Pub Date : 2026-02-28Epub Date: 2026-01-30DOI: 10.1016/j.jmapro.2026.01.095
Moaaz Safwa, Juho Bonifer, Hemantha Kumar Yeddu, Juha Varis, Ville Leminen
The extensive use of single-use food and beverage containers continues to raise environmental concerns, particularly due to the limited recyclability of polyolefin-coated paperboards. While polyethylene-coated structures remain prevalent in current manufacturing practices, improving their formability is essential for both optimizing existing processes and enabling the future adoption of more sustainable barrier materials. This study focuses on the bottom-forming stage of paper cup production, a process critical to achieving structural integrity and sealing performance. A finite element modeling approach was employed to simulate deformation behavior and stress distribution during forming, with experimental trials conducted to validate key results. Process parameters such as curling depth and production speed were systematically varied to evaluate their influence on forming outcomes. The results demonstrate that a curling depth of 4.3 mm provides improved seal consistency and structural uniformity, particularly at lower production speeds (80 cups/min), thereby minimizing the risk of defects commonly observed in fast, high-volume manufacturing settings. These findings contribute to a better understanding of process-structure relationships in fiber-based composite forming and offer valuable insights for reducing material waste and enhancing process reliability in sustainable packaging manufacturing.
{"title":"Process optimization of paper cup bottom-forming using FEM and experimental validation","authors":"Moaaz Safwa, Juho Bonifer, Hemantha Kumar Yeddu, Juha Varis, Ville Leminen","doi":"10.1016/j.jmapro.2026.01.095","DOIUrl":"10.1016/j.jmapro.2026.01.095","url":null,"abstract":"<div><div>The extensive use of single-use food and beverage containers continues to raise environmental concerns, particularly due to the limited recyclability of polyolefin-coated paperboards. While polyethylene-coated structures remain prevalent in current manufacturing practices, improving their formability is essential for both optimizing existing processes and enabling the future adoption of more sustainable barrier materials. This study focuses on the bottom-forming stage of paper cup production, a process critical to achieving structural integrity and sealing performance. A finite element modeling approach was employed to simulate deformation behavior and stress distribution during forming, with experimental trials conducted to validate key results. Process parameters such as curling depth and production speed were systematically varied to evaluate their influence on forming outcomes. The results demonstrate that a curling depth of 4.3 mm provides improved seal consistency and structural uniformity, particularly at lower production speeds (80 cups/min), thereby minimizing the risk of defects commonly observed in fast, high-volume manufacturing settings. These findings contribute to a better understanding of process-structure relationships in fiber-based composite forming and offer valuable insights for reducing material waste and enhancing process reliability in sustainable packaging manufacturing.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 581-590"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080428","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 : 2026-02-28Epub Date: 2026-01-20DOI: 10.1016/j.jmapro.2026.01.039
Yue Cao , Hai Lin , YuMing Zhang
Arc welding processes are vital for continuous fabrication but are susceptible to disturbances that cause defects and compromise weld quality. Real-time monitoring is therefore essential, yet remains challenging due to complex visual patterns and the nonlinear, time-varying nature of welding dynamics. While deep learning offers potential, its reliance on large, labeled datasets, and in particular, on process- and application-specific tuning, limits industrial scalability. We question whether there exists a common approach to characterize major arc processes across different applications. If so, by observing and monitoring characteristic variables as process states under a unified framework, scalability can be greatly improved. This paper presents a robust monitoring framework generalizable across arc welding processes. It integrates deep latent representation learning to extract compact features from weld pool views in an unsupervised manner and employs Bayesian filtering to enhance robustness against sensory disturbances that are both sustained and fluctuating in arc welding, such as arc radiation and specular reflection from the weld pool surface. To this end, we employ a Dynamic Variational Autoencoder (DVAE), composed of a Convolutional Neural Network (CNN)-based encoder-decoder and a Long Short-Term Memory (LSTM)-based transition model, to jointly learn compact latent representations of weld pool images and their evolution under control inputs. This setup fuses instantaneous visual representations with process dynamics modeling, enabling compact latent features that satisfy both objectives. To achieve robust real-time inference, a specialized Particle Filter (PF) is introduced to jointly propagate the latent state and the hidden state of the LSTM transition model, preserving both current and historical process information while suppressing sensor disturbances such as arc rotation and specular reflection. This design is well suited to the relatively slow and inertial dynamics of arc welding, allowing the PF to effectively fuse model-based predictions with real-time observations. The proposed framework is validated on both GTAW and GMAW processes without process-specific customization, demonstrating its generalizability and robustness.
{"title":"Robust monitoring of arc welding processes: A generalizable framework with DVAE and particle filter","authors":"Yue Cao , Hai Lin , YuMing Zhang","doi":"10.1016/j.jmapro.2026.01.039","DOIUrl":"10.1016/j.jmapro.2026.01.039","url":null,"abstract":"<div><div>Arc welding processes are vital for continuous fabrication but are susceptible to disturbances that cause defects and compromise weld quality. Real-time monitoring is therefore essential, yet remains challenging due to complex visual patterns and the nonlinear, time-varying nature of welding dynamics. While deep learning offers potential, its reliance on large, labeled datasets, and in particular, on process- and application-specific tuning, limits industrial scalability. We question whether there exists a common approach to characterize major arc processes across different applications. If so, by observing and monitoring characteristic variables as process states under a unified framework, scalability can be greatly improved. This paper presents a robust monitoring framework generalizable across arc welding processes. It integrates deep latent representation learning to extract compact features from weld pool views in an unsupervised manner and employs Bayesian filtering to enhance robustness against sensory disturbances that are both sustained and fluctuating in arc welding, such as arc radiation and specular reflection from the weld pool surface. To this end, we employ a Dynamic Variational Autoencoder (DVAE), composed of a Convolutional Neural Network (CNN)-based encoder-decoder and a Long Short-Term Memory (LSTM)-based transition model, to jointly learn compact latent representations of weld pool images and their evolution under control inputs. This setup fuses instantaneous visual representations with process dynamics modeling, enabling compact latent features that satisfy both objectives. To achieve robust real-time inference, a specialized Particle Filter (PF) is introduced to jointly propagate the latent state and the hidden state of the LSTM transition model, preserving both current and historical process information while suppressing sensor disturbances such as arc rotation and specular reflection. This design is well suited to the relatively slow and inertial dynamics of arc welding, allowing the PF to effectively fuse model-based predictions with real-time observations. The proposed framework is validated on both GTAW and GMAW processes without process-specific customization, demonstrating its generalizability and robustness.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 15-27"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146001830","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 : 2026-02-28Epub Date: 2026-01-21DOI: 10.1016/j.jmapro.2026.01.014
Duo Zhang , Shuaijie Ji , Zheng Wang , Yanfeng Yang , Heng Yang , Heng Li
Capillary tubes, valued for their large specific surface area and exceptional heat transfer efficiency, are widely used in aerospace, biomedical devices and chemical engineering. Mandrel-free drawing is essential for producing high-performance capillary tubes. In this process, through-thickness deformation is governed by external constraints on the outer surface, whereas the inner surface maintains a free boundary. This creates a radially asymmetric stress field, resulting in uneven deformation through the thickness. When superimposed with process parameter fluctuation, tube blank dimension change, and size effect, the uneven deformation was further exacerbated, making precise control of wall thickness in capillary tube fabrication more challenging. In this work, taking the mandrel-free drawing of GH4169 capillary tubes as the study case, through a series of well-designed simulations and experiments, the coupling effects of process parameters (section reduction, friction coefficient, die angle and sizing band length), tube blank dimension (D/t, the ratio of tube diameter to wall thickness), and size effect factor (t/d, the ratio of wall thickness to grain size) on wall thickness evolution were systemically investigated and revealed. The main findings include: 1) The section reduction, friction coefficient, die angle, and D/t exhibit significant influence on wall thickness evolution of capillary tubes during the mandrel-free drawing, while the sizing band length and t/d exert relatively minor effects. 2) The increase in friction coefficient and die angle raises the deformation gradient and axial stress, making the circumferential compressive strain more easily transform into the axial tensile strain to coordinate deformation, which decreases radial strain and alleviates wall thickness thickening. Conversely, the large section reduction and D/t decrease the deformation gradient and axial stress, resulting in an increase in wall thickness. 3) Due to the separation of the tube from the sizing band, the sizing band length has little influence on the evolution of wall thickness. As the t/d increases, the ratio of axial to radial deformation resistance remains constant, making the evolution of wall thickness independent of the t/d. 4) Based on the above insights, a novel wall thickness control strategy, employing increased friction coefficients (f = 0.12) and large die angles (α = 24°), was proposed to actively regulate the wall thickness of capillary tubes during the mandrel-free drawing process. The drawing experiments indicate that the absolute error of wall thickness was decreased from 0.041 mm to 0.013 mm, achieving a 68.29% improvement in forming accuracy. The developed method in this work will contribute to the high-precision manufacturing of high-performance capillary tubes.
{"title":"Mechanism and control of wall thickness evolution in mandrel-free drawing of capillary tubes","authors":"Duo Zhang , Shuaijie Ji , Zheng Wang , Yanfeng Yang , Heng Yang , Heng Li","doi":"10.1016/j.jmapro.2026.01.014","DOIUrl":"10.1016/j.jmapro.2026.01.014","url":null,"abstract":"<div><div>Capillary tubes, valued for their large specific surface area and exceptional heat transfer efficiency, are widely used in aerospace, biomedical devices and chemical engineering. Mandrel-free drawing is essential for producing high-performance capillary tubes. In this process, through-thickness deformation is governed by external constraints on the outer surface, whereas the inner surface maintains a free boundary. This creates a radially asymmetric stress field, resulting in uneven deformation through the thickness. When superimposed with process parameter fluctuation, tube blank dimension change, and size effect, the uneven deformation was further exacerbated, making precise control of wall thickness in capillary tube fabrication more challenging. In this work, taking the mandrel-free drawing of GH4169 capillary tubes as the study case, through a series of well-designed simulations and experiments, the coupling effects of process parameters (section reduction, friction coefficient, die angle and sizing band length), tube blank dimension (<em>D/t</em>, the ratio of tube diameter to wall thickness), and size effect factor (<em>t/d</em>, the ratio of wall thickness to grain size) on wall thickness evolution were systemically investigated and revealed. The main findings include: 1) The section reduction, friction coefficient, die angle, and <em>D/t</em> exhibit significant influence on wall thickness evolution of capillary tubes during the mandrel-free drawing, while the sizing band length and <em>t/d</em> exert relatively minor effects. 2) The increase in friction coefficient and die angle raises the deformation gradient and axial stress, making the circumferential compressive strain more easily transform into the axial tensile strain to coordinate deformation, which decreases radial strain and alleviates wall thickness thickening. Conversely, the large section reduction and <em>D/t</em> decrease the deformation gradient and axial stress, resulting in an increase in wall thickness. 3) Due to the separation of the tube from the sizing band, the sizing band length has little influence on the evolution of wall thickness. As the <em>t/d</em> increases, the ratio of axial to radial deformation resistance remains constant, making the evolution of wall thickness independent of the <em>t/d</em>. 4) Based on the above insights, a novel wall thickness control strategy, employing increased friction coefficients (<em>f</em> = 0.12) and large die angles (<em>α</em> = 24°), was proposed to actively regulate the wall thickness of capillary tubes during the mandrel-free drawing process. The drawing experiments indicate that the absolute error of wall thickness was decreased from 0.041 mm to 0.013 mm, achieving a 68.29% improvement in forming accuracy. The developed method in this work will contribute to the high-precision manufacturing of high-performance capillary tubes.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 28-49"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146001831","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 : 2026-02-28Epub Date: 2026-01-22DOI: 10.1016/j.jmapro.2026.01.062
Fan Zhang, Wenlong Hu, Yun Wang, Ji'’an Duan
3D defect detection by multimodal representations is vital in manufacturing field, but simple concatenation from multimodal feature may result in feature interference, thereby reducing fusion effectiveness. Hereby, this study proposes a dynamic background-guided asymmetric knowledge distillation network (DAK-Net) to realize 3D defect detection by using a multimodal fusion that fuses the features of RGB and depth images. The DAK-Net mainly consists of a 2D multi-scale feature extractor, spatial reorganization downsampling, foreground mask dynamic extraction, asymmetric feature fusion, and asymmetric knowledge distillation. The 2D multiscale feature extractor realizes the extraction of RGB image features through a multiscale feature splicing. The spatial reorganization downsampling module implements the spatial-to-channel dimension information reorganization. The foreground mask dynamic extraction module realizes the calculation of anomaly scores only in the foreground region to avoid background interference. The asymmetric feature fusion module is designed for merging features from both RGB and depth images. Concurrently, the framework employs an asymmetric knowledge distillation strategy, in which the teacher network employs conditional normalizing flows to learn a mapping that transforms the complex data distribution into a standard normal distribution, while the student network focuses on regressing the teacher's output specifically on normal, defect-free data. The experiments for DAK-Net achieved average image-level AUROC of 93.6% on MVTec-3D AD dataset and 57.35% on Anomaly-ShapeNet dataset, which demonstrated excellent 3D defect detection performance.
{"title":"Dynamic background-guided asymmetric knowledge distillation network for 3D defect detection","authors":"Fan Zhang, Wenlong Hu, Yun Wang, Ji'’an Duan","doi":"10.1016/j.jmapro.2026.01.062","DOIUrl":"10.1016/j.jmapro.2026.01.062","url":null,"abstract":"<div><div>3D defect detection by multimodal representations is vital in manufacturing field, but simple concatenation from multimodal feature may result in feature interference, thereby reducing fusion effectiveness. Hereby, this study proposes a dynamic background-guided asymmetric knowledge distillation network (DAK-Net) to realize 3D defect detection by using a multimodal fusion that fuses the features of RGB and depth images. The DAK-Net mainly consists of a 2D multi-scale feature extractor, spatial reorganization downsampling, foreground mask dynamic extraction, asymmetric feature fusion, and asymmetric knowledge distillation. The 2D multiscale feature extractor realizes the extraction of RGB image features through a multiscale feature splicing. The spatial reorganization downsampling module implements the spatial-to-channel dimension information reorganization. The foreground mask dynamic extraction module realizes the calculation of anomaly scores only in the foreground region to avoid background interference. The asymmetric feature fusion module is designed for merging features from both RGB and depth images. Concurrently, the framework employs an asymmetric knowledge distillation strategy, in which the teacher network employs conditional normalizing flows to learn a mapping that transforms the complex data distribution into a standard normal distribution, while the student network focuses on regressing the teacher's output specifically on normal, defect-free data. The experiments for DAK-Net achieved average image-level AUROC of 93.6% on MVTec-3D AD dataset and 57.35% on Anomaly-ShapeNet dataset, which demonstrated excellent 3D defect detection performance.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 185-199"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036929","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 : 2026-02-28Epub Date: 2026-01-30DOI: 10.1016/j.jmapro.2026.01.088
Juntao Shen , Baokai Ren , Ping Yao , Kang Zhou
Vehicle light-weighting has driven growing interests in joining aluminum alloys with press hardened steel (PHS), a metal widely used for its high strength. However, the significant physical and chemical differences between aluminum and PHS lead to low welding compatibility in conventional resistance spot welding (RSW) process, and some typically hard and brittle Fe-Al intermetallic compounds (IMCs) forming during the process, which can significantly deteriorate joint performance. This work proposes an improved method for RSW process of AA6061 aluminum alloy and PHS by introducing a stainless steel interlayer (thickness: 0.25–0.5 mm) and ultrasonic longitudinal vibration. The effects of different interlayer thicknesses and process combinations: conventional RSW, with interlayer only, and with both of interlayer and ultrasonic assistance (UA) were investigated in terms of dynamic resistance, metallurgical characteristics, joint microstructure, and mechanical properties of the Al/PHS welded joint. Results show that the stainless steel interlayer could effectively prevent direct Al-PHS contact. Using 0.45 mm interlayer made the IMC thickness reduced from ∼50 μm without the interlayer to ∼2.3 μm. Ultrasonic vibration further reduced the IMC layer to ∼1.4 μm, and enhanced Fe diffusion toward the aluminum side. With a 0.45 mm thick interlayer, the joint obtained from UA-RSW process could achieve a peak load of 6.32kN, which was 65.6% higher than that of the process using the same thickness interlayer without UA. The fracture energy was increased from 4.393 J to 14.806 J, and the fracture mode of the joint changed from interfacial fracture to better button fracture. These findings demonstrate that the synergistic using of a stainless steel interlayer and ultrasonic vibration can enable effective joining of PHS and aluminum alloys, and offer a promising solution for joining dissimilar metals and ultra-high-strength steels in advanced lightweight structures.
{"title":"Improving the weldability of press hardened steel to aluminum alloy in resistance spot welding using interlayer and ultrasonic assistance","authors":"Juntao Shen , Baokai Ren , Ping Yao , Kang Zhou","doi":"10.1016/j.jmapro.2026.01.088","DOIUrl":"10.1016/j.jmapro.2026.01.088","url":null,"abstract":"<div><div>Vehicle light-weighting has driven growing interests in joining aluminum alloys with press hardened steel (PHS), a metal widely used for its high strength. However, the significant physical and chemical differences between aluminum and PHS lead to low welding compatibility in conventional resistance spot welding (RSW) process, and some typically hard and brittle Fe-Al intermetallic compounds (IMCs) forming during the process, which can significantly deteriorate joint performance. This work proposes an improved method for RSW process of AA6061 aluminum alloy and PHS by introducing a stainless steel interlayer (thickness: 0.25–0.5 mm) and ultrasonic longitudinal vibration. The effects of different interlayer thicknesses and process combinations: conventional RSW, with interlayer only, and with both of interlayer and ultrasonic assistance (UA) were investigated in terms of dynamic resistance, metallurgical characteristics, joint microstructure, and mechanical properties of the Al/PHS welded joint. Results show that the stainless steel interlayer could effectively prevent direct Al-PHS contact. Using 0.45 mm interlayer made the IMC thickness reduced from ∼50 μm without the interlayer to ∼2.3 μm. Ultrasonic vibration further reduced the IMC layer to ∼1.4 μm, and enhanced Fe diffusion toward the aluminum side. With a 0.45 mm thick interlayer, the joint obtained from UA-RSW process could achieve a peak load of 6.32kN, which was 65.6% higher than that of the process using the same thickness interlayer without UA. The fracture energy was increased from 4.393 J to 14.806 J, and the fracture mode of the joint changed from interfacial fracture to better button fracture. These findings demonstrate that the synergistic using of a stainless steel interlayer and ultrasonic vibration can enable effective joining of PHS and aluminum alloys, and offer a promising solution for joining dissimilar metals and ultra-high-strength steels in advanced lightweight structures.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 591-610"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080331","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 : 2026-02-28Epub Date: 2026-01-27DOI: 10.1016/j.jmapro.2026.01.075
Huijuan Ma , Xiaoying Wei , Peiliao Wang , Zhiang Gong , Zhili Hu , Lin Hua
Pre-aged hardening warm forming (PHF) technology enables precise control of process parameters, allowing pre-hardened sheet to achieve superior formability compared to the O-temper condition. During subsequent forming stages, this process utilizes synergistic control of deformation and phase transformation, enabling the final component to attain mechanical properties comparable to the T6 temper. By employing pre-hardened technology in conjunction with warm forming process, the post-forming solution heat treatment and aging steps can be eliminated, thereby significantly reducing the component manufacturing cycle. However, as an emerging technique, past experience has limited guidance on excavating the mechanism and deducting the parameters of PHF process. Here, the Long Short-Term Memory network (LSTM) model of 7075 aluminum alloy (AA7075) is firstly established, which innovatively facilitates bidirectional prediction between process parameters and mechanical properties. Crucially, a constitutive model of PHF process based on dynamic precipitation and dislocation strengthening is proposed, considering the direct phase precipitation from the solid solution and the inherited precipitation from GPII zones to η' phases based on microstructure characterization utilizing the HRTEM, DSC, SAXS and the XRD. Moreover, the accuracy of the LSTM model is further improved through a novel pre-training approach that assimilates knowledge from the AA7075 constitutive model, followed by fine-tuning with experimental dataset. Embracing a “mechanism + data” fusion-driven approach, the mechanical properties prediction and the process parameters deduction of high-strength aluminum alloy components formed under the PHF process are achieved. Additionally, rapid and accurate deduction of process parameters for 7050 aluminum alloy (AA7050) with similar phase evolution is realized by transfer learning from the AA7075 LSTM model using little experimental data. This study not only accelerates the development of higher-performance aluminum alloy components, but also establishes a foundational framework for swiftly determining the process window under the cooperative control of deformation and phase transformation.
{"title":"Integrating mechanism and data-driven approaches in pre-aged hardening warm forming: Performance prediction and process parameters deduction","authors":"Huijuan Ma , Xiaoying Wei , Peiliao Wang , Zhiang Gong , Zhili Hu , Lin Hua","doi":"10.1016/j.jmapro.2026.01.075","DOIUrl":"10.1016/j.jmapro.2026.01.075","url":null,"abstract":"<div><div>Pre-aged hardening warm forming (PHF) technology enables precise control of process parameters, allowing pre-hardened sheet to achieve superior formability compared to the O-temper condition. During subsequent forming stages, this process utilizes synergistic control of deformation and phase transformation, enabling the final component to attain mechanical properties comparable to the T6 temper. By employing pre-hardened technology in conjunction with warm forming process, the post-forming solution heat treatment and aging steps can be eliminated, thereby significantly reducing the component manufacturing cycle. However, as an emerging technique, past experience has limited guidance on excavating the mechanism and deducting the parameters of PHF process. Here, the Long Short-Term Memory network (LSTM) model of 7075 aluminum alloy (AA7075) is firstly established, which innovatively facilitates bidirectional prediction between process parameters and mechanical properties. Crucially, a constitutive model of PHF process based on dynamic precipitation and dislocation strengthening is proposed, considering the direct phase precipitation from the solid solution and the inherited precipitation from GPII zones to η' phases based on microstructure characterization utilizing the HRTEM, DSC, SAXS and the XRD. Moreover, the accuracy of the LSTM model is further improved through a novel pre-training approach that assimilates knowledge from the AA7075 constitutive model, followed by fine-tuning with experimental dataset. Embracing a “mechanism + data” fusion-driven approach, the mechanical properties prediction and the process parameters deduction of high-strength aluminum alloy components formed under the PHF process are achieved. Additionally, rapid and accurate deduction of process parameters for 7050 aluminum alloy (AA7050) with similar phase evolution is realized by transfer learning from the AA7075 LSTM model using little experimental data. This study not only accelerates the development of higher-performance aluminum alloy components, but also establishes a foundational framework for swiftly determining the process window under the cooperative control of deformation and phase transformation.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 442-455"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080426","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 : 2026-02-28Epub Date: 2026-01-27DOI: 10.1016/j.jmapro.2026.01.066
Tao Wu , Yong Zhang , Yongfei Wang , Bin Hu , Chen Li
Converting micro-segment toolpaths into high-order curves can significantly enhance the stability of five-axis CNC machining processes. However, conventional toolpath optimization approaches tend to simultaneously cause both undercutting and overcutting on the workpiece surfaces. Overcutting leads to irreversible morphological damage to the workpiece, thereby resulting in scrapped parts. Furthermore, the asynchronous variation between speed limit trends and speed planning curves undermines the effectiveness of conventional speed planning strategies in the machining of complex structural workpieces. To achieve effective control over machining stability and accuracy in five-axis CNC machining of complex workpieces, this work proposed an error-controllable G3-continuous oriented toolpath optimization algorithm. Based on the G3-continuous quartic symmetric Bezier curve, the toolpath was directionally offset according to the viewing-angle theorem. To ensure toolpath reachability, the Sobolev seminorm method and CVE method were subsequently employed to further optimize toolpath stability. Additionally, an enhanced speed planning strategy with an extra verification mechanism was designed. By incorporating adaptive quintic Gauss-Legendre quadrature and S-shaped speed model, a numerical model was established to characterize the relationships among curvature radius, arc length, and motion time. The activation conditions for the verification mechanism were derived using quartic non-uniform difference formulas. The secant method was applied to dynamically adjust local snap parameters of current toolpath segments for speed profile modulation. Five-axis machining experiments on dentures were conducted to validate the effectiveness of optimization algorithms. Experimental results demonstrated that, compared with traditional strategies, the modified toolpath optimization and speed look-ahead algorithms reduced machine tool vibration by 6.62% and 19.46%, respectively, while increasing dimensional compliance rates by 283.79% and 439.774%, respectively. This work successfully mitigates the challenges of overcutting, machine chatter, and accuracy drift in the five-axis CNC machining of complex structural components, thereby offering theoretical support for the development of high-precision and stable machining technologies for such components.
{"title":"An error-controlled G3-continuous oriented toolpath optimization algorithm and modified speed planning for five-axis machining","authors":"Tao Wu , Yong Zhang , Yongfei Wang , Bin Hu , Chen Li","doi":"10.1016/j.jmapro.2026.01.066","DOIUrl":"10.1016/j.jmapro.2026.01.066","url":null,"abstract":"<div><div>Converting micro-segment toolpaths into high-order curves can significantly enhance the stability of five-axis CNC machining processes. However, conventional toolpath optimization approaches tend to simultaneously cause both undercutting and overcutting on the workpiece surfaces. Overcutting leads to irreversible morphological damage to the workpiece, thereby resulting in scrapped parts. Furthermore, the asynchronous variation between speed limit trends and speed planning curves undermines the effectiveness of conventional speed planning strategies in the machining of complex structural workpieces. To achieve effective control over machining stability and accuracy in five-axis CNC machining of complex workpieces, this work proposed an error-controllable G3-continuous oriented toolpath optimization algorithm. Based on the G3-continuous quartic symmetric Bezier curve, the toolpath was directionally offset according to the viewing-angle theorem. To ensure toolpath reachability, the Sobolev seminorm method and CVE method were subsequently employed to further optimize toolpath stability. Additionally, an enhanced speed planning strategy with an extra verification mechanism was designed. By incorporating adaptive quintic Gauss-Legendre quadrature and S-shaped speed model, a numerical model was established to characterize the relationships among curvature radius, arc length, and motion time. The activation conditions for the verification mechanism were derived using quartic non-uniform difference formulas. The secant method was applied to dynamically adjust local snap parameters of current toolpath segments for speed profile modulation. Five-axis machining experiments on dentures were conducted to validate the effectiveness of optimization algorithms. Experimental results demonstrated that, compared with traditional strategies, the modified toolpath optimization and speed look-ahead algorithms reduced machine tool vibration by 6.62% and 19.46%, respectively, while increasing dimensional compliance rates by 283.79% and 439.774%, respectively. This work successfully mitigates the challenges of overcutting, machine chatter, and accuracy drift in the five-axis CNC machining of complex structural components, thereby offering theoretical support for the development of high-precision and stable machining technologies for such components.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 456-478"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080397","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 : 2026-02-28Epub Date: 2026-01-28DOI: 10.1016/j.jmapro.2026.01.017
Muyang Ye , Haohua Xiu , Chung Ket Thein , Haotian Cui , Yongjie Zhao , Gongyu Liu , Jing Wang , Hao Nan Li
Laser beam drilling is widely employed in the aerospace industry due to its non-contact nature and efficient processing of various materials. While traditional fixed laser beam drilling methods such as Single Pulse Drilling (SPD) and Percussion Laser Drilling (PLD) are commonly used, mobile laser beam drilling techniques like Laser Trepanning and Helical Drilling are preferred for applications requiring precise control over hole geometric accuracy. This paper presents a new Dove-prism-based trepanning system model that enables analytical calculation of the laser trajectory within a 3D spiral domain. This model facilitates accurate prediction of drilled hole geometries, including diameter and taper. An innovative aspect of this study lies in the incorporation of a laser ablation effect into the prediction of hole geometry, which is often overlooked in other trepanning drilling research. By integrating a prediction function for ablation crater diameter, the accuracy of hole geometry prediction can be improved. The validity of the model is confirmed through extensive experiments, establishing its reliability while revealing important insights such as the impact of initial optomechanical conditions on hole geometry and the influence of laser parameters on hole circularity. Additionally, our compensation method enhances predictability and expands achievable geometry range when drilling holes. This research establishes a robust theoretical foundation for advancing mobile laser drilling technology, particularly in terms of system design and process optimization.
{"title":"The modelling and compensation method for dove-prism-based laser trepanning optomechanical system","authors":"Muyang Ye , Haohua Xiu , Chung Ket Thein , Haotian Cui , Yongjie Zhao , Gongyu Liu , Jing Wang , Hao Nan Li","doi":"10.1016/j.jmapro.2026.01.017","DOIUrl":"10.1016/j.jmapro.2026.01.017","url":null,"abstract":"<div><div>Laser beam drilling is widely employed in the aerospace industry due to its non-contact nature and efficient processing of various materials. While traditional fixed laser beam drilling methods such as Single Pulse Drilling (SPD) and Percussion Laser Drilling (PLD) are commonly used, mobile laser beam drilling techniques like Laser Trepanning and Helical Drilling are preferred for applications requiring precise control over hole geometric accuracy. This paper presents a new Dove-prism-based trepanning system model that enables analytical calculation of the laser trajectory within a 3D spiral domain. This model facilitates accurate prediction of drilled hole geometries, including diameter and taper. An innovative aspect of this study lies in the incorporation of a laser ablation effect into the prediction of hole geometry, which is often overlooked in other trepanning drilling research. By integrating a prediction function for ablation crater diameter, the accuracy of hole geometry prediction can be improved. The validity of the model is confirmed through extensive experiments, establishing its reliability while revealing important insights such as the impact of initial optomechanical conditions on hole geometry and the influence of laser parameters on hole circularity. Additionally, our compensation method enhances predictability and expands achievable geometry range when drilling holes. This research establishes a robust theoretical foundation for advancing mobile laser drilling technology, particularly in terms of system design and process optimization.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 480-497"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080392","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 : 2026-02-28Epub Date: 2026-01-26DOI: 10.1016/j.jmapro.2026.01.076
Shuhan Li , Xinqiang Lan , Zemin Wang
Extrusion-based additive manufacturing of powder-binder feedstocks offers a cost-effective route for indirectly fabricating metallic components. However, the sintering step remains a major bottleneck, typically requiring lengthy, powder-specific optimization that can exceed the effort spent on printing parameter tuning. This study overcomes this challenge by establishing a direct correlation between macroscopic shrinkage and microscopic porosity, enabling rapid assessment of sintering quality. For H13 steel, densification proceeds through solid-phase sintering at 1000–1350 °C and liquid-phase sintering at 1400–1450 °C. Thermodynamic analysis and experimental results confirm that by increasing the sintering temperature and avoiding excessive liquid-phase sintering (1300–1400 °C), both SSAM-5 and SSAM-10 powders (with medium particle size of 5.3 μm and 11.8 μm) can achieve ideal porosities of 0.45% and 0.96% after 1–3 h of holding. Macroscopic shrinkage was observed after sintering and approached a theoretical limit as porosity decreased. A quantitative model linking shrinkage to porosity was developed, enabling the immediate assessment of internal densification using easily accessible macroscopic data.
{"title":"Revealing the sintering behavior of H13 steel in semi-solid additive manufacturing through the correlation of shrinkage and porosity","authors":"Shuhan Li , Xinqiang Lan , Zemin Wang","doi":"10.1016/j.jmapro.2026.01.076","DOIUrl":"10.1016/j.jmapro.2026.01.076","url":null,"abstract":"<div><div>Extrusion-based additive manufacturing of powder-binder feedstocks offers a cost-effective route for indirectly fabricating metallic components. However, the sintering step remains a major bottleneck, typically requiring lengthy, powder-specific optimization that can exceed the effort spent on printing parameter tuning. This study overcomes this challenge by establishing a direct correlation between macroscopic shrinkage and microscopic porosity, enabling rapid assessment of sintering quality. For H13 steel, densification proceeds through solid-phase sintering at 1000–1350 °C and liquid-phase sintering at 1400–1450 °C. Thermodynamic analysis and experimental results confirm that by increasing the sintering temperature and avoiding excessive liquid-phase sintering (1300–1400 °C), both SSAM-5 and SSAM-10 powders (with medium particle size of 5.3 μm and 11.8 μm) can achieve ideal porosities of 0.45% and 0.96% after 1–3 h of holding. Macroscopic shrinkage was observed after sintering and approached a theoretical limit as porosity decreased. A quantitative model linking shrinkage to porosity was developed, enabling the immediate assessment of internal densification using easily accessible macroscopic data.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 359-370"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146080407","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 : 2026-02-28Epub Date: 2026-01-23DOI: 10.1016/j.jmapro.2026.01.040
Shengtao Lin , Kai Wang , Zhengcai Zhao , Yucan Fu
Geometric deviations in near-net-shape (NNS) parts often lead to the undercut defects and demanding numerical control (NC) programming, posing challenges for high-precision machining. To address these issues, this paper proposes a Gaussian mixture model (GMM)-driven non-rigid toolpath morphing framework integrating a novel skeleton-skin strategy. First, a design intent-preserving model is developed to construct feasible machining points (skeleton points) under nonlinear constraints of machining allowance and profile tolerance, addressing undercut regions with negative machining allowance. Second, the nominal toolpath cutter locations (skin points) are morphed to conform to the skeleton points through a GMM-based non-rigid morphing algorithm, bypassing conventional point-curve-surface reconstruction and enabling direct NC programming. Importantly, a Bayesian optimization method utilizing symmetric Hausdorff distance is introduced to determine the optimal parameters for non-rigid morphing. A comprehensive case study on a 3D-printed turbine blade, including the performance evaluations and milling experiments, is conducted to validate the proposed framework. Results show that the machined areas meet the ±0.10 mm profile tolerance requirement, while toolpath generation time is reduced by 31%. This work establishes a critical link between non-rigid shape compensation and efficient NC programming for NNS parts.
{"title":"Design intent-preserving non-rigid toolpath morphing: A novel skeleton-skin method for undercut compensation and rapid numerical control programming of near-net-shape parts","authors":"Shengtao Lin , Kai Wang , Zhengcai Zhao , Yucan Fu","doi":"10.1016/j.jmapro.2026.01.040","DOIUrl":"10.1016/j.jmapro.2026.01.040","url":null,"abstract":"<div><div>Geometric deviations in near-net-shape (NNS) parts often lead to the undercut defects and demanding numerical control (NC) programming, posing challenges for high-precision machining. To address these issues, this paper proposes a Gaussian mixture model (GMM)-driven non-rigid toolpath morphing framework integrating a novel skeleton-skin strategy. First, a design intent-preserving model is developed to construct feasible machining points (skeleton points) under nonlinear constraints of machining allowance and profile tolerance, addressing undercut regions with negative machining allowance. Second, the nominal toolpath cutter locations (skin points) are morphed to conform to the skeleton points through a GMM-based non-rigid morphing algorithm, bypassing conventional point-curve-surface reconstruction and enabling direct NC programming. Importantly, a Bayesian optimization method utilizing symmetric Hausdorff distance is introduced to determine the optimal parameters for non-rigid morphing. A comprehensive case study on a 3D-printed turbine blade, including the performance evaluations and milling experiments, is conducted to validate the proposed framework. Results show that the machined areas meet the ±0.10 mm profile tolerance requirement, while toolpath generation time is reduced by 31%. This work establishes a critical link between non-rigid shape compensation and efficient NC programming for NNS parts.</div></div>","PeriodicalId":16148,"journal":{"name":"Journal of Manufacturing Processes","volume":"160 ","pages":"Pages 302-316"},"PeriodicalIF":6.8,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036784","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