In situ consolidation automated fiber placement of thermoplastic composites (ICAT) is one of the key engineering technologies for the widespread application of environmentally friendly materials in important fields such as aerospace. Nevertheless, the high melting point and low viscosity characteristics of high-performance thermoplastic composites such as PEEK are key challenges limiting the large-scale industrialization of ICAT technology. As a consequence, researching the defect formation mechanism and suppression methods of the interlayer fusion process of ICAT technology in new high-performance low-melting-point thermoplastic matrix composites is an important direction for promoting the large-scale application of thermoplastic composites. Hence, this paper investigated the ICAT in situ forming process of low-melting-point polyaryletherketone matrix (LM/PAEK) composites. The temperature history was analyzed using a high-frequency temperature acquisition system. Simultaneously, the Taguchi algorithm was used to analyze the coupled effects of multiple consolidation parameters. The results indicate that consolidation temperature is the most significant factor affecting void and interlayer performance. The consolidation parameters primarily influence the extrusion and permeation behavior of the matrix. Furthermore, the optimized consolidation parameter combination is as follows: consolidation temperature of 400 degrees C, consolidation speed of 100 mm/s, and consolidation force of 300 N. Additionally, under the optimized consolidation parameter combination conditions, the porosity and interlayer shear strength were 1.42% and 39.53 MPa, respectively. The porosity was reduced by 78.39%, and the interlayer shear strength was increased by 50.59%.

Accurate prediction of cellulose concentration in lyocell slurry is crucial for process control and product quality in sustainable lyocell fiber production, yet the complex, nonlinear nature of the swelling process makes it challenging to model using conventional parametric methods. This study develops a hybrid ensemble machine learning approach to predict the cellulose concentration of lyocell slurry based on industrial production data. A data set comprising 350 samples was collected from an industrial pulping process, covering 18 input variables related to raw material properties and process conditions. Six conventional machine learning modelsGaussian process regression (GPR), support vector regression (SVR), kernel regression (KR), multivariate adaptive regression spline (MARS), random forest (RF), and artificial neural network (ANN)were first established and optimized using Bayesian optimization with 5-fold cross-validation. Subsequently, a hybrid ensemble model (HEM) was constructed by aggregating the predictions of the six base models using a random forest meta-learner selected through score analysis. The predictive performance of all models was evaluated using multiple metrics, including coefficient of determination (R 2), root-mean-square error (RMSE), and mean absolute error (MAE). The results show that the HEM achieves the best overall testing performance (R 2 = 0.761, RMSE = 0.067, MAE = 0.053), followed closely by the random forest model (R 2 = 0.754, RMSE = 0.068, MAE = 0.053). The ANN exhibits the smallest training-testing performance gap, confirming the effectiveness of L2 regularization. Kernel-based methods and MARS yield inferior accuracy (testing R 2 < 0.70), indicating the limitations of global smoothness or additive assumptions for this task. SHAP analysis identifies the NMMO-to-cellulose ratio, hydroxylamine concentration, and temperature parameters as the most influential features. The proposed HEM provides a reliable tool for online prediction of slurry cellulose concentration, enabling real-time process control and contributing to more efficient and sustainable lyocell fiber production.

During the fiber winding process, the combination of design parameters results in diverse winding structures, making the performance simulation and analysis for fiber wound composites (FWC) a challenging task. This paper presents a parametric modeling framework to characterize the cross-undulating features of various FWC, coupled with a progressive damage constitutive model that accounts for nonlinear shear. The mechanical behaviors of different winding structures were studied through the combination of experiments and simulations. The finite element (FE) models for three winding patterns were established and corresponding specimens for each pattern were fabricated to test under quasi-static tensile conditions, while Digital Image Correlation (DIC) technology was used to monitor the strain field variations. The results indicate that the FWC exhibit nonlinear behavior, with stress-strain distributions highly sensitive to winding patterns. The stiffness differences among these structures are closely linked to the proportion and distribution of cross-undulating features. Numerical simulation results for three winding patterns agree well with experimental load-displacement curves (secant stiffness within 7.11 %), strain fields (similar distribution) and failure modes (same shear failure). Then the numerical model was applied to winding structures with varying winding angles to predict their modulus and strength. The proposed model reproduces the mechanical behavior of diverse winding structures, facilitating reliable predictions for aerospace and new energy composite components.

3D woven composites are widely used in aerospace, rail transit and other fields. However, for complex components such as aero-engine blades, the accurate prediction of yarn paths and thickness distribution during the geometric transformation from twisted state to flattened state remains a critical bottleneck restricting their design. This paper presents a parametric extraction method for yarn trajectory prediction and thickness distribution suitable for complex 3D woven composites. Firstly, the composite model is reconstructed to accurately extract the upper surface, lower surface and centre surface; then, considering the torsional deformation effect along the yarn extension direction, the spatial baseline of the upper and lower surfaces are extracted; next, model interpolation is conducted on the upper and lower surfaces to establish warp and weft yarn trajectory prediction models, from which the yarn trajectory prediction models of the original surfaces are further extracted; finally, the spatial curved surface is flattened based on the yarn trajectory prediction models, and the spatial coordinates of interlacing points and thickness mapping data of the composite are obtained synchronously. Experimental results show that the proposed method can effectively predict the yarn trajectories and thickness distribution of various complex structures under different process parameters with satisfactory applicability. This study can provide key support for the design, manufacturing and performance optimization of complex 3D woven composites and promote their engineering applications.

Carbon fiber reinforced polymer (CFRP) composites are widely used in high-end equipment, yet their anisotropic and heterogeneous nature makes machining difficult to predict and control. While numerous reviews have addressed CFRP machining mechanics, damage, or modeling individually, a comprehensive synthesis that bridges traditional physics-driven models and emerging data-driven paradigms is lacking. This review systematically examines the state of the art in CFRP machining mechanisms, damage characterization, and modeling methods, with a particular focus on how material architecture, specifically laminated versus three-dimensional (3D) braided structures, influences modeling strategies. The machining mechanisms and tool wear phenomena are first synthesized as the physical foundation. Physics-driven modeling approaches are then critically reviewed across three scales: macro-scale (force and delamination models), micro-scale (RVE-based failure analysis), and multi-scale coupling methods. A parallel comparison between laminated and 3D braided CFRP is maintained throughout. Subsequently, the emerging paradigm of data-driven and physics-informed modeling is reviewed, with particular attention to the challenges of data quality, model generalizability, and the role of physical constraints in enhancing small-sample predictions. Key findings reveal that laminated CFRP modeling has reached relative maturity, whereas 3D braided CFRP machining models remain underdeveloped due to complex yarn architectures and difficulties in defining representative removal units. Future priorities include developing dynamic machining models for 3D braided CFRP, achieving bidirectional multi-scale coupling, and deepening the synergy between physical principles and data-driven approaches.

With the rapid development and widespread application of textile composites in aerospace, automotive, and civil engineering, braided preforms have become critical structural components that directly influence the mechanical properties and quality of the final product. However, automated defect detection in preforms remains challenging due to low contrast, complex textures, sensitivity to illumination variations, and the difficulty of detecting subtle defects near edges. To address this issue, this paper proposes Braided Preform You Only Look Once (BP-YOLO), a multimodal network based on You Only Look Once version 11 (YOLOv11) for defect inspection in braided preforms. By leveraging dynamic interaction between grayscale and depth features, BP-YOLO achieves enhanced accuracy and robustness under varying illumination conditions. We introduce a novel multimodal fusion module, the Bidirectional Interaction Attention Fusion (BIAF), that enables mutual feature enhancement by dynamically fusing grayscale and depth features via bidirectional spatial attention. Additionally, the Separated and Enhancement Attention Module (SEAM) is incorporated into the neck to suppress background interference and enhance defect regions. Experimental results show the proposed model achieves 87.47% precision, 83.0% recall and 86.53% mAP50, outperforming the baseline by 4.07%, 3.43% and 3.4%. Its inference speed reaches 100 FPS, fully satisfying real-time industrial inspection requirements. Moreover, the model exhibits strong robustness to illumination variations and enhanced sensitivity to edge-localized defects. These results indicate that the proposed method provides a promising solution for reliable and accurate defect inspection in textile composite manufacturing, while offering a novel approach for multimodal feature fusion.

The unique interlayer crosslinking characteristics and highly flexible carrier arrangement configurations of 3D braided structures confer exceptional design flexibility, while correspondingly increasing the complexity of failure mechanisms investigation. This paper proposes a full-process simulation mesoscale modeling approach to investigate the influence of carrier arrangements on the performance of 3D braided composites. A mesoscale model that considers fiber tow compression and layering of the 3D braided structure is generated through a parameterized braiding process model. Coupled with a progressive damage model, the mechanical response and damage evolution of the material under tensile loading are successfully predicted. Experimental validation demonstrated that this model exhibits high accuracy in both stress-strain response and fracture morphology. Based on this validated model, a systematic investigation is conducted on the influence of three different carrier arrangements (S1, S2, S3). Results indicate that while the elastic modulus of materials under different arrangements shows little variation, tensile strength exhibits significant differences. Mechanistic analysis reveals that the symmetry and continuity of surface yarns are the core factors driving this variation. Optimized carrier configurations may effectively regulate the mechanical properties of the composite. The findings of this study provide valuable insights for the structural design of high-performance 3D braided composites.

Atomic-level surface is required by interconnect for copper (Cu) in semiconductor and integrated circuit manufacturing, while the material removal rate (MRR) is usuary sacrificed. To achieve the atomic-level surface of Cu, the MRR is normally lower than 240 nm/min. To overcome this challenge, photocatalytic chemical mechanical polishing (PCMP) was developed for Cu, and the MRR is 379 nm/min. The CMP slurry consisted of silica, lanthana, hydrogen peroxide (H2O2), tartaric acid, Nile blue sulfide (NBS), triazole, and sodium bicarbonate. After PCMP, surface roughness Sa is 0.103 nm measured by atomic force microscopy. To the best of our knowledge, MRR is the highest for the atomic-level surface of Cu reported so far. Transmission electron microscopy reveals that the thickness of the damaged layer after PCMP is 1.69 nm. Under visible light irradiation, the corrosion current density of Cu increases from 57.54 to 72.89 mu A/cm(2) in H2O2 and NBS solution. X-ray photoelectron spectroscopy indicates that the fraction of Cu2+ enhances from 75.12 to 100% with adding NBS in H2O2. The absorption intensity of the fluorescent probe decreases 10.4, 50, and 65% in H2O2, NBS and H2O2 and NBS solution, respectively. Density functional theory (DFT) calculates that the Gibbs free energy of the system reduces 94.46 kcal/mol, revealing that Cu2+ forms coordination bonds with -COO- and -OH groups of tartaric acid. Fourier transform infrared (FTIR) spectroscopy exhibits that the -OH peak of tartaric acid shifts from 3411.35 to 3126.13 cm(-1), attributable to coordination bonding between -OH and Cu2+ ions. It also shows that -COOH deprotonates and transforms to -COO- with the disappearance of the peak at 1743.55 cm(-1) and the appearance of the peak at 1594.08 cm(-1), subsequently complexing with Cu2+ ions via coordination bonding. DFT-calculated results are in good agreement with those of FTIR measurements. Our proposed PCMP paves a route to achieve the atomic-level surface of Cu with a high MRR using visible light irradiation.

This paper investigates the elastoplastic contact behaviors considering the diverse loading modes and proposes a contact force model to describe these behaviors during the contact process. In this study, an expression, composed of the maximum contact force, maximum displacement and exponent a, is first assumed for the contact force-displacement relationships in the mixed elastic-plastic regime with reference to the Hertz contact force model. The peak displacement and contact force are derived as the functions of the exponent a under the law of conservation of energy. The unknown power a is defined under the approximation that the maximum displacement is much larger than the critical elastic displacement, with its initial value determined by analyzing the predictions with various conditions. Similarly, an improved equation for the restitution phase is proposed based on the model of Ma and Liu, with an explicit relation between the coefficient of restitution and the residual displacement. Comparisons with the FEM outcomes and the experimental data under various material properties and loading modes demonstrate the accuracy and adaptability of the proposed contact force model for elastoplastic contact problems.

Fiber patch placement technology automates the production of complex fiber composite structures. This study tackles issues such as uneven pressure during carbon fiber patch placement, which reduces interlaminar shear strength, and the high cost of monitoring end-effector states. A specialized end-effector was designed to improve pressure uniformity. Using simulation data and singular value decomposition, a real-time digital twin reduced-order model (ROM) was developed for state monitoring. An actuator with five evenly spaced airway holes demonstrated optimal pressure distribution. Experimental validation showed less than 10% error relative to simulations, confirming both finite element accuracy and the reliability of the digital twin model.

Two-dimensional triaxially braided composites (2DTBCs) exhibit pronounced anisotropy and complex failure behavior under biaxial loading, where strain-path dependence and fiber coupling lead to nonlinear and asymmetric strength responses. To address the limitations of classical failure models, this study develops a refined mesoscale finite element framework that captures the progressive damage evolution and stress redistribution across interacting fiber systems. Simulations reveal a systematic transition in failure modes—from axial-dominated fracture to coupled axial-transverse damage and ultimately to shear-driven collapse in the bias tows—as strain ratios and axial loading modes vary. Based on these observations, a mechanism-informed, piecewise failure envelope is proposed, integrating a modified Tsai–Wu formulation in coupling regimes with a maximum strain criterion elsewhere. This hybrid approach improves predictive accuracy and enhances physical interpretability for multiaxial strength assessment in complex braided architectures.

The Process–Equipment–In-Process State (PEI) architecture provides a conceptual framework for organizing process manufacturing systems using Process Structures (P), Equipment Entities (E), and In-Process States (I). However, practical application of the PEI architecture requires source engineering information to be systematically transformed into a consistent PEI representation, while explicit object-boundary, mapping, ownership, and structural rules for this transformation have not been established. To address this issue, this paper proposes a PEI-oriented information modeling method for process manufacturing. The proposed method consists of three stages: engineering information organization, PEI structure encoding, and PEI structure construction. Formal definitions, object-boundary criteria, cardinality constraints, and engineering-to-PEI mapping rules are introduced for Process Structures, Equipment Entities, In-Process States, and industrial variables. The encoded information is subsequently assembled into a machine-readable PEI structure through complementary relationship and object-oriented machine-readable representations. The Tennessee Eastman Process (TEP) is employed as a representative benchmark to demonstrate the proposed construction procedure and structurally verify the resulting PEI representation. All 12 identified manufacturing operations, 5 explicitly defined equipment objects, and 53 industrial variables were successfully mapped to valid PEI representations. In addition, all 53 variables satisfied the defined ownership rules, all 188 structural consistency checks were passed, and all 15 source topology relationships were preserved. For the investigated TEP case, these results show that the constructed PEI representation achieved complete coverage of the organized reference information, conformed to the predefined PEI structural rules, and preserved the organized process topology. The proposed method provides a structured information-modeling basis for subsequent PEI-based model construction, software implementation, and future automated PEI structure generation.

Although 3D woven composites have excellent performance, process design for complex components like aero-engine blades is inefficient and inaccurate due to empirical trial and error. This paper proposes a parametric method for thickness extraction and process design of complex 3D woven composites. First, the upper surface, lower surface and central plane of the composite model are extracted, and a predictive model for the warp and weft yarn trajectories on the upper and lower surfaces is constructed to extract the thickness distribution data of the interlacing point area; then, based on the thickness data, the distribution of warp and weft yarn layers and the position distribution of warp and weft yarn reduction points are extracted; furthermore, a weft yarn quantity matrix (WYQM) for each interlacing point is constructed, and a conversion model between WYQM and the jacquard pattern matrix (JPM) is established; finally, the shedding control matrix is extracted, the serialized arrangement of JPM is completed, and process files that can directly drive 3D weaving looms are generated. Experimental results demonstrate that this method adapts to various process parameters and achieves full-process weaving design, providing reliable digital support for the design and optimization of complex 3D woven composites.

Joints constitute critical assembly components in large 3D braided skin structure. Optimizing braided architectures to enhance bearing performance is essential for ensuring connection reliability of joints. Nevertheless, quantitative relationships between braided architectures and bearing performance remain poorly characterized. Moreover, the uniaxial loading in joints specimen tensile test differs significantly from the actual biaxial loading in components, leading to an overrated performance of joint. This study investigates the effect of braiding angle, thickness, edge-distance, and lap configuration on bearing performance to elucidating the effect of braided architecture on unidirectional bearing performance. Then, degradation in bearing performance under biaxial loading was analyzed to propose the strength criteria for biaxial bearing performance. Experiment shows that variations in braiding angle, edge-distance, lap configuration and thickness inducing up to 72% degradation in uniaxial bearing strength and transitions in failure modes. Additionally, biaxial loading of joints in components induces up to 35% degradation in bearing stiffness and up to 50% in bearing strength. The proposed criteria quantify the performance degradation relationship between specimens and components, providing more specific design numerical references for specimen to component design of joints.

In textile production, transparent tape is used for fabric fixing and quality marking, and its identification directly affects the accuracy of automated sorting. Because the tape’s color is similar to the fabric, with low contrast and strong reflectivity, traditional visual methods have difficulty achieving reliable recognition. To solve this, this paper introduces a high-precision transparent tape detection method based on the RT-DETR (Real-Time Detection Transformer) network. The efficient cross-stage partial Darknet53 (ECSPDarknet) is adopted as the backbone network, compressing model parameters significantly while enhancing feature extraction capabilities. The reflection-resistant attention module (RRAM) is integrated during the feature fusion stage to strengthen multiscale feature fusion, effectively solving the recognition challenges caused by the similarity between transparent tape and the background, as well as high reflection. The dynamic group shuffle transformer (DGST) replaces the reparameterization convolutional C3 (RepC3), resolving the high computational load and real-time bottlenecks introduced by the latter’s multibranch structure. In addition, the bounding box regression loss function is replaced with a weighted sum of SmoothL1 and FocalEIoU loss functions, optimizing the model’s convergence efficiency and improving detection accuracy. Ablation experiments were conducted with three sets of random seeds. The results show that the improved model reduces parameters by 43.8%, floating-point operations by 39.6%, and increases FPS by 13.2% compared with the baseline. Precision, recall, F 1-score, mAP@0.5, and mAP@[0.5:0.95] improve by 2.5%, 4.2%, 3.3%, 2.2%, and 1.0%, respectively. Meanwhile, the algorithm outperforms mainstream methods in terms of detection accuracy, providing a foundation for high-precision, transparent tape identification on fabrics.

3D overbraiding can achieve near-net shaping quickly based on the geometry of the mandrel. However, the highly flexible and high-speed dynamic production also introduces the challenge that the full field fiber tows distribution susceptible to shape-dependent variations, leading to the issue that the mechanical properties of composites are highly sensitive to the process configuration and mandrel geometry. This study systematically investigates the influence of mandrel geometry and process configurations on the tensile mechanical behavior and damage evolution of composites through integrated experimental and numerical approaches. A modeling approach toward establishing process-property correlations has been developed to model and predict tensile behavior. This model enables simulation of multiple process configuration throughout the full process chain. A comparison between simulation results and experimental results validates the feasibility and accuracy of the proposed approach. Using different process configurations on the same mandrel, there is a significant variation region versus a plateau region in the axial tensile properties of composites with different braiding angles. Composites braided directly on rectangular-section mandrels show circumferential property gradient phenomena, whereas off-center braiding process results in radial property gradient composites, which enables customized design of properties in different areas on three-dimensional braided composite components.

The morphology of the braided net formed during the laying and braiding processes performed by a hybrid braiding machine significantly affect the formation of the braided fabric and its mechanical properties. However, problems such as yarn accumulation during the braiding process, inability of braided webs to shrink, and yarn skipping remain unsolved. In this study, a dynamics-based braided net model is developed to predict the shape profile of the braided net using the given braiding parameters. In addition, a measurement method based on pixel-by-pixel comparison is proposed for extracting the topography of a braided net using an edge-detection technique. The accuracy of the model is verified by comparing its predictions with the topography of a braided net extracted using edge detection. The proposed model can be used to predict the braiding process and solve the problem of braided nets without shrinking and skipping yarns.

The performance of fiber-reinforced composites depends on the structure and continuity of high-performance fiber. However, fiber discontinuities due to insufficient carrier capacity during the braiding process occur frequently, which is a major obstacle to the high quality of preforms. This work proposes a general topology optimization method of carrier capacity to reduce fiber discontinuity due to yarn change. The movement path of the carrier was calculated. Collision detection of the carrier was performed by an interfering judgment between facets. The genetic algorithm was used to find the optimal structural parameter values of the carrier. To verify the effectiveness of the proposed method, braiding experiments were conducted using optimized carriers. The results show that the optimal carrier capacity of a three-layer spherical braiding machine is 24.3% higher than the carrier currently used. The frequency of yarn changes is significantly reduced, and fiber continuity is effectively ensured. The optimization method can provide reference values of carrier size for different types of braiding equipment, avoiding repeated trial-and-error costs. It can realize high-efficiency production and has great application value.

In response to the growing complexity of modern process manufacturing systems, this paper proposes a novel simulation framework named the Process–Equipment–In-Process State (PEI) simulation method, which introduces a unified and structured approach to modeling multi-stage industrial processes. Unlike conventional simulation approaches that rely on ad hoc or loosely organized modules, the PEI method decomposes the simulation system into three core and interoperable modules: Process Structure (P), Equipment Behavior (E), and In-Process State (I). This modular abstraction facilitates the decoupling of model logic. It also enables a structure-driven simulation execution mechanism. In this structure, the process topology governs task scheduling; equipment models translate control inputs into physical conditions; and state models simulate material evolution accordingly. A complete simulation case involving water mixing, heat exchange, and slurry transformation demonstrates the method’s capability to support traceable state evolution, logical task flow, and extensible model binding. The results demonstrate that the proposed method enables module decoupling, clear simulation pathways, and traceable state changes, providing effective support for structured modeling and behavioral evolution analysis in process manufacturing.

This paper establishes a physical model for the non-contact rotary screen coating process based on a spacecraft structural plate and proposes a theoretical expression for the adhesive thickness of the non-contact rotary screen coating. The thickness of the adhesive is a critical factor influencing the quality of the optical solar reflector (OSR) adhesion. The thickness of the adhesive layer depends on the equivalent fluid height and the ratio of the fluid flow rate to the squeegee speed below the squeegee. When the screen and fluid remain constant, the fluid flow rate below the squeegee depends on the pressure at the tip of the squeegee. The pressure is also a function related to the deformation characteristics and speed of the squeegee. Based on the actual geometric shape of the wedge-shaped squeegee, the analytical expression for the vertical displacement of the squeegee is obtained as the actual boundary of the flow field. The analytical expression for the deformation angle of the squeegee is used to solve the contact length between the squeegee and the rotary screen. It reduces the calculation difficulty compared with the previous method. Based on the theory of rheology and fluid mechanics, the velocity distribution of the fluid under the squeegee and the expression of the dynamic pressure at the tip of the squeegee were obtained. The dynamic pressure at the tip of the squeegee is a key factor for the adhesive to pass through the rotary screen. According to the continuity equation of the fluid, the theoretical thickness expression of the non-contact rotary screen coating is obtained. The simulation and experimental results show that the variation trend of coating thickness with the influence of variables is consistent. Experimental and simulation errors compared to theoretical values are less than 5%, which proves the rationality of the theoretical expression of the non-contact rotary screen coating thickness under the condition of considering the actual squeegee deformation. The existence of differences proves that a small part of the colloid remains on the rotary screen during the colloid transfer process. The expression parameterizes the rotary screen coating model and provides a theoretical basis for the design of automatic coating equipment.