Enabling high throughput molecular dynamics with automation and machine learning for development of advanced engineering polymers

计算建模在材料科学和工程中越来越重要。与实验室工作相比，建模有明显的优势：它是无限可重复的，产生丰富的可用数据，不会产生浪费材料，并且在成本和时间投资方面都相当便宜。然而，目前的建模模拟仅限于它们设计的长度尺度（例如宏观尺度或原子尺度），这最终限制了它们的应用范围。这项工作旨在开发一种方法，使尺度之间的桥梁产生一个真正的多尺度建模实验。这一新的进步将使复杂结构的高精度模型的生产成为可能。这些模型将促进快速原型制作，减少对物理测试的依赖。此外，这项工作的另一个目标将是扩大复合结构的建模潜力，特别是在低长度尺度模拟中，并将其纳入多尺度模型方法。这项工作将利用分子动力学（MD）对聚合物、纳米材料和碳纤维的原子化学结构进行建模。MD非常适合模拟聚合物结构的固化过程，以及它们如何与其他材料（如碳纤维表面）相互作用。由MD确定的最终结构和材料特性将被输入到中尺度模型中，将尺度从纳米提高到微米。这些模型将揭示聚合物基体和碳纤维在单一复合材料层中的作用，并可用于解释材料的机械性能。最后，中尺度模型的结果将用于产生高质量的有限元（FE）模型。有限元分析能够展示大型有形结构，如梁，板或桁架，如何响应外部负载。使用最初在MD中建模的材料允许在FE模型中详细描述材料属性，提高模型的准确性并识别有可能满足给定任务的材料。这整个过程可以用不同的初始材料重复，以完善被建模结构的最终特性。这项工作的应用是广泛的，因为建模技术的任何发展不仅增强了无机材料的建模（例如金属，电池，超导体），而且还增强了药物设计领域的生物建模。在复合材料领域，这种多尺度建模将对航空航天、汽车和可再生能源行业产生重大影响，有助于设计风力涡轮机等复杂结构。同样重要的是，该技术可能会引起材料制造商的兴趣，因为它有可能加速材料的发现和进一步开发定制材料。

Computational modelling is of increasing importance to materials science and engineering. Modelling has clear advantages over laboratory work: it is infinitely repeatable, generates a wealth of useable data, doesn't produce waste material, and is considerably cheaper in terms of both cost, and time investment. However, current modelling simulations are limited to the length scale (e.g. macroscale or atomistic) for which they were designed, which ultimately limits their application scope. This work aims to develop a methodology that enables bridging between scales to produce a true multiscale modelling experiment. This novel advancement would enable the production of high accuracy models for complex structures. These models would facilitate rapid prototyping and reduce reliance on physical testing. Furthermore, an additional objective of this work will be to expand the modelling potential of composite structures, particularly in low length scale simulations, and incorporate this into the multiscale model methodology.This work will model atomistic chemical structures of polymers, nanomaterials and carbon fibre by using molecular dynamics (MD). MD is well suited to modelling the curing process of polymer structures and how they interact with other materials such as at the surface of carbon fibre. The final structure and material properties determined by MD will then be fed into mesoscale models, raising the scale from nanometre to micrometre. These models will reveal how a polymer matrix and carbon fibre act in a single composite ply and can be used to explain the mechanical properties of a material. Finally, the results of the mesoscale model will be used to produce a high-quality finite element (FE) model. FE analysis is able to demonstrate how large, tangible structures, such as beams, panels or trusses, respond to external loads. Using material originally modelled in MD allows for detailed material properties in the FE model, enhancing the accuracy of the model and identifying materials that have potential to cater to a given task. This whole process can be repeated with different starting materials to refine the final properties of structure being modelled.Application of this work is broad as any development of modelling technology not only enhances inorganic materials modelling (e.g. metals, batteries, superconductors) but also biological modelling for the world of drug design. In composites, this multiscale modelling would be of significant interest to the aerospace, automotive and renewables industries to aid in the design of complex structures such as wind turbines. Of equal importance, the technology could be of interest to materials manufacturers as it has the potential to accelerate materials discovery and further the development of bespoke materials.