Multiscale Simulation-Based Life Prediction

Vanderbilt University

Caglar Oskay


The ability to accurately predict the performance of structures made of composites and other heterogeneous materials, particularly the long-term performance, has tremendous benefits in both economical and safety terms for civil infrastructure and many other structural systems. Physics-based modeling of the multiple failure mechanisms, including the long-term evolution of these failure mechanisms, is a multiscale problem in space and time. Multiple spatial scales exist since many failure mechanisms initiate and grow at the scale defined by the material heterogeneity, whereas the overall failure is assessed at the scale of the structure. Multiple temporal scales exist because of the disparity between the characteristic period of the applied loads, and the overall life of the structure. The problem is further complicated by the interaction of the structure with its environment, causing reaction-induced changes and deterioration in the mechanical state. Currently, the computational capabilities and tools for prediction of the long-term performance are limited, and the available tools are largely phenomenologically-based life models that typically view the prediction problem separate and only weakly coupled to the response models.


At MCML, we are developing a computational life prediction framework that accounts for progressive failure mechanisms that operate at multiple time and length scales. The life prediction framework is built on the application of computational homogenization method concurrently applied in time and space. The computational complexity of this multiscale approach is overcome using spatial and temporal reduced order modeling. The multiscale approach is integrated into a Bayesian probabilistic framework, which links the uncertainties at various time and length scales to the structural life prediction.


The computational life prediction framework was formulated and implemented to predict the progressive damage accumulation and life of fiber reinforced polymer composites subjected to fatigue loading. Novel reduced order models developed to accelerate solution in space and time have been demonstrated to achieve efficiencies that make largescale structural life predictions possible. We have demonstrated the performance of the framework in the context of graphite fiber reinforced epoxy. A limited set of experiments conducted on simple unidirectional composite layups was employed to calibrate the constituent properties at a probabilistic setting. The calibrated probabilistic multiscale model was exercised to predict the mean strength as well as the probability distributions of strength and strainto-failure under monotonic and fatigue loads.


MCML's research is building the foundational knowledge and the computational toolset needed for a paradigm shift in design and maintenance of composite structures in civil infrastructure and other structural systems. The new paradigm is damage tolerance. Many decades of research in metals allowed the possibility of tolerance for damage in metallic structures. MCML's research is contributing to bringing this paradigm to the world of composite structures, which will lead to safer and more economical structures.

Core competencies

  • Multiscale computational failure modeling of solids and structures
  • Life prediction and performance assessment of structures
  • Extreme environment modeling of composites and other heterogeneous materials
  • Modeling of multiphysics systems

Current research team members

  • Dr. Robert D. Crouch (Postdoc)
  • Michael J. Bogdanor (Ph.D. Student)
  • Tong Hui (Ph.D. Student)
  • Matthew G. Pike (Ph.D. Student)
  • Theodore M. Russell (Undergraduate)
  • Paul A. Sparks (Ph.D. Student)
  • Shuhai Zhang (Ph.D. Student)
  • Xiang Zhang (Ph.D. Student)
  • Hao Yan (Ph.D. Student)

Recent PhD graduates

  • Robert D. Crouch, Ph.D., 2012 (Now at Vanderbilt University)
  • Arun Krishnan, Ph.D., 2010 (Now at Simulia)