Análise inversa para identificação de parâmetros do modelo de drucker-prager em compacto verde de alumina

Abstract

Powder pressing followed by sintering is widely used in industry. In low-confinement stages, such as die ejection, handling, and green machining, the mechanical behavior of the green compact can be described by an elastoplastic constitutive model based on the Drucker-Prager (DP) yield criterion, incorporating isotropic hardening. This study investigated the identifiability of the linear DP model parameters with Voce hardening law, which include the elastic modulus (E), internal friction angle (β), yield stress (σy), saturation stress (σ∞), and hardening parameter (b), using cyclic Brazilian test data on alumina green compacts, adopting an associative plastic flow hypothesis (ψ = β). The Finite Element Model Updating (FEMU) technique was employed in two stages: a numerical sensitivity analysis based on synthetic data and a force-based experimental application (FEMU-F) using experimental data of three Brazilian tests available in the literature. The numerical analysis demonstrated that global force data enables the identification of E and β with high sensitivity but are unable to uniquely constrain the hardening parameters due to a critical coupling between σy, σ∞, and b, creating a valley of indistinguishable solutions. The incorporation of displacement fields in the combined approach (FEMU-UF) proved capable of mitigating this ill-conditioning, numerically validating the cyclic Brazilian test as a viable setup for elastoplastic characterization. In the experimental application, E and β were identified with a variability of less than 4%, while σy exhibited a variability of 38%, experimentally confirming the coupling predicted by the Hessian eigenvalue decomposition. Despite this parametric instability, a generalized model using average parameters reproduced the force-time curves with less than a 7% deviation compared to the individual fits. In conclusion, the model reproduces faithfully the mechanical response of the Brazilian test. Nevertheless, its global predictive capacity remains limited to the specific test conditions and the adopted hypothesis of associative flow.