Inverse characterization of composite materials via surrogate modeling. (15th November 2015)

• Record Type: Journal Article
• Authors: Steuben, John; Michopoulos, John; Iliopoulos, Athanasios; Turner, Cameron
• Is Part Of: Composite structures. Volume 132(2016)
• Page Start: 694
• Page End: 708
• Publication Date: 2015-11-15
• Subjects: Material characterization -- Composite materials -- Surrogate models -- Inverse problems; Composite construction -- Periodicals; Composites -- Périodiques; 624.18
• DOI: 10.1016/j.compstruct.2015.05.029
• Languages: English
• ISSNs: 0263-8223

Abstract

Abstract: Methods for the inverse characterization of mechanical properties of materials have recently seen significant growth, largely because of the availability of enabling technologies such as automated testing, full-field measurement techniques, and inexpensive computing resources. Unfortunately, as the complexity of the material systems and their associated behaviors increase, even the most advanced methods for inverse characterization require impractically large computation time to produce results. To overcome this limitation we present a method that employs Non-Uniform Rational B-spline (NURBs) based surrogate modeling to generate a very efficient representation of the combined constitutive and structural model response required for inverse characterization. Building on our previous work, we present an inversion method for identifying the constitutive material properties that minimize an appropriate objective function. Verification of this methodology is achieved through synthetic numerical experiments that include material systems of isotropic-elastic, orthotropic-elastic, and orthotropic-hyperelastic with damage nature on selected geometries. Statistical analyses on the effects of experimental noise supplement our analysis. We then proceed to demonstrate the use of this approach to characterize actual specimens tested using a multiaxial robotic system. In conclusion, we discuss the effectiveness and limitations of the surrogate model-based methodology and outline further research required to advance this approach.