Regularized least squares for the building of a sizing criterion based on damage mechanics

Abstract

mance and efficiency is evaluated both for training and validation data using error measures based on the residuals. Preliminary results showed acceptable accuracy for the linear least squares and Multi-Stage Least Squares methods. An important key aspect of VTA is that off-line computations for the advanced laminates can be stored and used for different full-scale structures as a simple failure criterion.

Design of Artificial Neural Networks for Damage Estimation of Composite Laminates: Application to Delamination Failures in Ply Drops

“…This section presents a general overview of the finite element modeling and the DCE conducted by Friderikos et al [3] since the dataset obtained therein are re-purposed for the present analysis.…”

“…As such, the design space to be explored is defined by the nine design variables θ i , t i and u i for i = 1, 2, 3. The methodology employed by Friderikos et al [3] consists in sampling each of the three sub-spaces independently and subsequently generating all the possible combinations of variables. First, the design space for the fiber orientations angles is limited to θ i ∈ [0, 90]…”

“…In a previous work by Friderikos et al [3], a data driven approach combining regularized regression models and finite element analysis was developed to minimize the exhaustive experimental testing as well as to control the complexity of damage patterns of thermoplastic QSP composite laminates. This methodology was shown to be able to predict the damage evolution of internal ply drop laminates.…”

“…To this end, high nonlinear FEM simulations have been utilized for the prediction of the complex failure modes. As stated by Friderikos et al [3], even though only delamination failure is taken into account, other interlaminar mechanisms described at the meso-scale can be incorporated into the FEM model as well. Moreover, the proposed methodology therein can be applied to other critical laminate substructures, i.e., skin-stiffener joints, external ply drops, corners, etc.…”

“…In this work, a data driven approach has been developed using the dataset derived from nonlinear FE analysis of ply-drop laminates provided by Friderikos et al [3]. The aim is to construct a metamodel by implementing ANN as they become relatively competitive to conventional regression models for high nonlinear applications, multi-dimensional and big data.…”