Optimization of crystal plasticity parameters with proxy materials data for alloy single crystals
Optimization of crystal plasticity parameters with proxy materials data for alloy single crystals
复制标题
DOI：
10.1016/j.ijplas.2024.103894
复制
发表时间：
2024-03
影响因子：
9.8
通讯作者：
Shahram Dindarlou;G. Castelluccio
中科院分区：
文献类型：
--
作者：
Shahram Dindarlou;G. Castelluccio
多尺度建模方法在理解、预测和工程材料响应方面显示出了巨大的价值。虽然计算能力的提高有助于从第一原理出发模拟原子行为，但模拟介观机制，如晶间破坏或裂纹萌生，仍然强烈依赖相关模型。晶体塑性模型已被广泛用于关联金属材料的工艺-性能-结构，包括织构、微塑性和失效变异性等细观尺度效应。然而，模型在颗粒尺度上的预测能力仍然很低，这导致在实验校准集之外的损伤预测很差。除了模型形状误差之外，中尺度的不确定性是由不充分的模型参数引起的，这完全是由于对宏观实验数据的校准引起的。这项工作探索了晶体塑性模型中的参数不确定性，并提出了一种基于物理和数值优化的混合方法来识别与面心立方金属和合金中的介观强化相关的参数。该方法的强度和新颖性依赖于使用单晶和多晶应力-应变曲线独立地校准参数。我们进一步证明，可以将多个材料同时合并到单个优化算法中，以稳健地量化中尺度材料不变参数。然后，这些值被用来盲目预测单晶和多晶工程合金的响应。因此，我们的方法通过增加来自具有相似位错结构的不同材料(即替代材料)的单晶实验的校准数据来减少建模的不确定性。这些结果为晶体塑性模型的稳健参数化提供了基础，即使在没有直接实验数据的情况下，该模型也可以预测工程合金的单晶和多晶响应。
Multiscale modeling approaches have demonstrated ample value in understanding, predicting, and engineering materials response. While increasing computational power has aided in modeling atomic behavior from first principles, modeling mesoscale mechanisms such as intergranular failure or crack initiation still rely strongly on correlative models. Crystal Plasticity models have been extensively used to relate process-property-structure in metallic materials including mesoscale effects such as texture, microplasticity, and failure variability. However, models still suffer from low predictive power at the grain scale, which leads to poor damage prognosis outside the experimental calibration set. In addition to model form error, mesoscale uncertainty is dominated by an inadequate model parameterization that arises from calibration exclusively to macroscopic experimental data. This work explores parameter uncertainty in Crystal Plasticity models and proposes a hybrid physic-based and numerical optimization approach to identify parameters associated to mesoscale strengthening in FCC metals and alloys. The strength and novelty of the approach rely on calibrating parameters independently using single-crystal and polycrystal stress–strain curves. We further demonstrate that multiple materials can be incorporated simultaneously into a single optimization algorithm to robustly quantify mesoscale material-invariant parameters. These values are then used to blindly predict the response of single- and poly-crystals engineering alloys. As a result, our approach mitigates modeling uncertainty by augmenting the data for calibration with single crystal experiments from different materials with similar dislocation structures (i.e., proxy materials). The results provide the basis for a robust parameterization of crystal plasticity models that can predict single- and poly-crystal responses for engineering alloys even in the absence of direct experimental data.