Answers are grounded in this paper and cite their sources.
Computational modeling study demonstrates robust 6D failure predictions in orthotropic composites, highlighting viable data-driven modeling for thick shell and solid finite elements.
A new data-driven failure criterion has been developed to improve modeling of orthotropic structural composites. Rather than using an analytical expression traditionally employed to predict failure, a point cloud failure surface is constructed in the stress/strain space using a combination of laboratory testing and virtual testing with multiscale models based on representative volume elements. In the current work, the failure predictions are extended from a three-dimensional stress space that is suitable for use with thin shell finite elements to a six-dimensional stress space, suitable for use with thick shell and solid finite elements. A crucial step in using the material characterization failure data is the ability of a method to correctly predict when failure onset in a finite element takes place. This predictive capability needs to balance efficiency and accuracy. Three predictive methods are compared in this paper: approximate nearest neighbor, neural network, and a new custom exact nearest neighbor that was developed specifically to be accurate and efficient in low-dimensional search space. Point cloud data from a unidirectional composite, the T800-F3900, commonly used for aerospace applications, are used for comparative evaluation of these methods. Results indicate that the custom exact nearest neighbor and neural network implementations are robust, efficient, and accurate.
No takes yet. Share an insight, caveat, or question.
Maurya et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: