An artificial intelligence based approach to account for the effects of microstructure gradients and residual stresses on fatigue performance of additively manufactured aluminum

该联盟项目致力于开发一种新的数值框架，将基于晶体塑性的有限元方法(CPFEM)和基于人工智能(AI)的机器学习(ML)方法相结合，以研究微结构梯度和残余应力对添加制造(AM)部件疲劳性能的影响。AM制造零件中的一个主要问题是在激光粉床熔化(LPBF)加工的零件中经常产生的过大和各向异性残余应力。这些材料的机械性能，如疲劳寿命，很大程度上取决于残余应力，而残余应力往往与材料的屈服强度一样高。此外，这些应力是由于局部微观结构效应而形成的，例如颗粒形态、织构和滑移行为的局部差异，并且由于固有的复杂和异质的微观结构，这些应力是添加制造的合金所关注的原因。由于连续有限元可以综合宏观和微观层面的残余应力、微观组织属性(即晶粒形态、微结构梯度、织构)，并且可以在滑移系统水平模拟诱发塑性变形的累积，因此它是目前微结构敏感应用的理想候选者。然而，由于疲劳加载的重复性和耗时的性质，CPFE建模的计算量很大，并且应用于疲劳寿命预测使问题进一步复杂化。该项目结合了晶体塑性和人工智能的最新进展，使研究微观结构梯度和残余应力对添加制造的铝合金疲劳性能的影响成为可能。新的数值框架将使用人工智能算法显著加快CPFE模拟，从而允许对疲劳进行数值模拟，这种模拟可以解释通过添加制造获得的微结构的全部复杂性。

This Alliance project focuses on developing a new numerical framework to couple crystal plasticity-based finite element method (CPFEM) with artificial intelligence (AI) based machine learning (ML) approaches to investigate the effects of microstructural gradients and residual stresses on the fatigue performance of additively manufactured (AM) components. One major concern in AM build parts is the excessive and anisotropic residual stresses that frequently develop in laser powder bed fusion (LPBF) processed components. The mechanical properties such as the fatigue life of these materials are strongly governed by residual stresses that can often be as high as the material's yield strength. Furthermore, these stresses form due to local microstructural effects, such as local differences in grain morphologies, texture, and slip behavior, and are a cause of concern for additively manufactured alloys due to the inherent complex and heterogeneous microstructures. Since CPFEM can incorporate macro and micro-level residual stresses, microstructural attributes (i.e., grain morphologies, microstructural gradient, texture), and can model the accumulation of the induced plastic deformation at the slip system level, it is an ideal candidate for the present microstructure sensitive application. However, CPFEM modeling is computationally expensive, and its application to fatigue life predictions further complicates the problem due to the repetitive and time-consuming nature of fatigue loading. This project combines the latest advancements in crystal plasticity and artificial intelligence to enable the investigation of the effects of microstructural gradients and residual stresses on the fatigue performance of additively manufactured aluminum alloys. The new numerical framework will use artificial intelligence algorithms to significantly accelerate CPFEM simulations and thus permit numerical simulations of fatigue that can account for the full complexity of microstructures obtained by additive manufacturing.