Inversion of the fracture toughness of zirconium alloy cladding interface in nuclear fuel using splitting method via general regression neural network

Abstract

For nuclear fuel elements, the interface mechanical properties of zirconium alloy cladding is critical to the safety and reliability of reactors. However, due to the small thickness of the fuel plates (<2 mm), accurately capturing the behavior of interface cracks is challenging, complicates the measurement of interface fracture toughness. This study developed a data-driven inversion method using the generalized regression neural network (GRNN) to rapidly and accurately determine the fracture toughness of zirconium alloy cladding interface. The database was established by combining splitting experiments with numerical simulations. The cohesive zone model was utilized to accurately simulate crack propagation paths and fracture modes in numerical simulations. The influences of key parameters such as cohesive strength, stiffness, and interface fracture energy were analyzed in detail. After extensive training, the prediction model accurately forecasted the interface fracture toughness. The results indicate that the proposed GRNN-based inversion approach is feasible and effective for predicting the fracture toughness of zirconium alloy cladding interface, and can be extended to determinations of other mechanical properties in the nuclear fuel element.