Detecting Multi-Scale Defects in Material Extrusion Additive Manufacturing of Fiber-Reinforced Thermoplastic Composites: A Review of Challenges and Advanced Non-Destructive Testing Techniques

Polymers, Год журнала: 2024, Номер 16(21), С. 2986 - 2986

Опубликована: Окт. 24, 2024

Additive manufacturing (AM) defects present significant challenges in fiber-reinforced thermoplastic composites (FRTPCs), directly impacting both their structural and non-structural performance. In structures produced through material extrusion-based AM, specifically fused filament fabrication (FFF), the layer-by-layer deposition can introduce such as porosity (up to 10-15% some cases), delamination, voids, fiber misalignment, incomplete fusion between layers. These compromise mechanical properties, leading reduction of up 30% tensile strength and, cases, 20% fatigue life, severely diminishing composite's overall performance integrity. Conventional non-destructive testing (NDT) techniques often struggle detect multi-scale efficiently, especially when resolution, penetration depth, or heterogeneity pose challenges. This review critically examines FRTPCs, classifying FFF-induced based on morphology, location, size. Advanced NDT techniques, micro-computed tomography (micro-CT), which is capable detecting voids smaller than 10 µm, health monitoring (SHM) systems integrated with self-sensing fibers, are discussed. The role machine-learning (ML) algorithms enhancing sensitivity reliability methods also highlighted, showing that ML integration improve defect detection by 25-30% compared traditional techniques. Finally, potential self-reporting equipped continuous fibers for real-time situ SHM, investigated. By integrating ML-enhanced accuracy efficiency be significantly improved, fostering broader adoption AM aerospace applications enabling production more reliable, defect-minimized FRTPC components.

Язык: Английский

---

Prediction of microstructural-dependent mechanical properties, progressive damage, and stress distribution from X-ray computed tomography scans using a deep learning workflow

Computer Methods in Applied Mechanics and Engineering, Год журнала: 2024, Номер 424, С. 116878 - 116878

Опубликована: Март 11, 2024

Creating computationally efficient models that link processing methods, material structures, and properties is essential for the development of new materials. Translating microstructural details to macro-level mechanical often proves be an arduous challenge. This paper introduces a novel deep learning-based framework predict 3D stress fields, behavior, progressive damage in ceramic materials informed by features material. We construct dataset synthetic representative volume elements utilizing X-ray computed tomography scans employ automated finite element (FE) modeling approach generate datasets alumina ceramics with varying inclusion morphologies. The learning model, U-Net based convolutional neural network (CNN), trained understand structure-property linkages responses directly from FE-generated data without transforming them into image format. CNN's architecture optimized capturing both local global contextual information data, enabling accurate prediction fields evolution. Inclusions within are shown play crucial role initiation propagation damage. CNN model demonstrated robust performance predicting field, stress-strain curve, training test showing high consistent similarity between predictions ground truth. Overall, this research offers generalized can adapted different structures toward creating digital replicas optimizing real-world applications.

Язык: Английский

---

Advancing Additive Manufacturing through Deep Learning: A Comprehensive Review of Current Progress and Future Challenges

IISE Transactions, Год журнала: 2024, Номер unknown, С. 1 - 44

Опубликована: Дек. 19, 2024

This paper presents the first comprehensive literature review of deep learning (DL) applications in additive manufacturing (AM). It addresses need for a thorough analysis this rapidly growing yet scattered field, aiming to bring together existing knowledge and encourage further development. Our research questions cover three major areas AM: (i) design AM, (ii) AM modeling, (iii) monitoring control AM. We use step-by-step approach following Preferred Reporting Items Systematic Reviews Meta-Analyses (PRISMA) guidelines select papers from Scopus Web Science databases, aligning with our questions. include only those that implement DL across seven categories - binder jetting, directed energy deposition, material extrusion, powder bed fusion, sheet lamination, vat photopolymerization. reveals trend towards using generative models, such as adversarial networks, also highlights an increasing effort incorporate process physics into models improve modeling reduce data requirements. Additionally, there is interest 3D point cloud monitoring, alongside traditional 1D 2D formats. Finally, summarizes current challenges recommends some promising opportunities domain investigation special focus on generalizing wide range geometry types, managing uncertainties both overcoming limited, imbalanced, noisy issues by incorporating (iv) unveiling potential interpretable

Язык: Английский

---

Deep-Learning-Based Optimization of Non-Periodic Porous and Composite Materials for Aerospace Structures

AIAA SCITECH 2022 Forum, Год журнала: 2025, Номер unknown

Опубликована: Янв. 3, 2025

Язык: Английский

---

A high-throughput framework for predicting three-dimensional structural-mechanical relationships of human cranial bones using a deep learning-based method

Journal of the mechanical behavior of biomedical materials/Journal of mechanical behavior of biomedical materials, Год журнала: 2025, Номер 168, С. 107007 - 107007

Опубликована: Апрель 25, 2025

Язык: Английский

---

Virtual rapid prototyping of materials with deep learning: spatiotemporal stress fields prediction in ceramics employing convolutional neural networks and transfer learning

Virtual and Physical Prototyping, Год журнала: 2024, Номер 19(1)

Опубликована: Сен. 9, 2024

Язык: Английский

---

Data-driven integration of synthetic representative volume elements and machine learning for improved microstructure-property linkage and material performance in ceramics

Deleted Journal, Год журнала: 2024, Номер 4, С. 100011 - 100011

Опубликована: Авг. 23, 2024

Язык: Английский

---

Curing simulation and data-driven curing curve prediction of thermoset composites

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Дек. 30, 2024

Molding has been widely used to manufacture thermoset composite structures in the aerospace and automotive industries owing its efficiency reducing number of parts manufacturing cost. For such molded parts, degree-of-cure curve is generally evaluate solidification resin. Nevertheless, simulation cure not model itself, but rather knowing initial conditions as fiber volume fraction, curing degree, convective boundary etc. Additionally, solving heat transfer coupled with kinetics presents additional requirements for time, making artificial intelligence tools promising these problems. This paper focuses on developing a data-driven approach predicting curve. The simulated corresponds specific temperature-time was verified by published value. Then, resulting degree-of-cure-time curves obtained from finite element simulations were created training prediction models using machine learning approaches support vector regression (SVR), back propagation (BP) neural network BP optimized genetic algorithm (GA-BP). validation evaluation indices illustrate that trained GA-BP yields highest accuracy.

Язык: Английский

---