Data driven residual stress simulation for additive manufacturing - ENGIN-X

Residual stress (RS) remains a critical challenge in metal additive manufacturing (AM), potentially undermining component performance and reliability. Neutron diffraction offers a non-destructive alternative that probes deep within components and captures all three stress components. Leveraging this capability, the proposed work integrates neutron diffraction measurements, surrogate finite element analysis (FEA), and machine learning (ML) to refine and validate RS predictions in laser powder bed fusion (LPBF) steel specimens.

Two distinct sample geometries will be examined: one for model refinement (via extensive RS mapping) and another for validation. The neutron diffraction data will serve as ground-truth inputs to improve the ML-driven FEA, ensuring enhanced predictive accuracy. The outcomes of this research not only demonstrate the efficacy of neutron diffraction in facilitating robust RS assessment and model validation but also promote greater confidence in AM processes for industrial adoption.

残留应力（Residual Stress, RS）仍是金属增材制造（Metal Additive Manufacturing, AM）领域的核心挑战之一，其可能会对构件的服役性能与可靠性造成不利影响。中子衍射作为一种无损检测技术，可深入构件内部开展表征，并可同步获取全三向应力分量。本研究依托该技术特性，整合中子衍射测量、代理有限元分析（Finite Element Analysis, FEA）与机器学习（Machine Learning, ML）手段，对激光粉末床熔融（Laser Powder Bed Fusion, LPBF）钢制试样的残留应力预测结果开展优化与验证工作。

本次研究将采用两类差异化的试样几何构型：一类用于模型优化（通过大范围残留应力测绘），另一类用于模型验证。中子衍射数据将作为基准真值输入，用于优化机器学习驱动的有限元分析模型，从而显著提升预测精度。本研究成果不仅验证了中子衍射在实现精准可靠的残留应力评估与模型验证方面的应用效能，同时也为工业界推广应用增材制造工艺提供了更坚实的信心支撑。