across all three materials were analyzed using the random forest method. The maximum likelihood method assessed the fragility of these containment structures, providing statistical forecasts of the capacities to withstand internal pressures. This comprehensive analysis supports the application of UHPC and ECC in practical engineering contexts for containment structures.