Title: Inverse Method for Characterization of Material Properties  

Speaker: Miguel A. Aguilo, PhD., Cornell University

Date/Time: Monday, June 14, 2010, 9:00 – 10:00 am       

Location: CSRI Building/Room 90 (Sandia NM)

Brief Abstract: This work presents two numerical approaches to inversely and non-invasively characterize material properties. Since characterizing material properties directly is not always desirable or even possible, these techniques—the L2-adjoint inverse approach and the concept of error in constitutive equation—are gaining popularity in many scientific and engineering areas. For example, in aerospace engineering, noninvasive techniques can be used to inversely characterize material properties from the structural response of a remotely excited aircraft component. Variations in wing or fuselage response could give an indication of structural damage. In medicine, the same techniques can be used to detect material properties of human tissue, which may be healthy or unhealthy. In a joint effort, the Mayo Clinic solicited Cornell’s assistance to determine if these approaches could be applied to detect unhealthy tissue which could give an indication of disease; both techniques were exercised. With the L2-adjoint inverse approach, the objective function measured the difference between experimental measured responses and the finite element representation of the human tissue. Then, the adjoint method was used to compute the gradient with respect to the design variables. With the concept of error in the constitutive equation method, the objective function measured the extent of disagreement between dynamic and kinematic admissible solutions. The gradient was computed by taking the derivative of the error in the constitutive equation functional with respect to the design variables. In both numerical frameworks, total variation regularization was used to stabilize the inverse solution and a Quasi-Newton optimization algorithm was exercised to solve the optimization problem. While both methods provide an effective framework to characterize material properties, better results were obtained faster with the concept of error in constitutive equation method than with the adjoint approach.

CSRI POC: James Stewart, (505) 844-8630