Image-based Modelling and Simulation

成像模式越来越多地用于以非破坏性方式监测或测试材料和过程的状态。然而，从成像到计算模拟和分析的步骤需要劳动密集型的分割和网格划分过程，每个过程都有自己的不确定性。此外，成像数据本身存在不确定性，这对模拟结果有很大影响。我们建议重点开发一个框架，以缓解图像和模型之间的差距。将开发工具，使用机器学习技术识别/提取图像特征/统计（例如材料属性，边界条件，几何形状），然后将其转化为一系列有代表性的计算模型，通过HPC探索成像不确定性的因果影响。设想的项目将使用材料微观结构（静态）的微ct数据和流动（动态）的超声图像来创建和测试框架。由此产生的框架应消除在基于图像的建模和仿真管道中典型产生的一些不确定性，同时通过简化过程和参与来简化诊断中成像模式的集成。

Imaging modalities are increasingly used to monitor or test the state of materials and processes in a non-destructive manner. However, the step from imaging to computational simulation and analysis requires labour intensive segmentation and meshing process each with their own uncertainties. Furthermore, there is uncertainty in the imaging data itself, which can have a large effect on the simulation results. We propose to focus on the development of a framework to ease the gap between image and model. Tools will be developed that identify/extract image characteristics/statistics (e.g. material properties, boundary conditions, geometry) using machine learning techniques, which will then translate into a range of representative computational models that explore the causal effects of the imaging uncertainties through HPC. Envisaged projects will use microCT data on material microstructures (static) on and ultrasound images on flows (dynamic) to create and test the framework. The resulting framework should eliminate some of the uncertainties typical incurred in image-based modelling and simulation pipelines, whilst easing the integration of imaging modalities in diagnostics by simplifying the processes and involvement.