18 results
Now showing1 - 10 of 18
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , MV-GEL: Language-driven multi-view geometric entity localization on meshesIdentifying and grounding precise geometric entities, such as edges, planar regions, and curved surfaces within 3D objects, is foundational to computer-aided design (CAD), robotic manipulation, and scientific simulation. Although modern Vision Language Models (VLMs) have advanced referring segmentation (RIS) in the image domain, extending such language-driven localization to structured 3D geometry is substantially harder. The 3D object appearance is highly sensitive to viewpoints; a single perspective may render a target entity clearly observable, while another may suffer from severe occlusion or foreshortening. In this work, we attempt to solve these challenges with MV-GEL (Multi-View Geometric Entity Localization), a framework for localizing fine-grained geometric entities on polygon meshes from natural language queries. Our key insight is that reliable CAD entity (i.e., faces, edges or solids) localization depends on selecting views that make the queried entity maximally interpretable. We introduce GELviews, a prompt-conditioned ranking module that prioritizes viewpoints based on language prompted observability of geometric CAD entities. Selected views are processed by a VLM-based reasoning segmentation backbone, and predicted masks are lifted to the corresponding meshes via geometry-aware ray casting. Our framework is completely CAD agnostic and relies only on 3D meshes. Experiments show up to a 1.7× improvement in face-level IoU and over 4.5× gains in edge-level F1 compared to vanilla baselines, substantially outperforming CLIP-based and random view sampling, particularly for thin and view-sensitive structures. The dataset, code and trained checkpoints are available at https://github.com/kbali1297/MV-GEL.Publicationtype: Conference Paper13 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Concurrently coupling particle and continuum simulations to study block copolymer membrane fabrication(2026-06-02); ; ; ; We present a concurrent multiscale simulation framework for membrane fabrication from a block copolymer solution via self-assembly and nonsolvent-induced phase separation, combining simulations of a soft, coarse-grained particle model with the continuum Uneyama-Doi model. The computationally intensive particle model provides a molecularly resolved description of micro- and macrophase separation, including thermal fluctuations, while the computationally efficient continuum model captures the process-driven self-assembly of nonequilibrium membrane morphologies on large length and time scales. Central to the approach is a machine learning-guided adaptive coupling strategy, implemented through a coordinator library, which predicts the evolving spatiotemporal subdomain where high-fidelity particle simulations are required and dynamically allocates computational resources accordingly. The two complementary models are concurrently coupled through a consistent exchange of the solvents' fluxes, enabling the treatment of spatially inhomogeneous, multicomponent systems with diffusive transport, micro- and macrophase separation, and vitrification. This adaptive strategy enables predictive simulations of membrane formation on experimentally and technologically relevant length and time scales, reaching micrometers and minutes. As an application, we examine the influence of polymer concentration in the initial casting solution. The framework is general and extensible, providing a computational tool for investigating nonequilibrium structure formation in complex multicomponent soft-matter systems.Publicationtype: Journal Article17 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , From model training to model raising(Association for Computing Machinery (ACM), 2026-01-28); ; ; A call to reform AI model-training paradigms from post hoc alignment to intrinsic, identity-based development. Current AI training methods align models with human values only after their core capabilities have been established, resulting in models that are easily misaligned and lack deep-rooted value systems. We propose a paradigm shift from “model training” to “model raising,” in which alignment is woven into a model’s development from the start. We identify several key components for this paradigm, all centered around redesigning the training corpus: reframing training data from a first-person perspective, recontextualizing information as lived experience, simulating social interactions, and scaffolding the ordering of training data. We expect that this redesign of the training corpus will lead to an early commitment to values from the first training token onward, such that knowledge, skills, and values are intrinsically much harder to separate. In an ecosystem in which large language model capabilities start overtaking human capabilities in many tasks, this seems to us like a critical need.Publicationtype: Journal Article32 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , 18 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Simulated annealing framework for generating representative volume elements of materials with complex ligamentous microstructuresAt the microscale, various materials from biological tissues to nanoporous metals are formed by networks of ligaments. Here we propose a highly efficient simulated annealing (SA) framework for generating synthetic representative volume elements (RVE) of such materials. It can produce RVE where the microstructural characteristics both on the network level (e.g., node valency and ligament length) and on the level of individual ligaments (e.g., curvature) can be predefined by the user via probability distributions. As an application example of our framework, we generate a large variety of RVEs, analyze their mechanical properties by the finite element method, and establish through this approach links between microstructural descriptors and macromechanical properties of materials with ligamentous microstructures.Publicationtype: Journal Article48 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , 21 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Constitutive artificial neural networks: a general anisotropic constitutive modeling framework utilizing machine learningIn this contribution, a novel machine learning architecture for data-driven modeling of the mechanical constitutive behavior of materials, constitutive artificial neural networks (CANNs) [1], will be introduced. CANNs incorporate basic material modeling fundamentals from continuum mechanics while relying on artificial neural networks for material-specific relations. Their architecture allows them to process stress-strain curves and arbitrary additional information (e.g., about the microstructure or manufacturing parameters). With only a low-to-moderate amount of training data and training time, they can predict the constitutive behavior of complex nonlinear and anisotropic materials. The ability to utilize additional material-specific information enables CANNs to predict the mechanical behavior of previously unseen materials if the CANN was sufficiently trained with many similar materials.Publicationtype: Conference Paper57 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Machine learning-augmented and microspectroscopy-informed multiparametric MRI for the non-invasive prediction of articular cartilage compositionBackground: Articular cartilage degeneration is the hallmark change of osteoarthritis, a severely disabling disease with high prevalence and considerable socioeconomic and individual burden. Early, potentially reversible cartilage degeneration is characterized by distinct changes in cartilage composition and ultrastructure, while the tissue's morphology remains largely unaltered. Hence, early degenerative changes may not be diagnosed by clinical standard diagnostic tools. Methods: Against this background, this study introduces a novel method to determine the tissue composition non-invasively. Our method involves quantitative MRI parameters (i.e., T , T , T and T maps), compositional reference measurements (i.e., microspectroscopically determined local proteoglycan [PG] and collagen [CO] contents) and machine learning techniques (i.e., artificial neural networks [ANNs] and multivariate linear models [MLMs]) on 17 histologically grossly intact human cartilage samples. Results: Accuracy and precision were higher in ANN-based predictions than in MLM-based predictions and moderate-to-strong correlations were found between measured and predicted compositional parameters. Conclusion: Once trained for the clinical setting, advanced machine learning techniques, in particular ANNs, may be used to non-invasively determine compositional features of cartilage based on quantitative MRI parameters with potential implications for the diagnosis of (early) degeneration and for the monitoring of therapeutic outcomes.Publicationtype: Journal ArticleScopus© Citations 13 66 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Computer-controlled biaxial bioreactor for investigating cell-mediated homeostasis in tissue equivalentsSoft biological tissues consist of cells and extracellular matrix (ECM), a network of diverse proteins, glycoproteins, and glycosaminoglycans that surround the cells. The cells actively sense the surrounding ECM and regulate its mechanical state. Cell-seeded collagen or fibrin gels, so-called tissue equivalents, are simple but powerful model systems to study this phenomenon. Nevertheless, few quantitative studies document the stresses that cells establish and maintain in such gels; moreover, most prior data were collected via uniaxial experiments whereas soft tissues are mainly subject to multiaxial loading in vivo. To begin to close this gap between existing experimental data and in vivo conditions, we describe here a computer-controlled bioreactor that enables accurate measurements of the evolution of mechanical tension and deformation of tissue equivalents under well-controlled biaxial loads. This device allows diverse studies, including how cells establish a homeostatic state of biaxial stress and if they maintain it in response to mechanical perturbations. It similarly allows, for example, studies of the impact of cell and matrix density, exogenous growth factors and cytokines, and different types of loading conditions (uniaxial, strip-biaxial, and biaxial) on these processes. As illustrative results, we show that NIH/3T3 fibroblasts establish a homeostatic mechanical state that depends on cell density and collagen concentration. Following perturbations from this homeostatic state, the cells were able to recover biaxial loading similar to homeostatic. Depending on the precise loads, however, they were not always able to fully maintain that state.Publicationtype: Journal ArticleScopus© Citations 14 100 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , The accuracy of laser flash analysis explored by finite element method and numerical fittingLaser flash analysis (LFA) has become over the last decades a widely used standard technique to measure the thermal diffusivity of bulk materials under various conditions like different gases, atmospheric pressures, and temperatures. A curve fitting procedure forms the heart of LFA. This procedure bases on a mathematical model that should ideally account for inherent shortcomings of the experimental realization such as: duration of the heating pulse, heat losses to the environment and sample holder, non-opaque samples, and radiative heat transfer. The accuracy of the mathematical model and curve fitting algorithm underlying LFA defines an upper bound of the accuracy of LFA in general. Unfortunately, not much is known about the range of parameters and conditions for which this accuracy is acceptable. In this paper, we examine the limits of accuracy of LFA resulting from its underlying computational framework. To this end, we developed a particularly accurate and comprehensive computational framework and applied it to data from simulated experiments. We quantify the impact of different (simulated) experimental conditions on the accuracy of the results by comparing the fit results of our computational framework to the known simulation input parameters. This way we demonstrate that a state-of-the-art computational framework for LFA admits determining thermal conductivities of materials in a broad range from at least 0.16 W/mK to 238 W/mK with relative errors typically well below 4% even in the presence of common undesired experimental side effects.Publicationtype: Journal Article212