CAREER: Manufacturing USA: Deep Learning to Understand Fatigue Performance and Processing Relationship of Complex Parts by Additive Manufacturing for High-consequence Applications

金属增材制造（AM），如激光粉末床熔合（LPBF），不仅在产品创新方面得到了越来越多的探索，而且在车间生产方面也得到了越来越多的探索，从各种行业中获得了越来越多的成功。然而，由于缺乏对LPBF部件疲劳失效和性能不确定性的了解，这对LPBF在高后果应用中部署的潜力构成了重大挑战。该学院早期职业发展（Career）奖支持基础研究，以了解LPBF加工对缺陷和随后的疲劳行为的影响，推进复杂几何形状和受多轴载荷影响的LPBF零件的疲劳散射知识。这项工作将为疲劳寿命预测建立一个以物理为中心的机器学习框架，作为未来金属增材制造动态承载应用的技术基础，从而提高美国工业的竞争力。CAREER项目还将整合教育和外展项目，旨在通过积极吸引K-12学生参加STEM教育，招募女性和少数族裔参与研究，扩大代表性不足群体的参与，为未来几代多样化的工程师提供制造业创新和大数据时代不可或缺的知识和技能。这项早期工作的最终目标是了解复杂LPBF部件在多轴载荷下的疲劳失效，以进行数据驱动的疲劳寿命预测。该研究将研究在最高应力集中的塑性变形和裂纹萌生引起的疲劳失效的性质，并使用多尺度方法将疲劳寿命预测转化为评估脆弱区域的裂纹扩展。在微观尺度上，将根据与疲劳失效的相关性识别具有裂纹启动特征的关键缺陷（通过x射线计算机断层扫描或光学轮廓术）；关键缺陷及其空间相互作用对裂纹扩展的影响将使用断裂力学和数据密集统计进行检查。在局部尺度上，将通过解耦多轴加载对应力和应变行为进行有限元建模，以检查最高应力集中的薄弱区域。然后，将临界缺陷和脆弱部位主应力的影响纳入深度学习的分层图卷积网络中，以模拟它们对裂纹扩展的协同影响，并通过高级数据分析计算LPBF零件的疲劳寿命。研究结果有望产生与LPBF部件疲劳性能相关的缺陷形成的新知识，揭示多尺度因素对疲劳断裂的协同影响，并进一步将LPBF应用于高后果应用。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Metal additive manufacturing (AM) such as laser powder-bed fusion (LPBF) has been increasingly explored not only for product innovation, but also shop-floor production, demonstrated by growing success from a variety of industries. However, the lack of knowledge in both fatigue failure and the performance uncertainty of LPBF parts poses a significant challenge and undermines the potential of deploying LPBF for high-consequence applications. This Faculty Early Career Development (CAREER) award supports fundamental research to understand the effects of LPBF processing on defects and subsequent fatigue behavior, advance the knowledge of fatigue scattering of LPBF parts that are complex in geometry and subject to multiaxial loading. The effort will establish a physics-centric, machine learning framework for fatigue life predictions, serving as a technological foundation for future metal AM production of dynamic load-bearing applications, and thus, enhance the competitiveness of U.S. industry. This CAREER project will also integrate education and outreach programs designed to broaden the participation from underrepresented groups through actively engaging K-12 students for STEM education and recruiting women and minorities into research, priming future generations of diverse engineers with the knowledge and skills indispensable in the age of manufacturing innovation and big data.The ultimate goal of this early career effort is to understand fatigue failures of complex LPBF parts under multiaxial loading for data-driven fatigue life predictions. The research will investigate the nature of fatigue failures from plastic deformation and crack initiation at the highest stress concentrations and translate fatigue life predictions into evaluating the crack growth at the vulnerable zones using a multiscale approach. On the micro-scale, critical defects with crack-initiating features (by x-ray computed tomography or optical profilometry) will be identified based on the correlation with fatigue failures; both the effects of critical defects and their spatial interactions on crack growth will be examined using fracture mechanics and data-intense statistics. On the part scale, the weak regions of the highest stress concentrations will be examined by finite element modeling of stress and strain behaviors through decoupling multiaxial loading. The effects of critical defects and the principal stresses at vulnerable localities will then be incorporated into a hierarchical graph convolutional network of deep learning to model their synergistic impacts on crack growth and calculate the fatigue life of LPBF parts with advanced data analytics. The findings are expected to generate new knowledge of defect formation relevant to fatigue performance of LPBF parts, uncover the synergistic impacts of multiscale factors on fatigue fractures, and further LPBF adoption for high-consequence applications.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.