Crack Growth During Fatigue in Ni Superalloys: Physical Origin of Stochastic Jumps and Their Predictive Role Using Statistical Approaches

坚固耐用的材料是我们社会不可或缺的一部分。在航空航天和海洋工业中使用的涡轮发动机中发现的一类材料被称为高温合金。高温合金具有优异的机械性能（强度、抗蠕变、耐腐蚀）。然而，由于所需的材料制造过程，这些材料涉及很大程度的结构混乱，这是一个问题。这种混乱的影响在涡轮机的高温下变得更加明显，由于持续的负载条件，导致材料的微观损伤和裂缝。这些裂纹会随着机器的使用寿命而加剧。因此，了解这些裂纹的行为和防止灾难性的机械失效是至关重要的。在非均质材料中裂纹的产生和扩展不仅是有害的，而且是不可预测的，因此需要统计方法和协议来评估部件在疲劳加载的各个阶段的可靠性。该项目将推进金属非均质材料在循环加载（疲劳）过程中的随机裂纹扩展跳变科学，特别是Ni高温合金。这种小裂缝产生的机械噪声的有用之处在于，它可能包含独特的统计特征，可以识别涡轮机部件的损坏程度。一个由工程师和科学家组成的团队将结合多尺度建模方法、统计方法和实验，最终开发出结合实验和理论的协议，以表征疲劳引起的“开裂噪音”，并评估机械部件的损伤水平。除了高温合金之外，本研究的结果还将促进对疲劳损伤的基本认识和开发非侵入性结构预测方法。此外，还计划在代表性不足的EPSCoR州西弗吉尼亚州开展一项教育推广计划，涉及研究生、本科生、高中生以及普通公众。本项目将促进对Ni高温合金疲劳加载过程中随机跳变的认识。将采用多尺度建模方法，将密度泛函理论（DFT）预测与相场建模相结合。机器学习方法将被纳入相场模型，该模型将基于已进行的实验进行训练。这项研究的结果将是对疲劳损伤的基本理解，可用于预测灾难性失效，特别是在有限的统计抽样情况下。一组工程师和科学家将开发一种新的途径来预测金属高温合金疲劳载荷期间裂纹扩展的预测建模：通过在恒定应力短时间测试的假设下，统计采样疲劳各个阶段的噪声相关性，我们将使用直接多步骤预测策略构建预测机器学习框架。在此过程中，我们将研究随机裂纹增长跳跃的基本起源，并将开发一个概率模型，该模型将结合密度泛函理论预测和相场建模产生的内聚能的第一性原理关系。为了验证我们的模型，我们将使用原位扫描电镜进行一系列控制良好的实验，我们将使用直流电阻降测量来跟踪裂纹扩展。裂纹扩展噪声在不同阶段随温度和环境压力的统计特性将与多尺度模型预测结果进行比较。验证后的多尺度模型将用于研究前几个循环中裂纹扩展事件（按裂纹长度变化分类）的概率分布，以预测后期阶段的裂纹扩展。结果将是一个经过训练的模型，可以根据早期疲劳事件预测故障。基于金属高温合金疲劳加载时裂纹扩展跳跃的基本物理起源，该研究项目具有社会影响，金属高温合金通常用于飞机涡轮和其他硬件。目的是制定通用方案，以促进早期，安全的预测金属合金的裂纹扩展。除了社会影响外，还计划开展一项教育推广计划，包括在代表性不足的EPSCoR州西弗吉尼亚州培训研究生、本科生和高中生以及普通公众。培训的重点将放在材料科学的计算建模的使用上，以及对裂纹扩展、断裂和非平衡罕见事件的基本物理性质的深刻理解。PI将设计一门课程，向多学科本科工程环境介绍非平衡统计力学和断裂的基础知识。

Non-Technical AbstractStrong, durable materials are an integral part of our society. One such class of materials found in turbine engines, used in the aerospace and marine industries, are known as superalloys. Superalloys exhibit excellent mechanical properties (strength, creep resistance, corrosion resistance). However, there is a catch in that these materials involve a large degree of structural disorder as a result of the required material manufacturing process. The effects of such disorder become even more pronounced at the high temperatures of turbines, due to sustained loading conditions, leading to microscopic damage and cracks in the material. These cracks are exacerbated over the lifetime of the machinery. Therefore, it is crucial to understand the behavior of these cracks and prevent catastrophic mechanical failure.Crack initiation and growth in very heterogeneous materials not only can be detrimental but also very unpredictable, thus it requires statistical methods and protocols for assessing the reliability of components at various stages of fatigue loading. This project will advance the science of stochastic crack growth jumps during cyclic loading (fatigue) of metallic heterogeneous materials, with a particular focus on Ni superalloys. The usefulness of the mechanical noise produced by such little cracks is that it might contain distinctive statistical features that can identify the damage level in a turbine component. A team of engineers and scientists will combine multi-scale modeling approaches, statistical methods, and experiments to ultimately develop combined experiment and theory protocols for characterizing the fatigue-induced "cracking noise" and assessing the damage levels of mechanical components. Beyond superalloys, the very outcome of this research is to promote the progress of the fundamental understanding of fatigue damage and develop non-invasive structural prognosis methods. An educational outreach program is also planned that involves graduate, undergraduate, and high-school students, as well as the general public, in the under-represented EPSCoR state of West Virginia.Technical Abstract This project will advance the understanding of stochastic jumps during fatigue loading of Ni superalloys. A multi-scale modeling approach will be employed that will combine density functional theory (DFT) predictions with phase-field modeling. Machine-learning methods will be incorporated into the phase field model, which will be trained based on conducted experiments. The outcome of this research will be the fundamental understanding of fatigue damage that may be used to predict catastrophic failures, especially when there is limited statistical sampling.A team of engineers and scientists will develop a novel pathway to predictive modeling of crack growth during fatigue loading in metallic superalloys: By statistically sampling the noise correlations at various stages of fatigue under the assumption of constant-stress short-time tests, we will build a predictive machine-learning framework using a direct multi-step forecasting strategy. In doing so, we will investigate the fundamental origin of stochastic crack growth jumps and will develop a probabilistic model that will incorporate a first-principles relationship of the cohesive energy, generated by density functional theory predictions and phase-field modeling. To validate our models, we will conduct a series of well-controlled experiments using in-situ SEM and we will track crack growth using DC resistance drop measurements. The statistical properties of crack growth noise at various stages as a function of temperature and environmental pressure will be compared to the multi-scale model predictions. The validated multi-scale model will then be used to investigate the probability distributions of crack growth events (classified in terms of crack-length changes) during the first few cycles to predict crack growth at late stages. The outcome will be a trained model that can predict failure based on early fatigue events.This research project has a societal impact based on the fundamental physical origin of crack growth jumps during fatigue loading of metallic superalloys, which are commonly used on aircraft turbines and other hardware. The aim is to develop general protocols to promote early, safe prediction of crack growth in metallic alloys. In addition to societal impact, an educational outreach program is planned that involve training graduate, undergraduate, and high-school students, as well as the general public, in the under-represented EPSCoR state of West Virginia. The focus of training will be on the use of computational modeling materials science as well as the deep understanding of basic physical properties of crack growth, fracture, and non-equilibrium rare events. The PI will design a course that will introduce the fundamentals of non-equilibrium statistical mechanics and fracture to multidisciplinary, undergraduate engineering environments.