AI surrogate model ForgeNet for cold-forging simulation presented at international conference

On July 22, 2026, Godelblock Inc. presented research findings on ForgeNet, a deep-learning surrogate model for cold-forging simulation, at WCCM ECCOMAS Munich 2026, an international conference held in Munich, Germany. According to the company’s press release, ForgeNet addresses the challenge of maintaining node correspondence during adaptive remeshing by projecting finite element analysis results onto a fixed Eulerian grid. It enables fast predictions using two models, one for node displacement and the other for stress.

For engineers using cold-forging analysis in product design and die studies, a key challenge is that analyses that rebuild elements as a part deforms can disrupt node correspondence, making the data difficult to use directly for training. Godelblock says its research enables fast surrogate predictions under these conditions by mapping results onto a nondeforming reference grid and predicting displacement and stress separately.