Image-based Mechanics: an Overview of Experimental and Numerical Approaches

The proposed course is aimed at drawing an overview of recent developments in the field of image-based mechanics to give the participants clues for understanding how images are acquired, used or analysed and how they can be used for understanding and validation purposes. Having a global vision of the whole pipeline from the lab to the computing center is essential for practitioners and researchers to make their work more effective. The course is divided into several blocks to answer the following questions: how are images acquired? What can we do with images? How is image analysis formulated/implemented? How robust are the results? To help the participants in their future use of images, this comprehensive overview of fullfield measurements deployment and quantitative image analysis is coupled with an in-depth exploration of best practices in formulation and implementation. The immense potential of these techniques is showcased through discussions on related inverse methods for material characterisation, multimodal setups, and in-situ/in-operando implementations but also on the evaluation of their performance. Among the possible uses of images in mechanical analysis, full-field displacement measurements and microstructure analysis are covered. Concerning full kinematic field measurements, an overview of different versions and their associated practical implementation in the lab is presented together with an introduction of the numerical implementation of digital image correlation (DIC), localized spectrum analysis (LSA) and optical flow (OF) methods. Applications and specific features of 2D, stereo and volume imaging will be discussed. Quantitative image analysis is discussed for microstructure characterization and crack detection using 3D images. Various image processing techniques, from classical mathematical morphology to machine learning-aided image segmentation, are introduced and discussed. Approaches for model-based quantitative image analysis will also be showcased. Requirements on image resolution and methods for correcting sampling biases will be addressed. From the results obtained using the previously mentioned techniques, it is now common to identify the mechanical behaviour of the studied materials at the chosen observation scale. The course thoroughly explores advanced identification methods, offering participants a deep understanding of techniques such as Finite Element Model Updating (FEMU) and Virtual Fields Method (VFM). Through practical demonstrations and theoretical discussions, participants will gain proficiency in implementing these advanced methods to enhance material characterisation and simulation accuracy. Recently proposed non-parametric (without postulating a constitutive relation) identification techniques based on the data-driven paradigm will also be introduced theoretically and practically. These approaches require a computational model numerically twinning the sample. A short overview of the computational issues regarding geometrical description and computational efficiency is also proposed. Last, the question of robustness is covered. To this end, uncertainty quantification will be considered as an ongoing issue.