Simulation and experimental analysis of material failure in large scale yielding – SEMAFOR

Technologies for analyzing failures processes have recently made great improvements through improved methods for monitoring complex tests, development of new models for damage and fracture analysis, and robust numerical algorithms that take benefits of hybrid parallel machine architecture. Thus, the SEMAFOR project aims to bring together expertise from several teams, specialized in the involved thematics (experimental, material modeling and numerical methods) to develop new models and computational strategies capable of simulating cracking phenomena observed in combustion chambers with much better accuracy than the currently available methods in the literature. To achieve this objective, focusing on the cobalt-base superalloy HA188, strong interaction research activities will be carried on experimental aspects, material modeling, and high-performance numerical simulation.

Regarding the experimental part, as some previous work delivered us an anisothermal model of the HA188 in a wide temperature range, we will focus on the experimental analysis of cracking processes in generalized plasticity in isothermal conditions for loads in mode I (on both plain and multiperforated samples) and mixed-mode I + II using a coplanar biaxial fatigue test facility. These tests will be declined with temperature gradients for the biaxial test and anisothermal conditions for mode I, to establish reference cases. Monitoring of the crack growth will make use of the electric potential method, optical measurements, and image analysis.

For modeling aspects, we propose, on the one hand, to develop skills acquired over a decade on models and numerical tools for addressing the problem of large scale yielding crack growth with cohesive approaches or using integrals invariant for standard generalized behavior or even non-local damage. On the other hand, the development of a bifurcation criterion adapted to generalized plasticity crack growth under multiaxial loading will be undertaken in order to be successfully applied to the three kinds of models.

The modeling efforts lead to numerical simulations that face a multitude of difficulties: internal variables behavior, severe nonlinearities, complex propagation laws, and very fine temporal and spatial discretizations. In all cases, the complexity of the proposed models leads to discretized problems of very large dimension. It is therefore essential to design approaches to get the most out of the new computer architectures with massively multi-core algorithms through several levels of parallelism, error estimation, and mesh adaptation. As the entities involved in the project already possess hardware for distributed computing, this project will aim to demonstrate the feasibility of simulations, in the context of nonlinear crack growth, up to hundreds of millions of unknowns.

One of the strong contributions of this project is based on complementarity, testing, modeling, calculation with the ambition of wanting to deal, through the developed models, with problems of several hundred million unknowns. While it is becoming increasingly clear to both the academic and the industrial world that the architectures of computing machines have irreversibly evolved into a resolutely parallel and multicore hybrid structure, the mechanics community is also awaiting for truly robust high-performance computation strategies to study more and more fine and complex physical problems. In this delicate context, where industrial codes do not seem to be able to adapt effectively to these new computing resources, while their robustness and reliability remain crucial, we need to show the feasibility of such ambition. Therefore, one of the actions of the sustainability of this project will be to include all the development carried out within the Z-set / Zebulon code in order to be able to conduct numerical simulations of cracking with much better robustness and precision than what allow the currently available tools.

The aim of the SEMAFOR project is to develop a set of models, through complex tests, enabling, in the long term, the numerical simulation of fatigue failure phenomena in generalized plasticity within aeronautical components.

The tests, models, and simulations obtained at the end of the project aim to constitute elements allowing a significant evolution of methodologies of prediction of the lifetime in cracking in this type of component and therefore an application in the sizing practices used by the manufacturers of field. Developments carried out in this framework are the subject of scientific papers (publications in international peer-reviewed journals and congresses) and thus contribute to the understanding of generalized plasticity cracking (which is a major challenge for other industrial players And academic). Also, the partners of this project, whose entities are members of the GdR FATACRACK, will be able to share their progress on these issues within this exchange platform.

The SEMAFOR project aims to make significant contributions to experimental characterization, modeling and numerical simulation of the propagation of fatigue cracks under large scale yielding conditions. This work is motivated by strong industrial requirements for predicting the lifetime of critical aircraft components, particularly combustion chambers where cracks may appear in service. Degradation of this type of structure is the result of a complex set of phenomena whose interactions are difficult to take into account. The observed cracks propagate in perforated zones which produce large thermal gradients and microstructural evolutions related to large scale yielding, in an oxidizing environment and under fatigue multiaxial loads. In such conditions, where the failure phenomena can not be considered separately from plastic material evolution, we talk about crack propagation in generalized plasticity.

Technologies for analyzing these processes have recently made great improvements through improved methods for monitoring complex tests, development of new models for damage and fracture analysis, and robust numerical algorithms that take benefits of hybrid parallel machine architecture. Thus, the SEMAFOR project aims to bring together expertise from several teams, specialized in the involved thematics (experimental, material modeling and numerical methods) to develop new models and computational strategies capable of simulating cracking phenomena observed in combustion chambers with a much better accuracy than the currently available methods in the literature. To achieve this objective, focusing on the cobalt base superalloy HA188, strong interaction research activities will be carried on experimental aspects, material modeling and high performance numerical simulation.

Regarding the experimental part, as some previous work delivered us an anisothermal model of the HA188 in a wide temperature range, we will focus on the experimental analysis of cracking processes in generalized plasticity in isothermal conditions for loads in mode I (on both plain and multiperforated samples) and mixed-mode I + II using a coplanar biaxial fatigue test facility. These tests will be declined with temperature gradients for the biaxial test and anisothermal conditions for mode I, to establish reference cases. Monitoring of the crack growth will make use of the electric potential method, optical measurements and image analysis.

For modeling aspects, we propose, on the one hand, to develop skills acquired over a decade on models and numerical tools for addressing the problem of large scale yielding crack growth with cohesive approaches or using integrals invariant for standard generalized behavior or even non-local damage. On the other hand, the development of a bifurcation criterion adapted to generalized plasticity crack growth under multiaxial loading will be undertaken in order to be successfully applied to the three kinds of models.

The modeling efforts lead to numerical simulations that face a multitude of difficulties: internal variables behavior, severe nonlinearities, complex propagation laws and very fine temporal and spatial discretizations. In all cases, the complexity of the proposed models leads to discretized problems of very large dimension. It is therefore essential to design approaches to get the most out of the new computer architectures with massively multi-core algorithms through several levels of parallelism, error estimation and mesh adaptation. As the entities involved in the project already possess hardware for distributed computing, this project will aim to demonstrate the feasibility of simulations, in the context of nonlinear crack growth, up to hundreds of millions of unknowns.