arXiv Science⌕ Search
Home/Archive/arXiv

arXiv · 2509.20011

Characterizing failure morphologies in fiber-reinforced composites via k-means clustering based multiscale framework

Harpreet Singh
View original source ↗Read source document ↗Export metadata ↓

Abstract

A novel homogenization methodology is proposed for analyzing the failure of fiber-reinforced composite materials, utilizing elastic and eigen influence tensors within a damage informed transformation field analysis (D-TFA) framework. This approach includes a technique for calculating macroscopic damage under uniform stress and strain conditions, offering more realistic simulations. Computational efficiency is enhanced through a reduced-order modeling strategy, while elastic and eigen strain distribution driven k-means clustering methods are employed to partition the microscale domain. The model's performance is assessed by simulating the response of a representative volume element (RVE) treated as a homogenized continuum. Subsequently, a comparative assessment is carried out to check the efficacy of two clustering schemes. Damage morphologies are calculated using proposed framework and compared with predictions obtained using finite element method. Furthermore, open-hole specimen tests are simulated and failure paths are predicted for the domains with different fiber layups. Ultimately, we show that D-TFA can accurately capture damage patterns and directional strengths, providing improved predictions of the mechanical behavior of composite materials. It has been demonstrated that higher cluster counts are crucial for capturing a more accurate stress-strain response, especially for complex microstructures.

Explore related subjects

cs.CEphysics.comp-ph

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Harpreet Singh. 2025-09-24. Characterizing failure morphologies in fiber-reinforced composites via k-means clustering based multiscale framework. https://doi.org/10.1016/j.compstruc.2025.107928

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Cross-Sectional Asset Retrieval via Future-Aligned Soft Contrastive Learning

Asset retrieval (finding similar assets in a financial universe) is central to quantitative investment decision-making. Existing approaches define similarity through historical price patterns or sector classifications, but such backward-looking criteria provide no guarantee about future behavior. We argue that effective asset retrieval should be future-aligned: the retrieved assets should be those most likely to exhibit correlated future returns. To this end, we propose Future-Aligned Soft Contrastive Learning (FASCL), a representation learning framework whose soft contrastive loss uses pairwise future return correlations as continuous supervision targets. We further introduce an evaluation protocol designed to directly assess whether retrieved assets share similar future trajectories. Experiments on 5,631 US-listed securities against 14 baselines show that FASCL attains the best future return correlation at every retrieval depth and the best rank information coefficient at every depth and horizon, leads on trend consistency in 13 of 16 cells, and gives the highest gross Sharpe ratio in a spread trading backtest at every basket size. Code is available at https://github.com/HyeongminLEE/fascl .

cs.CE↗
arXiv2026

On the Minimum Number of Control Laws for Nonlinear Systems with Input-Output Linearisation Singularities

We study how many distinct feedback-linearising control laws are needed for a single-input single-output nonlinear system whose exact input-output linearising law is singular on a nonempty subset of the state space. The question is posed as a coverage problem: a family of exact and approximate (Hauser-Sastry-Kokotovic) linearising laws covers a region if every state of the region lies in the validity domain of at least one law. Using transversality arguments we prove that, for $p$ laws whose singular hypersurfaces are in general position, the uncovered set is a finite union of embedded submanifolds of codimension exactly $p$. Hence $n+1$ laws always suffice for complete coverage of an $n$-dimensional state space, two laws already reduce the uncovered set to a thin set that generic trajectories never meet, and complete coverage with fewer than $n+1$ laws depends on specific intersections rather than on the number of components of the singularity manifold. The theory is applied exhaustively to the ball-and-beam system, whose exact linearising law is singular on the union of two transverse hypersurfaces $\{r=0\}\cup\{\dotθ=0\}$. We construct three explicit laws (the exact law and the two canonical Hauser-Sastry-Kokotovic approximations), compute their validity domains in closed form and prove that (i) every periodic tracking task forces the state through the singular set, so the exact law can never be used alone; (ii) a single approximate law covers the entire admissible region $|θ|<π/2$, and any two of the three laws cover $\mathbb{R}^4$ up to a codimension-two set; (iii) the three laws together cover $\mathbb{R}^4$ minus an explicit obstructed set on which the output is not linearisable by any static feedback, because the ball motion there is invariant under reversal of the input. Simulations confirm each statement.

cs.CE↗
arXiv2026

Globally Adaptive and Locally Regular Point Discretization of Implicit Surface Geometries

Point discretization of implicitly defined curved surfaces is required throughout computer-aided design, engineering analysis, and scientific computing. These applications often impose that surfaces are sampled with local regularity and global curvature adaptivity for unambiguous geometry representation. Finding a numerically well-conditioned point discretization, however, is not trivial, even for simple analytic surfaces. We present an algorithm for finding near-optimal surface point distributions governed by a prescribed resolution field on a curved surface. The algorithm works by approximately minimizing a global potential over local point-point interactions. We leverage a high-order meshfree level-set method to implicitly describe the surface and to perform all required projections without additional surface-attractive forces. To accelerate convergence, the algorithm dynamically fuses and inserts points where a local excess or lack is detected, for which it introduces an integral support measure. We test the proposed algorithm on a variety of shapes, ranging from parametric to non-parametric surfaces. We find point discretizations with different curvature adaptivity and low deviation from the prescribed target spacing. In all cases, the presented algorithm rapidly and robustly converges to the final number and distribution of surface points.

cs.CE↗