|
|
|
5 lectures In-depth Analysis of a Hopkinson Bar Experiment, Pr. Gérard Gary, Ecole Polytechnique The presentation will be student-oriented. The objective is to propose a critical approach to that of Kolsky, almost systematically quoted in all the articles presenting results of tests on Hopkinson bars. We will first recall the hypotheses, in particular those which are not stated. For this, a reminder of the basic knowledge necessary for understanding will be given. Various subtleties can then be approached through examples of test analyzes (*). Interventions and questions from the public will be welcome and even desired. (*) If participants wish to have data from their trials analyzed, we can respond to their request provided that they are provided in "David" format - mainly in ASCII - which I can specify on request. Metal Response to Shock Loading: From Microscale to Applications, Dr. Bolis Cyril, CEA Equation of state, strength and damage models are the main components needed when you want to perform shock simulations. If some part of theses models can easily be adjusted by simple experiments and considerations, it is always difficult to obtain some data along all the thermo-mechanical path encounter during our simulations. It is also difficult to understand some effects using only macroscopic considerations such as phases changes, kinetics. That's why some multi-scales approaches have been developed in the last decades, in particular with the extensive use of supercomputers. In our speech, we will present a quick overview of what can be done with theses methods. The Addition of DIC measurement to the Split Hopkinson Bar Experiment, Pr. Amos Gilat, The Ohio State University The Digital Image Correlation (DIC) technique for full field measurement of deformation is probably the most significant advance in experimental mechanics in recent years. When applied to the SHB experiment it provides means for examining the validity of the strain measurement from the recorded waves that assumes uniform deformation in the specimen during the test, and it provide means for conducting tests where the deformation is intentionally not uniform. Both issues are addressed, and several experimental setups where DIC has a crucial role are presented. Composite Materials and Multi-material Assemblies under Dynamic Stresses at High Strain-rates, Pr. Michel Arrigoni, ENSTA Bretagne The requirements for the service life of some high value-added structures in the aeronautics or aerospace fields, as well as for defense applications, are becoming higher and higher. The expected mechanical performances are increased to resist to severe shocks and impacts for lower and lower expected structural weigh. This antagonism gives birth to the paradigm of physical protection by advanced materials. Among these materials, composite materials remain a preferred choice but cannot be sufficient on their own in some situations involving extreme loads such as detonation waves or shock waves resulting from ballistic impacts or hypervelocity fragments. In these cases, the dynamic behaviour must be revisited to consider the supersonicity related to shock waves propagating in the target material. This presentation aims at presenting experimental methods allowing the establishment of numerical models dedicated to this kind of events. These models are used in predictive tools of numerical modelling in order to limit the number of experiments, generally dangerous, expensive and where some key parameters cannot be easily tuned. The numerical tool also allows optimization steps. A few examples taken from the research work carried out at the Dupuy de Lôme Research Institute illustrate the modeling approach. These examples highlight the progress of the experimental and numerical methods used to study the mechanical effects of impacts and explosions. An Introduction to Machine-learning for Modeling the Dynamic Behavior of Materials and Structures, Pr. Dirk Mohr, ETHZ Machine learning is a powerful computational modeling tool whose application is currently pursued in many branches of engineering science. In this lecture, we introduce basic deep learning techniques such as fully-connected neural networks (FCNNs), graph-neural networks (GNNs), convolutional neural networks (CNNs) and recurrent neural networks (RNNs). We then ask the question about their potential application in the field of dynamic behavior of materials. It is demonstrated that FCCNs provide a very useful tool for modeling the rate- and temperature dependent hardening of both metals and polymers. The merits of FCCNs for predicting the dynamic crushing response of thin-walled structures is also shown. We the proceed to the application of GNNs to facilitate the design of optimal lightweight structures such as truss- and shell-lattices. And finally, the development of new RNNs is discussed that address the mechanics-specific challenges of modeling history dependent behavior. The outcome is universal material model which can capture the large deformation response of a wide spectrum of engineering material.
|
| Online user: 2 | Privacy |
|
