Work

Linkage from Mold Flow Simulation to Crash Simulation, Part One: Prediction of Mechanical Properties for Die Castings Based on Physical Feature Enhancement and Random Forest

Year: 2026
Type: conference-paper
Abstract: Traditional evaluation of mechanical properties in die-cast components relies heavily on exhaustive physical testing. However, the structural complexity of integrated castings significantly complicates sampling design, while discrete experimental data often fail to capture global performance gradients. These limitations lead to a critical technical bottleneck: “spatial characterization blind spots — prediction inaccuracies — engineering failure risks.” To bridge this gap, this study proposes a predictive framework for the spatial distribution of mechanical properties by integrating manufacturing process parameters, material constitutive models, and mechanical performance. Leveraging physical fields from mold-flow simulations and geometric feature distributions, a strategic mechanical test matrix was developed and experimentally validated. Material constitutive modeling and fracture parameter identification were performed using LS-DYNA to establish reference material cards. Through systematic small-scale butterfly-notch tensile tests and simulation-based inverse calibration, constitutive and fracture parameters across diverse locations were derived. Finally, an optimized machine learning model was implemented to predict the full-field distribution of mechanical properties, achieving multi-dimensional correlations between material, process, and structural performance.
Language: English
Open Access status: hybrid
OpenAlex ID: W7204207481
FWCI:
Cites: 1
Cited by:
Related to: 10