Developing a "microstructural fingerprint" of titanium alloys - metallurgy in the information age

我们生活中几乎每个方面都要用到的金属材料的性能都高度依赖于它们的微观结构。金属内部的晶粒和相界模式，材料内部的晶粒形状和缺陷分布（如层错（二维），位错（一维）和点缺陷（零维））对于确定这些特性非常重要。现代工程合金在成分和加工（合金配方的成分和步骤）方面的复杂性很大程度上是由于需要在非常具有挑战性的服务环境中实现高度优化的性能。因此，奇怪的是，我们没有一种普遍认可的语言来描述材料的微观结构。例如，通常，多晶中的晶粒图案可以用平均晶粒尺寸和晶粒形状的一些度量来描述。显然，这遗漏了详细微观结构中固有的大部分信息。现代高分辨率、高通量的实验表征设备可以以非常高的速率生成微观结构的详细图像，但大多数信息都被有效地立即丢弃：原始数据文件太大而无法处理（通常太大甚至无法保留），我们缺乏一种描述语言来详细捕捉微观结构的本质。本博士项目将着手解决这一不足。您将开发一种方法，其中计算机视觉和图像分析工具与机器学习方法一起使用，以产生合金系统的“微观结构指纹”。该项目由劳斯莱斯公司赞助，将以用于喷气发动机风扇叶片的Ti6/4合金为例材料。这种材料具有丰富的微观结构，需要在多个长度尺度上进行描述。此外，微观结构直接影响到几个关键的性能特征，劳斯莱斯公司拥有一个庞大的材料、性能和性能数据数据库。

The properties of the metallic materials that we rely on in almost every aspect of our lives are highly dependent on their microstructures. The patterns of grain and phase boundaries within the metals, the grain shapes and the distribution of defects within the material (such as stacking faults (2-dimensional), dislocations (1-d) and point defects (0-d)) are hugely important in determining these properties. Much of the complexity in modern engineering alloys in terms of composition and processing (the ingredients and steps of the alloy recipe) is a result of the need to achieve highly optimised properties for deployment in very challenging service environments.It is therefore curious that we have no universally agreed language for describing material microstructure. Often, for example, a pattern of grains in a polycrystal might be described by no more than an average grain size and some measure of the grain shape. Clearly this misses most of the information inherent in the detailed microstructure. Modern high-resolution, high-throughput experimental characterisation equipment can generate detailed images of microstructure at a very high rate, but most of the information is effectively thrown away immediately: the raw data files are too big to handle (and often too big even to retain) and we lack a descriptive language for capturing the essence of the microstructure in detail.This PhD project will begin to address this deficiency. You will work to develop a methodology in which the tools of computer vision and image analysis are used alongside machine learning methods to produce a "microstructural fingerprint" of an alloy system. The project is sponsored by Rolls-Royce and will take as an example material the Ti6/4 alloy used for fan blades in jet engines. This material has a rich microstructure, requiring description on multiple length scales. Furthermore, the microstructure directly influences several key performance characteristics and Rolls-Royce has available a large database of material and properties and performance data.