On-line Nonlinear Characterization of Anisotropic Materials
This dissertation proposes a new framework to characterize the nonlinear behavior of anisotropic materials in an on-line manner. The proposed framework applies recursive estimation and a multi-linear model to characterize the nonlinear behavior of anisotropic materials on-line using full-field strains, which are capable of capturing the multi-axial information of anisotropic materials.
A stochastic method is developed to characterize the linear behavior of anisotropic materials under the influence of full-field strain measurement noise. This method first derives stochastic equations based on the formulas of energy-based characterization that utilizes the principle of ener-gy conservation, and then recursively estimates elastic constants at every acquisition of measure-ment using a Kalman filter (KF). Since the measurement model is expressed nonlinearly, the KF utilizes a Kalman gain, which is newly derived in this dissertation through variance minimization, to achieve optimal characterization. The aforementioned method, namely stochastic linear characteri-zation in this dissertation, becomes a basis of the multi-linear characterization method. This method utilizes a multi-linear model, which is defined by partitions, to characterize the nonlinear constitu-tive relations. The multi-linear characterization scales up the number of estimates and identifies the coefficients of each linear partition using the previously derived KF. The recursive updates in measurements not only removes uncertainty through sensor measurements, but also enables the on-line capability of the nonlinear characterization of anisotropic materials.
A series of numerical and experimental studies were performed to demonstrate the performance of the proposed framework in characterizing the nonlinear behavior of anisotropic materials. The validity and applicability of the proposed framework were confirmed by the comparison with the known values of the characterized constitutive relations. It was found that the proposed framework identified elastic constants that were in good agreement with known values irrespective of the spec-imen geometry. The results of the multi-linear characterization method were well correlated with known nonlinear stress-strain relations and concluded that the proposed framework is capable of characterizing adequate nonlinear behavior on-line.