Deriving Ogden Constitutive Parameters from Biaxial Stretch Testing Data
Deriving Ogden constitutive parameters pairs non-linear biaxial stretch data with Drucker stability constraints to ensure reliable finite element simulations.
Levenberg Marquardt Optimization
Meaning
Non-linear least-squares numerical regression solves non-convex optimization problems by interpolating between gradient descent and Gauss-Newton methods. Applying levenberg marquardt optimization fits hyperelastic constitutive model parameters to empirical stress-strain test datasets derived from elastomer tensile tests. The iterative solver handles moderately non-linear objective functions with robust convergence characteristics, but struggles when parameter spaces exhibit multiple local minima or ill-conditioned Jacobian matrices.
Algorithm Mechanics
The algorithm adjusts step size dynamically using a damping parameter that shifts search direction based on current residual error slope. Performing levenberg marquardt optimization starts with high damping for safe gradient descent steps when far from the solution, transitioning to rapid Gauss-Newton convergence near the minimum. Parameter updates continuously balance convergence speed against numerical stability during complex non-linear curve fitting.
Improper initial parameter guesses cause the solver to get trapped in local energy minima, yielding non-physical material constants.
Objective Function
Residual minimization targets the sum of squared differences between experimental stress points and strain energy model predictions across multiple deformation modes. Weighting factors prevent high-strain stress values from dominating the sum and degrading low-strain predictions.
Parameter Bounds
Constraining optimization variables prevents parameters from shifting into unphysical negative regimes during iteration cycles. Unbounded parameter searches frequently generate unstable strain energy potentials that fail Drucker stability criteria.