Machine learning guided alloy design and thermomechanical process optimisation for high performance automotive aluminium alloys

传统的基于直觉和试错实验用于开发新合金和/或优化加工条件是非常耗时和昂贵的方法。基于物理学原理的材料建模和工艺模拟有助于在一定程度上减少时间和成本，以寻找满足特定工程应用技术要求的给定金属材料的合金成分和加工参数方面的最佳解决方案。然而，合金化学和材料加工方法的日益复杂性可能会对当前基于多物理学的计算建模提出挑战，因为缺乏合金成分/加工和性能之间的定量唯象关系。机器学习（ML）的强大之处在于它可以对高维数据进行操作，以理解多元素合金成分和多步骤处理的复杂性。由机器学习方法生成的数据驱动模型已广泛用于材料工程领域，以成功找到导致有效材料设计的复杂相关性（1-6）。

Traditional intuition-based and trial-and-error experimentation used in the development of new alloys and/or optimisation of processing conditions are very time-consuming and costly approaches. Materials modelling and process simulations guided by physics-based principles have helped to reduce the time and cost to some extend in the search for optimal solutions in terms of alloy composition and processing parameters for a given metallic material that meets the technical requirements of specific engineering applications. However, the increasing complexity of alloy chemistry and materials processing methodologies can be challenging to current multiphysics-based computational modelling due to the lack of quantitative phenomenological relationships between the composition/processing and properties of alloys. The power of machine learning (ML) is that it can operate on high-dimensional data to make sense of the complexity of multi-element alloy compositions and multi-step processing. The data-driven models generated by machine learning approach have been used extensively in the materials engineering domain to successfully find such complex correlations leading to effective materials design(1-6).