AI for inverse problems in solid mechanics

固体力学反问题的人工智能

• 批准号: 2117845
• 金额: --
• 依托单位: University of Bristol
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2018
• 资助国家: 英国
• 起止时间: 2018 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2117845
• 关键词: AI; inverse; problems; solid; mechanics

Inverse problems use measurements from a system response to calculate unknown values for the system and its inputs. This is advantageous when there is limited information about the system, unknown inputs, multiple solutions or indeed no solution exists. Essentially, you start with something that has failed and use information from this to work out the failure conditions. Unknown properties are everywhere. All seemingly well-defined material properties are distributions. Many safety critical industrial applications in power generation, transport and construction, have complex loading cases that are not fully defined. Therefore, applying the inverse problem technique will present an advantage for improving designs in these areas. Having multiple unknowns in multiple areas creates computational demand. Due to this, using the inverse problem technique is often dismissed. Thus, I will be utilising the highest form of computational power available through High Performance Computers (HPC). To build a model for HPC, I will be considering solid mechanic theory, mathematical theory, computational theory and software development. Once a basic model has been built, an optimisation process will take place through benchmarking and profiling. The augmentation of artificial intelligence (AI) will be an integral part of the project. Using the statistical methods designed for AI can be used to optimise and reduce the number of iterations required for a solution. The Solid Mechanics Research Group has the expertise and knowledge to facilitate this project. Dr. Anton Shterenlikht has a wide expertise in HPC methods. His knowledge of with FORTRAN coarrys will be invaluable for this project.

逆问题使用来自系统响应的测量来计算系统及其输入的未知值。这在关于系统的信息有限、未知输入、多个解或实际上不存在解时是有利的。从本质上讲，你从失败的东西开始，并使用从中获得的信息来计算失败的条件。到处都是未知的属性。所有表面上明确定义的物质属性都是分布。在发电、运输和建筑领域的许多安全关键工业应用具有复杂的载荷情况，这些载荷情况尚未完全定义。因此，应用反问题技术将有利于改善这些领域的设计。在多个领域有多个未知数会产生计算需求。因此，使用反问题技术往往被驳回。因此，我将利用高性能计算机（HPC）提供的最高形式的计算能力。为了建立HPC模型，我将考虑固体力学理论，数学理论，计算理论和软件开发。一旦建立了一个基本模型，将通过基准测试和分析进行优化。人工智能（AI）的增强将是该项目不可或缺的一部分。使用为AI设计的统计方法可以用于优化和减少解决方案所需的迭代次数。固体力学研究小组拥有促进该项目的专业知识和知识。Anton Shterenlikht博士在HPC方法方面拥有广泛的专业知识。他在FORTRAN语言方面的知识对这个项目将是非常宝贵的。