Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105031
Chuxiao Meng , Conor Porter , Sina Malakpour Estalaki , Seongyeon Yang , Garrett Mathesen , Jian Cao
Pore formation in Laser Powder Bed Fusion (L-PBF) significantly impacts mechanical properties, often caused by unstable keyhole collapse due to excessive laser energy density. Traditional ex-situ CT scanning for pore detection is time-consuming, costly, and often limited in size. We present a rapid pore detection method, predicting pore counts in L-PBF-fabricated samples efficiently and non-destructively. Our approach employs a peak detection algorithm for coaxial photodiode melt pool monitoring (MPM) signals to identify keyhole collapse patterns, precursors to keyhole pore formation. By smoothing MPM signals with cubic polynomials and applying interval-based thresholding, we identify peaks precisely. Validated against CT scans, the method shows strong correlation (R² = 0.95) between MPM peaks and pore counts across eight samples, with an average prediction time of 36 s for a 2400 mm³ sample. The method works best in the high energy density, unstable keyhole regime, and overpredicts pores in the near-optimal regime. This highly scalable, cost-effective solution outperforms prior pores detection techniques in speed and simplicity, with high potential for industrial applicability, particularly if more attempts from diverse research groups or industrial applications are shared to advance the overall coverage of material types, machine types and part geometry.
{"title":"Part-scale keyhole pore detection in laser powder bed fusion using coaxial photodiodes","authors":"Chuxiao Meng , Conor Porter , Sina Malakpour Estalaki , Seongyeon Yang , Garrett Mathesen , Jian Cao","doi":"10.1016/j.addma.2025.105031","DOIUrl":"10.1016/j.addma.2025.105031","url":null,"abstract":"<div><div>Pore formation in Laser Powder Bed Fusion (<span>L</span>-PBF) significantly impacts mechanical properties, often caused by unstable keyhole collapse due to excessive laser energy density. Traditional ex-situ CT scanning for pore detection is time-consuming, costly, and often limited in size. We present a rapid pore detection method, predicting pore counts in <span>L</span>-PBF-fabricated samples efficiently and non-destructively. Our approach employs a peak detection algorithm for coaxial photodiode melt pool monitoring (MPM) signals to identify keyhole collapse patterns, precursors to keyhole pore formation. By smoothing MPM signals with cubic polynomials and applying interval-based thresholding, we identify peaks precisely. Validated against CT scans, the method shows strong correlation (R² = 0.95) between MPM peaks and pore counts across eight samples, with an average prediction time of 36 s for a 2400 mm³ sample. The method works best in the high energy density, unstable keyhole regime, and overpredicts pores in the near-optimal regime. This highly scalable, cost-effective solution outperforms prior pores detection techniques in speed and simplicity, with high potential for industrial applicability, particularly if more attempts from diverse research groups or industrial applications are shared to advance the overall coverage of material types, machine types and part geometry.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105031"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145616194","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105025
Junfang Qi , Xue Liu , Peng Dong , Quan Li , Yabin Yang
This study proposes a novel inherent strain approach, termed the Improved Dynamic Inherent Strain (IDIS) method, for efficient and accurate prediction of residual stresses and deformations in metal additive manufacturing (MAM). Building upon the previously developed Dynamic Inherent Strain (DIS) method, the IDIS method incorporates a more comprehensive treatment of three-dimensional stress/strain interactions based on the theory of total plasticity. The accuracy of the IDIS model is evaluated through two benchmark experiments reported in the literature: a wire and arc additive manufacturing (WAAM) process with single-track deposition per layer, and a standard laser powder bed fusion (LPBF) test. Results show that both the DIS and IDIS models achieve good accuracy in deformation prediction, with a maximum error of approximately 5 %. Compared to the DIS model, the IDIS model significantly improves stress prediction accuracy, particularly in capturing stress distribution trends more precisely. The study also examines the influence of different layer activation thickness (LAT) on prediction accuracy in LPBF simulations. It is found that a larger LAT yields higher stress prediction accuracy at the bottom of the part, while a smaller LAT provides better accuracy at the top. A mixed-LAT technique is proposed, which achieves high accuracy at both the bottom and top of the part while improving computational efficiency by 99.5 % without compromising accuracy. This work offers an effective and practical framework for rapid residual stress and deformation prediction in large-scale MAM components.
{"title":"An improved dynamic inherent strain method for efficient prediction of residual stresses and deformations in metal additive manufacturing","authors":"Junfang Qi , Xue Liu , Peng Dong , Quan Li , Yabin Yang","doi":"10.1016/j.addma.2025.105025","DOIUrl":"10.1016/j.addma.2025.105025","url":null,"abstract":"<div><div>This study proposes a novel inherent strain approach, termed the Improved Dynamic Inherent Strain (IDIS) method, for efficient and accurate prediction of residual stresses and deformations in metal additive manufacturing (MAM). Building upon the previously developed Dynamic Inherent Strain (DIS) method, the IDIS method incorporates a more comprehensive treatment of three-dimensional stress/strain interactions based on the theory of total plasticity. The accuracy of the IDIS model is evaluated through two benchmark experiments reported in the literature: a wire and arc additive manufacturing (WAAM) process with single-track deposition per layer, and a standard laser powder bed fusion (LPBF) test. Results show that both the DIS and IDIS models achieve good accuracy in deformation prediction, with a maximum error of approximately 5 %. Compared to the DIS model, the IDIS model significantly improves stress prediction accuracy, particularly in capturing stress distribution trends more precisely. The study also examines the influence of different layer activation thickness (LAT) on prediction accuracy in LPBF simulations. It is found that a larger LAT yields higher stress prediction accuracy at the bottom of the part, while a smaller LAT provides better accuracy at the top. A mixed-LAT technique is proposed, which achieves high accuracy at both the bottom and top of the part while improving computational efficiency by 99.5 % without compromising accuracy. This work offers an effective and practical framework for rapid residual stress and deformation prediction in large-scale MAM components.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105025"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145616191","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105041
H. Yeung
Laser-Based Powder Bed Fusion (LB-PBF) is a widely used additive manufacturing technique known for its ability to produce intricate geometries, optimized structures, and lightweight designs. However, its broader industrial adoption is hindered by persistent quality issues arising from inadequate process control. Conventional LB-PBF systems rely on predefined laser power settings for each scan vector, which do not account for dynamic thermal variations caused by scan sequence, part geometry, and localized heat accumulation. This limitation often leads to defects in resulting parts such as lack-of-fusion in underheated regions or keyhole porosity in overheated areas. This study explores the feasibility of real-time, in-line laser power control using in-situ monitoring of the melt pool area as feedback. The proposed method continuously adjusts laser power within approximately 118 μs of measurement acquisition, ensuring corrections are made fast enough to potentially alleviate disturbances or instabilities in the melt pool. Experiments conducted on nickel superalloy 625 plates under various laser process conditions validate the effectiveness of this technique. The results demonstrate that real-time feedback control significantly improves process stability, laying the foundation for future advancements in adaptive additive manufacturing and defect mitigation strategies.
{"title":"Toward real-time feedback control for powder bed fusion additive manufacturing","authors":"H. Yeung","doi":"10.1016/j.addma.2025.105041","DOIUrl":"10.1016/j.addma.2025.105041","url":null,"abstract":"<div><div>Laser-Based Powder Bed Fusion (LB-PBF) is a widely used additive manufacturing technique known for its ability to produce intricate geometries, optimized structures, and lightweight designs. However, its broader industrial adoption is hindered by persistent quality issues arising from inadequate process control. Conventional LB-PBF systems rely on predefined laser power settings for each scan vector, which do not account for dynamic thermal variations caused by scan sequence, part geometry, and localized heat accumulation. This limitation often leads to defects in resulting parts such as lack-of-fusion in underheated regions or keyhole porosity in overheated areas. This study explores the feasibility of real-time, in-line laser power control using in-situ monitoring of the melt pool area as feedback. The proposed method continuously adjusts laser power within approximately 118 μs of measurement acquisition, ensuring corrections are made fast enough to potentially alleviate disturbances or instabilities in the melt pool. Experiments conducted on nickel superalloy 625 plates under various laser process conditions validate the effectiveness of this technique. The results demonstrate that real-time feedback control significantly improves process stability, laying the foundation for future advancements in adaptive additive manufacturing and defect mitigation strategies.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105041"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145681661","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105008
Jinsi Yuan , Bowen Hu , Peng Cai , Jinxing Sun , Xiaoteng Chen , Haijiang Wang , Xiewen Wen , Jiaming Bai
Multi-material 3D printing holds transformative potential for fabricating complex functional components, yet current vat photopolymerization (VPP) techniques remain limited in material compatibility. Resin tank switching-based VPP methods are restricted to low-viscosity resins, while hybrid strategies integrating Direct Ink Writing with VPP enable the printing of high-viscosity pastes. Nevertheless, these strategies remain insufficient to accommodate the broad spectrum of material viscosities required for diverse multi-material printing applications. Here, we introduce a novel top-down multi-material VPP technique based on the liquid surface supported printing (LSSP) method, which exhibits broad slurry compatibility. The LSSP system accommodates a wide range of materials, including low-viscosity hydrogels, high-viscosity resins, and ceramic slurries. Moreover, it enables continuous gradient material printing—a capability unattainable with conventional VPP. By overcoming limitations in material adaptability and gradient structure fabrication, the LSSP system opens new avenues for manufacturing high-performance, multi-material, and functionally gradient structures.
{"title":"Multi-material top-down vat photopolymerization 3D printing based on liquid surface supported printing","authors":"Jinsi Yuan , Bowen Hu , Peng Cai , Jinxing Sun , Xiaoteng Chen , Haijiang Wang , Xiewen Wen , Jiaming Bai","doi":"10.1016/j.addma.2025.105008","DOIUrl":"10.1016/j.addma.2025.105008","url":null,"abstract":"<div><div>Multi-material 3D printing holds transformative potential for fabricating complex functional components, yet current vat photopolymerization (VPP) techniques remain limited in material compatibility. Resin tank switching-based VPP methods are restricted to low-viscosity resins, while hybrid strategies integrating Direct Ink Writing with VPP enable the printing of high-viscosity pastes. Nevertheless, these strategies remain insufficient to accommodate the broad spectrum of material viscosities required for diverse multi-material printing applications. Here, we introduce a novel top-down multi-material VPP technique based on the liquid surface supported printing (LSSP) method, which exhibits broad slurry compatibility. The LSSP system accommodates a wide range of materials, including low-viscosity hydrogels, high-viscosity resins, and ceramic slurries. Moreover, it enables continuous gradient material printing—a capability unattainable with conventional VPP. By overcoming limitations in material adaptability and gradient structure fabrication, the LSSP system opens new avenues for manufacturing high-performance, multi-material, and functionally gradient structures.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105008"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570886","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105022
Dmitry Chernyavsky , Denys Y. Kononenko , Julia Kristin Hufenbach , Jeroen van den Brink , Konrad Kosiba
Identifying optimal processing parameters remains a major challenge in additive manufacturing (AM), limiting its potential and broader industrial adoption. In this work, we present a Bayesian machine learning (ML) framework designed to efficiently determine optimal parameters for AM processes. We demonstrate its effectiveness through the successful processing of the AA2024 alloy into high-density components, known for its difficulty in processing, using laser powder bed fusion (PBF-LB/M). Our approach begins with Bayesian Optimization (BO) applied to an initial dataset containing only five processing parameter sets. Despite the limited data, the method accurately predicts conditions for producing crack-free components with a remarkably high density resulting in tensile properties similar to cast counterparts. We further extend the framework to perform bi-objective optimization, targeting both maximum build-up rate (BUR) and density. Experimental validation confirms that the framework can identify new parameter sets that significantly enhance BUR while maintaining high part quality. This work underscores the potential of BO strategies for accelerating optimal processing conditions discovery, especially for challenging materials and multi-objective scenarios.
{"title":"Bayesian optimization for laser powder bed fusion of defect-free AA2024","authors":"Dmitry Chernyavsky , Denys Y. Kononenko , Julia Kristin Hufenbach , Jeroen van den Brink , Konrad Kosiba","doi":"10.1016/j.addma.2025.105022","DOIUrl":"10.1016/j.addma.2025.105022","url":null,"abstract":"<div><div>Identifying optimal processing parameters remains a major challenge in additive manufacturing (AM), limiting its potential and broader industrial adoption. In this work, we present a Bayesian machine learning (ML) framework designed to efficiently determine optimal parameters for AM processes. We demonstrate its effectiveness through the successful processing of the AA2024 alloy into high-density components, known for its difficulty in processing, using laser powder bed fusion (PBF-LB/M). Our approach begins with Bayesian Optimization (BO) applied to an initial dataset containing only five processing parameter sets. Despite the limited data, the method accurately predicts conditions for producing crack-free components with a remarkably high density resulting in tensile properties similar to cast counterparts. We further extend the framework to perform bi-objective optimization, targeting both maximum build-up rate (BUR) and density. Experimental validation confirms that the framework can identify new parameter sets that significantly enhance BUR while maintaining high part quality. This work underscores the potential of BO strategies for accelerating optimal processing conditions discovery, especially for challenging materials and multi-objective scenarios.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105022"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570884","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This study was undertaken to address dominant microstructural factors in the melt-pool structure formed in powder bed fusion using laser beam (PBF-LB) for high strength and the mechanical behavior of Al-2.5 %Fe binary alloy samples additive-manufactured via PBF-LB. Single-crystal micropillars with a mean diameter of approximately 2 μm were fabricated at different regions of the melt-pool structure (melt-pool inside: MPI, melt-pool boundary: MPB), and compression tests were performed at various initial strain rates controlled by loading rate. The PBF-LB process produced numerous Al6Fe metastable phases (with a few tens of nanometers in size) distributed in the α-Al supersaturated solid solutions containing highly concentrated Fe in a large part of the melt-pool inside (MPI). A relatively coarsened microstructure with a thickness of approximately 3 µm (composed of many granular α-Al phase with a few hundred nanometer size surrounded by Al6Fe-phase particles) was localized at the melt-pool boundary (MPB) region. The MPI micropillars exhibited a higher 0.2 % proof stress than the MPB micropillars, whereas the following strain hardening appeared similar due to the activation of multiple slip systems. The MPI micropillars exhibit almost the same mechanical behavior after the 300 ℃ annealing for the formation of nanoscale precipitates consuming solute Fe, indicating that the refined Al6Fe phase is a dominant contributor to the strengthening by the PBF-LB process. Intriguingly, MPI micropillars exhibited a negative strain rate sensitivity of flow stress at an early stage of deformation, whereas the negative strain rate sensitivity transitioned to positive after annealing at 300 ℃. TEM characterization revealed the dynamic precipitation of nanoscale Fe-rich precipitates inside the α-Al(fcc) matrix with high solute Fe contents even at ambient temperature. The nanoscale precipitates interact with introduced dislocations, resulting in the enhanced flow stress at an early stage of plastic deformation. The present study provided new insights into a local variation in strain-rate dependent strength depending on the location of the melt pool structure in Al alloys processed via PBF-LB, in terms of solute alloy-element contents (driving force for the dynamic precipitation) controlled by laser conditions of the PBF-LB process.
{"title":"Microstructural factors in melt-pool structure for mechanical behavior of Al-Fe alloy manufactured by laser-beam powder bed fusion: Single-crystal micropillar compression test approach","authors":"Dasom Kim , Akihiro Choshi , Yuhki Tsukada , Naoki Takata","doi":"10.1016/j.addma.2025.105035","DOIUrl":"10.1016/j.addma.2025.105035","url":null,"abstract":"<div><div>This study was undertaken to address dominant microstructural factors in the melt-pool structure formed in powder bed fusion using laser beam (PBF-LB) for high strength and the mechanical behavior of Al-2.5 %Fe binary alloy samples additive-manufactured via PBF-LB. Single-crystal micropillars with a mean diameter of approximately 2 μm were fabricated at different regions of the melt-pool structure (melt-pool inside: MPI, melt-pool boundary: MPB), and compression tests were performed at various initial strain rates controlled by loading rate. The PBF-LB process produced numerous Al<sub>6</sub>Fe metastable phases (with a few tens of nanometers in size) distributed in the α-Al supersaturated solid solutions containing highly concentrated Fe in a large part of the melt-pool inside (MPI). A relatively coarsened microstructure with a thickness of approximately 3 µm (composed of many granular α-Al phase with a few hundred nanometer size surrounded by Al<sub>6</sub>Fe-phase particles) was localized at the melt-pool boundary (MPB) region. The MPI micropillars exhibited a higher 0.2 % proof stress than the MPB micropillars, whereas the following strain hardening appeared similar due to the activation of multiple slip systems. The MPI micropillars exhibit almost the same mechanical behavior after the 300 ℃ annealing for the formation of nanoscale precipitates consuming solute Fe, indicating that the refined Al<sub>6</sub>Fe phase is a dominant contributor to the strengthening by the PBF-LB process. Intriguingly, MPI micropillars exhibited a negative strain rate sensitivity of flow stress at an early stage of deformation, whereas the negative strain rate sensitivity transitioned to positive after annealing at 300 ℃. TEM characterization revealed the dynamic precipitation of nanoscale Fe-rich precipitates inside the α-Al(fcc) matrix with high solute Fe contents even at ambient temperature. The nanoscale precipitates interact with introduced dislocations, resulting in the enhanced flow stress at an early stage of plastic deformation. The present study provided new insights into a local variation in strain-rate dependent strength depending on the location of the melt pool structure in Al alloys processed via PBF-LB, in terms of solute alloy-element contents (driving force for the dynamic precipitation) controlled by laser conditions of the PBF-LB process.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105035"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145616193","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105029
Xingjie Yang, Ping Chen, Zhikun Liu
Laser direct writing (LDW) of fully single-crystalline silicon nanostructures has been successfully demonstrated for the first time, using liquid-phase cyclohexasilane (CHS) as the precursor. The fully epitaxial, single-crystal nature of these nanostructures - including vertical pillars and horizontal wires - is confirmed by transmission electron microscopy (TEM), which reveals seamless lattice continuity with the substrate and identical diffraction patterns across various substrate orientations. This maskless, additive process yields subwavelength features - pillar and wire widths smaller than the laser wavelength - and supports vertical growth of pillar at the rate of 2900 nm/s. CHS proves to be a highly effective silicon precursor for LDW due to its low thermal decomposition temperature (∼200–300 °C) and higher silicon content (Si:H = 1:2 vs 1:4 for SiH4), which together enable rapid deposition at relatively low temperatures. Crucially, the use of tightly focused, pulsed laser irradiation provides localized, transient heating that confines the reaction to the liquid-substrate interface. This local energy delivery promotes surface-limited precursor decomposition and epitaxial growth while suppressing bulk nucleation, thereby yielding continuous single-crystal silicon structures. This work establishes a foundational methodology for the direct, on-demand additive manufacturing of high-quality single-crystal silicon nanostructures, opening new pathways for creating complex, high-performance semiconductor devices.
以液相环己硅烷（CHS）为前驱体，首次成功地实现了全单晶硅纳米结构的激光直接写入。透射电子显微镜（TEM）证实了这些纳米结构（包括垂直柱和水平线）的完全外延，单晶性质，揭示了与衬底无缝的晶格连续性和不同衬底方向上相同的衍射图案。这种无掩模的增材工艺产生了亚波长特性——柱和线的宽度小于激光波长——并支持柱的垂直生长速度为2900 nm/s。CHS被证明是LDW的高效硅前驱体，因为它的热分解温度低（~ 200-300°C）和硅含量高（Si:H = 1:2 vs SiH4的1:4），两者共同使在相对较低的温度下快速沉积。至关重要的是，使用紧密聚焦的脉冲激光照射提供了局部的瞬态加热，将反应限制在液体-衬底界面。这种局部能量传递促进表面受限前驱体分解和外延生长，同时抑制体成核，从而产生连续的单晶硅结构。这项工作为高质量单晶硅纳米结构的直接、按需增材制造建立了基础方法，为创造复杂、高性能的半导体器件开辟了新的途径。
{"title":"Laser direct writing of single-crystal silicon nanostructures from liquid cyclohexasilane","authors":"Xingjie Yang, Ping Chen, Zhikun Liu","doi":"10.1016/j.addma.2025.105029","DOIUrl":"10.1016/j.addma.2025.105029","url":null,"abstract":"<div><div>Laser direct writing (LDW) of fully single-crystalline silicon nanostructures has been successfully demonstrated for the first time, using liquid-phase cyclohexasilane (CHS) as the precursor. The fully epitaxial, single-crystal nature of these nanostructures - including vertical pillars and horizontal wires - is confirmed by transmission electron microscopy (TEM), which reveals seamless lattice continuity with the substrate and identical diffraction patterns across various substrate orientations. This maskless, additive process yields subwavelength features - pillar and wire widths smaller than the laser wavelength - and supports vertical growth of pillar at the rate of 2900 nm/s. CHS proves to be a highly effective silicon precursor for LDW due to its low thermal decomposition temperature (∼200–300 °C) and higher silicon content (Si:H = 1:2 vs 1:4 for SiH<sub>4</sub>), which together enable rapid deposition at relatively low temperatures. Crucially, the use of tightly focused, pulsed laser irradiation provides localized, transient heating that confines the reaction to the liquid-substrate interface. This local energy delivery promotes surface-limited precursor decomposition and epitaxial growth while suppressing bulk nucleation, thereby yielding continuous single-crystal silicon structures. This work establishes a foundational methodology for the direct, on-demand additive manufacturing of high-quality single-crystal silicon nanostructures, opening new pathways for creating complex, high-performance semiconductor devices.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105029"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145536895","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105028
Ali Kassab , Meriam Chelbi , Sajad Shirzad , Christopher Pannier , Pravansu Mohanty , Georges Ayoub
Predicting shrinkage and deformation during sintering in metal additive manufacturing (AM) is challenging due to complex interactions among thermal gradients, material properties, and part geometry. Finite element methods (FEM) provide high-fidelity simulations but are computationally intensive, limiting their use for real-time control and design iteration. To address this, we present a hybrid deep learning framework that couples a physics-encoded recurrent neural network (RNN) with a physics-informed graph neural network (GNN) for efficient and accurate sintering simulation. The RNN component, a Physics-Gated Autoregressive LSTM, models the global, time-dependent density evolution. Its architecture explicitly embeds physical principles, such as thermal activation, diffusion state memory, and density saturation, through custom gating mechanisms, achieving an R² score of 0.9954 on the test set by accurately resolving all three distinct phases of densification. The GNN component takes the predicted density evolution to extend this framework by predicting localized, three-dimensional deformation, incorporating spatial relative positions, directional vectors, and boundary condition information from mesh-based graph representations. To ensure physically consistent predictions, the GNN uses a physics-informed loss function that enforces mass conservation. Trained on a large FEM-generated dataset containing 175 geometries simulated using a hybrid, experimentally calibrated viscoplastic model based on the Olevsky-Skorohod framework, the proposed machine learning framework significantly reduces computation time. While FEM simulations require approximately 52 min per geometry, our hybrid model completes predictions in just 24.64 s, achieving a speed-up of nearly 127 times. By leveraging high-fidelity, physics-informed data, this scalable, data-driven method enables rapid sintering prediction, supporting real-time process optimization, quality control, and the design of metal additive manufacturing components.
{"title":"Physics-informed hybrid deep learning-driven sintering simulation for next-generation metal additive manufacturing","authors":"Ali Kassab , Meriam Chelbi , Sajad Shirzad , Christopher Pannier , Pravansu Mohanty , Georges Ayoub","doi":"10.1016/j.addma.2025.105028","DOIUrl":"10.1016/j.addma.2025.105028","url":null,"abstract":"<div><div>Predicting shrinkage and deformation during sintering in metal additive manufacturing (AM) is challenging due to complex interactions among thermal gradients, material properties, and part geometry. Finite element methods (FEM) provide high-fidelity simulations but are computationally intensive, limiting their use for real-time control and design iteration. To address this, we present a hybrid deep learning framework that couples a physics-encoded recurrent neural network (RNN) with a physics-informed graph neural network (GNN) for efficient and accurate sintering simulation. The RNN component, a Physics-Gated Autoregressive LSTM, models the global, time-dependent density evolution. Its architecture explicitly embeds physical principles, such as thermal activation, diffusion state memory, and density saturation, through custom gating mechanisms, achieving an R² score of 0.9954 on the test set by accurately resolving all three distinct phases of densification. The GNN component takes the predicted density evolution to extend this framework by predicting localized, three-dimensional deformation, incorporating spatial relative positions, directional vectors, and boundary condition information from mesh-based graph representations. To ensure physically consistent predictions, the GNN uses a physics-informed loss function that enforces mass conservation. Trained on a large FEM-generated dataset containing 175 geometries simulated using a hybrid, experimentally calibrated viscoplastic model based on the Olevsky-Skorohod framework, the proposed machine learning framework significantly reduces computation time. While FEM simulations require approximately 52 min per geometry, our hybrid model completes predictions in just 24.64 s, achieving a speed-up of nearly 127 times. By leveraging high-fidelity, physics-informed data, this scalable, data-driven method enables rapid sintering prediction, supporting real-time process optimization, quality control, and the design of metal additive manufacturing components.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105028"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570836","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105027
Yayati Jadhav , Peter Pak , Amir Barati Farimani
Industry 4.0 has revolutionized manufacturing by driving digitization and shifting the paradigm toward additive manufacturing (AM). Material extrusion (MEX), a core AM method, produces customized and cost-effective products with minimal waste, challenging traditional subtractive manufacturing. Despite its advantages, MEX remains susceptible to defects that can compromise part quality and function, often requiring expert intervention. Existing rule-based and machine learning approaches struggle to generalize across different printers and sensors, while deep learning methods depend on large labeled datasets, limiting their scalability and adaptability. To address these challenges, we introduce a process monitoring and control framework that employs Large Language Models (LLMs) as autonomous controllers for additive manufacturing. Unlike rule-based or heavily data-dependent approaches, our method requires no domain-specific fine-tuning or training. Instead, the LLM leverages in-context learning, self-prompting, and iterative prompt-reason refinement to evaluate print quality from sequential image captures, detect and classify emerging failure modes, and query and modify the printer for relevant operating parameters. Through this adaptive reasoning process, the LLM not only interprets defects but also improves its own decision-making logic, autonomously formulating and executing corrective actions. This demonstrates a rule-free, self-improving approach to process control that extends beyond traditional quality assurance methods. We validated the effectiveness of the proposed framework by comparing it with a control group of engineers with different levels of AM expertise. The evaluation showed that LLM-based agents not only reliably identified common 3D printing errors such as inconsistent extrusion, stringing, warping, and poor layer adhesion, but also determined their causes and corrected them without human intervention. In addition to matching expert-level accuracy, the LLM was able to recognize emerging print errors earlier than human experts, highlighting its value as a proactive controller. To further demonstrate generalizability, we deployed and tested the framework on two different 3D printers with distinct sensor setups, confirming its adaptability across hardware. We also performed compression tests on baseline prints and on prints optimized by the LLM, with the optimized parts showing clear improvements in mechanical performance.
{"title":"LLM-3D print: Large Language Models to monitor and control 3D printing","authors":"Yayati Jadhav , Peter Pak , Amir Barati Farimani","doi":"10.1016/j.addma.2025.105027","DOIUrl":"10.1016/j.addma.2025.105027","url":null,"abstract":"<div><div>Industry 4.0 has revolutionized manufacturing by driving digitization and shifting the paradigm toward additive manufacturing (AM). Material extrusion (MEX), a core AM method, produces customized and cost-effective products with minimal waste, challenging traditional subtractive manufacturing. Despite its advantages, MEX remains susceptible to defects that can compromise part quality and function, often requiring expert intervention. Existing rule-based and machine learning approaches struggle to generalize across different printers and sensors, while deep learning methods depend on large labeled datasets, limiting their scalability and adaptability. To address these challenges, we introduce a process monitoring and control framework that employs Large Language Models (LLMs) as autonomous controllers for additive manufacturing. Unlike rule-based or heavily data-dependent approaches, our method requires no domain-specific fine-tuning or training. Instead, the LLM leverages in-context learning, self-prompting, and iterative prompt-reason refinement to evaluate print quality from sequential image captures, detect and classify emerging failure modes, and query and modify the printer for relevant operating parameters. Through this adaptive reasoning process, the LLM not only interprets defects but also improves its own decision-making logic, autonomously formulating and executing corrective actions. This demonstrates a rule-free, self-improving approach to process control that extends beyond traditional quality assurance methods. We validated the effectiveness of the proposed framework by comparing it with a control group of engineers with different levels of AM expertise. The evaluation showed that LLM-based agents not only reliably identified common 3D printing errors such as inconsistent extrusion, stringing, warping, and poor layer adhesion, but also determined their causes and corrected them without human intervention. In addition to matching expert-level accuracy, the LLM was able to recognize emerging print errors earlier than human experts, highlighting its value as a proactive controller. To further demonstrate generalizability, we deployed and tested the framework on two different 3D printers with distinct sensor setups, confirming its adaptability across hardware. We also performed compression tests on baseline prints and on prints optimized by the LLM, with the optimized parts showing clear improvements in mechanical performance.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105027"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145681657","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-25DOI: 10.1016/j.addma.2025.105032
Louis Remes , Thierry J. Massart , Caroline Fréderix , Philippe Hendrickx , Péter Berke
Inter-filament voids are critical defects in fused deposition modelling, leading to stress concentrations with undesired effects on the strength of the printed part. This work presents a novel methodology for the computational analysis of the mechanical behaviour of 3D printed materials incorporating these voids by exploiting computed tomography scans to automatically generate conformal finite element (FE) meshes with user control on the geometrical details accounted for in the model. This is achieved through an advanced curvature-based local smoothing algorithm, which preserves mechanically relevant morphological characteristics while reducing computational complexity and filtering out irrelevant geometrical features. The proposed approach is applied to 3D printed short carbon fibre reinforced polyether ether ketone (SCFR PEEK) samples subjected to realistic loading conditions. Results show inter-filament void induced stress concentrations and the occurrence of plastic events in the microstructural volume, leading to guidelines for choosing the appropriate level of the geometrical details and features to embark in the FE model.
{"title":"CT-scan based mechanical finite element analysis of inter-filament voids in fused deposition modelling","authors":"Louis Remes , Thierry J. Massart , Caroline Fréderix , Philippe Hendrickx , Péter Berke","doi":"10.1016/j.addma.2025.105032","DOIUrl":"10.1016/j.addma.2025.105032","url":null,"abstract":"<div><div>Inter-filament voids are critical defects in fused deposition modelling, leading to stress concentrations with undesired effects on the strength of the printed part. This work presents a novel methodology for the computational analysis of the mechanical behaviour of 3D printed materials incorporating these voids by exploiting computed tomography scans to automatically generate conformal finite element (FE) meshes with user control on the geometrical details accounted for in the model. This is achieved through an advanced curvature-based local smoothing algorithm, which preserves mechanically relevant morphological characteristics while reducing computational complexity and filtering out irrelevant geometrical features. The proposed approach is applied to 3D printed short carbon fibre reinforced polyether ether ketone (SCFR PEEK) samples subjected to realistic loading conditions. Results show inter-filament void induced stress concentrations and the occurrence of plastic events in the microstructural volume, leading to guidelines for choosing the appropriate level of the geometrical details and features to embark in the FE model.</div></div>","PeriodicalId":7172,"journal":{"name":"Additive manufacturing","volume":"114 ","pages":"Article 105032"},"PeriodicalIF":11.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145681658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号