The development and application of Automated Fiber Placement (AFP) technology have laid a solid foundation for the large-scale application of composite materials in aircraft manufacturing. However, layup defects are inevitably present during the AFP process, directly affecting the performance of the manufactured composite products. To address this problem, this paper creates an Automated Fiber Placement Defect Detection Network (AFP-Net) for implementation in domestically manufactured gantry-type AFP machines. AFP-Net is an end-to-end object detection framework that integrates three key innovations: a novel multi-scale feature fusion convolutions module, an adaptive feature pyramid network with weighted multi-scale information aggregation, and a lightweight detection head specifically tailored for AFP defect detection. Specifically, this study introduces three innovative designs for AFP defect detection: (1) a Multi-scale Feature Fusion with Convolutions (MFFC) module that employs multi-scale convolution kernels without dilation to extract features at different scales and capture local context; (2) a Gap-foreign Aimed Detection Feature Pyramid Network (GADFPN) that enhances multi-level feature utilization through efficient weighted fusion; and (3) a Gap-Aimed Detection (GAD) Head that uses shared convolutions instead of redundant independent ones, improving gap defect detection accuracy while maintaining computational efficiency. Experimental results demonstrate that the proposed AFP-Net exhibits excellent performance in defect detection in the AFP of composite materials. Moreover, comparison experiments on the publicly available surface defect dataset validate that AFP-Net outperforms YOLO and RT-DETR, demonstrating that AFP-Net has outstanding generalization ability.

Polymer zippers are widely used in consumer and industrial products, yet their performance is governed by complex contact-dominated interactions between molded teeth, sliders, and stitched joints, which are difficult to assess prior to manufacturing. This study develops a process-informed finite element framework that links polymer manufacturing and assembly parameters to product-level mechanical performance of PET zippers. The framework integrates strain-energy-guided adaptive mesh refinement, a velocity-dependent friction formulation tailored to slider-tooth nonlinear interactions, and a multiscale representation of molded teeth and stitched connections. Experimental testing supports the predictive capability of the model, with relative deviations within 2.40%, while computational time is reduced by 26.92% compared with uniform meshing. A parametric study indicates that small geometric deviations can noticeably affect interlocking efficiency and global load transfer. Overall, the proposed framework provides a practical virtual verification tool for high-performance zipper product development and supports simulation-assisted design of process-sensitive polymer zipper assemblies.

Aero-engines typically operate under time-varying conditions, which obscure degradation patterns in sensor data and significantly increase the difficulty of Remaining Useful Life (RUL) prediction. Approaches based on Graph Neural Networks (GNNs) extract spatial-temporal features by constructing homogeneous graphs and refining node feature update mechanisms. However, these studies do not adequately model how operating conditions influence sensor data, thereby weakening the methods' robustness. Therefore, we propose an Operating Condition-Aware Heterogeneous Graph Neural Network (OCA-HGNN) for RUL prediction of aero-engines under time-varying operating conditions. First, we introduce an expert knowledge-driven method for constructing a condition-sensor heterogeneous graph that explicitly accounts for the engine's mechanical structure and operating principles. Then, a Spatial-Temporal and Operating Condition-Aware Message Passing (STOCA-MP) mechanism is developed to update node features and select edges. STOCA-MP computes and fuses the central node's historical message, spatial messages from neighboring nodes at the same time step, and influence messages from operating-condition nodes, while identifying the edges that provide more information to the central node. Finally, STOCA-MP is embedded into neural network layers to form OCA-HGNN for RUL prediction. Experiments on the C-MAPSS and N-CMAPSS datasets demonstrate that the proposed method is effective for spatial-temporal feature extraction, modeling the influence of operating conditions, and aggregating multisource messages, and it achieves high prediction accuracy under time-varying operating conditions.

Operating conditions of aero-engines often change over time, complicating the prediction of their remaining useful life (RUL). Most RUL prediction methods overlook the impact of time-varying operating conditions on feature extraction and fail to account for the potential interference of operating condition information in sensor data features. Additionally, deep-learning-based methods lack sufficient interpretability. To address these issues, we model sensor data as a coupling of operating condition and equipment degradation information, and propose an interpretable model named variational attention-weighted feature decoupling network (VAFD-Net) for RUL prediction of engines under both discrete and continuous time-varying operating conditions. VAFD-Net separately extracts operating condition features and sensor data features, then uses variational attention weights representing operating condition information to weight sensor data features, improving the model’s robustness. VAFD-Net also introduces three constraints to decouple operating condition and degradation information in latent space, mitigating the impact of signal non-stationarity. Experimental results on the Commercial Modular Aero-Propulsion System Simulation (CMAPSS) and N-CMAPSS datasets indicate that VAFD-Net not only achieves high prediction accuracy under time-varying operating conditions but also reveals the contribution of physical quantities to the engine’s health states through weighted feature maps. Furthermore, latent variable distribution plots enable users to directly infer the RUL based on the position of the latent variables.

This study investigates the fatigue behavior of aircraft countersunk riveted joints. Riveting tests and their numerical counterparts were conducted to explore the effects of hole diameter and squeeze force on the driven head dimensions and the distribution of interference magnitude. Fatigue tests were performed to analyze the fatigue behavior, including fatigue life, failure mode, etc. The microstructural characteristics of the fractures on the rivets and plates were analyzed to observe and measure the distribution and chemical composition of fretting debris, which effectively revealed the mechanisms of fatigue failure and fretting of the riveted joints. The results demonstrate that appropriately increasing the squeeze force and reducing the hole diameter can significantly improve the fatigue life of the riveted structure. There are two failure modes during fatigue: rivet fracture and outer plate fracture. Cracks were observed on both the rivets and outer plates during the fatigue failure, indicating a competitive relationship between their crack propagation. Fretting wear on the contact surfaces of the inner and outer plates initiates fatigue cracks. The fretting debris contributes to the initiation and propagation of the cracks, which expand along the thickness and width directions of the riveted joints.

Aero-engine nacelle acoustic liners are complex curved surface subassemblies with tens of thousands of dense acoustic holes for noise reduction. Traditional robotic drilling systems with a single spindle and conventional teaching programming cannot meet the high-quality and high-efficiency drilling requirements for nacelle acoustic liners. This paper developed a novel robotic multi-spindle drilling system, which integrates standard industrial robots with a multi-spindle drilling end-effector to drill acoustic holes on an acoustic liner. The task planning strategy and offline programming method are investigated for the robotic multi-spindle drilling of the acoustic liner. A dedicated offline programming software capable of quickly generating robotic machining programs has been designed and developed. A precise acoustic hole arrangement applicable to curved drilling zones is demonstrated. The virtual surrogate hole is introduced to target the array layout of the developed end-effector's multiple spindles. A virtual surrogate hole generation method is proposed based on the minimum mutually exclusive set cover and the greedy algorithm. After that, a virtual surrogate hole layout optimization method is developed using an adaptive genetic annealing algorithm. Then, the frame chain is constructed as the kinematic foundation in robotic multi-spindle drilling of a nacelle acoustic liner. The drilling pose planning of the multi-spindle end-effector based on normal vectors and coordinates of acoustic hole points covered by virtual surrogate holes is explored. And a multi-spindle drilling path planning algorithm is developed using the genetic algorithm. To drill the aero-engine nacelle of a large aircraft, an offline programming software for robotic multi-spindle drilling of the nacelle acoustic liner is developed by integrating the methods above. A case study of multi-spindle drilling task planning of a nacelle acoustic liner is carried out. The case study results indicate that, compared to traditional single-spindle robotic drilling systems, the developed multi-spindle drilling system increases drilling efficiency by approximately 347 %. Using the proposed multi-spindle drilling path optimization algorithm, the drilling path has been shortened by 84.8 %. These results validate the effectiveness of the proposed series of drilling task planning methods and the developed offline programming software, which significantly enhance drilling efficiency while meeting the technical requirements for drilling acoustic holes on the acoustic liner.

Laying acoustic liners is one of the most effective ways to reduce aero-engine noise. Traditional single-spindle drilling systems are inadequate for high-quality and efficient drilling for large-scale sound-absorbing holes on acoustic liners. Therefore, this paper develops a robotic multi-spindle drilling cell (RMDCell) that integrates a multi-spindle drilling end-effector, a standard industrial robot, and a rotatable clamping fixture. A robot performance-oriented layout optimization method is studied for the RMDCell to improve the drilling quality of acoustic liners in aero-engine nacelles. Firstly, a weighted comprehensive performance model of the multispindle drilling robot is built, integrating joint limit avoidance, singularity avoidance, and robot stiffness performances. Thereinto, a new robot stiffness performance index is introduced, and a variable robot stiffness weight is derived considering the end-effector's spindle configuration. Secondly, a dimensionality reduction method of layout optimization parameters and a drilling zone division method are proposed for the RMDCell of nacelle acoustic liners. A bi-level optimization algorithm for RMDCell layout parameters is developed. The algorithm fuses particle swarm optimization (PSO) and deep hierarchical feature learning on point sets in a metric space (PointNet++). Finally, drilling experiments are conducted based on the developed RMDCell. The results show that the comprehensive robot performance index is reduced by 39.8 % after optimization compared to the comprehensive robot performance before layout optimization. The layout optimization method effectively reduces defects such as burrs, delamination, and tearing in composite sound-absorbing holes. The average delamination factor decreases by 16.5 %. The robot performance-oriented layout optimization significantly improves the drilling quality, which meets the manufacturing quality requirements of the acoustic liner of an aero-engine nacelle.

Remaining Useful Life (RUL) prediction is a crucial task in Prognostics and Health Management (PHM) of aero-engines, aiming to improve reliability and reduce maintenance costs. However, existing deep learning methods overlook the differences in feature distributions between healthy and degraded states, assuming that they follow the same simple distribution, which limits model’s flexibility and interpretability. To address these challenges, we propose SADF-VNet, a State-Aware and Dual-Flow Variational Network for interpretable RUL prediction. Specifically, SADF-VNet separately models healthy and degraded states through private invertible flows, while maintaining structural consistency and probabilistic alignment via shared layers and a common Gaussian prior. In this model, a local-global feature extractor captures multi-scale patterns, and an adaptive fusion module integrates them based on inter-feature correlations. After reparameterization, a state classifier predicts each latent variable’s degradation probability and assigns it to a dual-flow transformation: a lightweight affine flow for healthy samples and a spline-based flow for degraded samples, introducing greater distributional complexity. Finally, the latent representations are decoded by a regressor for RUL estimation, and the latent variable’s distribution provide intuitive RUL assessment. Experimental results on the C-MAPSS dataset demonstrate that SADF-VNet achieves both high prediction accuracy and interpretability.

The large-scale tiny acoustic holes densely distributed on acoustic liners are essential in aero-engine noise reduction. Accurate segmentation of those holes is fundamental for a robotic multi-spindle drilling system. This paper introduces a novel semi-supervised segmentation method for acoustic holes on composite acoustic liners. This method uses perturbation consistency loss to ensure output consistency while solving the problem of data volume imbalance. Afterward, a multi-reliability enhancement method and a pseudo-label reliability enhancement module are utilized to enhance the model’s robustness and accuracy. The segmentation experiments show that our method is superior to UniMatch and FixMatch. When using only 30% of labeled data, our method achieves an IoU of 96.39%, and the porosity differs only 0.038% from the ground truth, which is better than the fully-supervised segmentation method using all data. Our method can meet the accuracy and efficiency requirements of aero-engine nacelle production in China with only limited labeled data.

A new method, to our knowledge, is proposed to achieve high-precision measurement of parallel edge spacing for sheet metal parts in the complex industrial environment of aviation manufacturing. First, the sub-pixel edges of sheet metal parts are extracted by combining a what we believe to be a novel adaptive rolling bilateral filter and a sub-pixel edge detection algorithm based on the Canny–Steger algorithm. Then, the acquired edge data are denoised by using the clustering algorithm. Finally, a parallel line fitting algorithm, which combines an improved K-medoids algorithm with composite constraints on points and slopes, is proposed to calculate the parallel edge spacing. The results show that the method is robust to introducing noise in the edge data due to uneven illumination and various types of defects such as wear, scratches, and stains. The detection accuracy is high, with an average detection error of only 0.015 mm.

为了实现复杂环境下航空零件孔特征的高效高精度检测，提出了一种集成视觉显著性和群决策的检测方法。在经典FT显著性检测算法中引入图像增强步骤，并为每个像素赋予以最大显著区域中心为参考的权重，使用改进后的方法对图像进行孔区域分割。设计具有多尺度多结构元素的新型数学形态学边缘检测算法，结合轮廓细化算法对孔区域进行轮廓提取。最后，利用Meanshift算法寻找轮廓点的圆心位置，建立新的基于群决策的圆半径计算模型，获得孔特征的关键几何参数。结果表明：改进的视觉显著性特征检测算法能够生成更加突显孔特征的全分辨率显著图；新型数学形态学边缘检测算法能获得简化且可靠的轮廓点；该方法在不均匀光照、各类孔缺陷和孔内壁干扰等条件下均显示出较好的稳定性；即使在噪声密度高达30%时仍能成功完成孔检测，且圆心坐标和半径的误差均小于0.012 mm；平均检测时间仅为0.236 s。该方法能够在复杂环境下对航空零件孔特征进行准确、稳定的检测。

The stack-ups of composite and metal are used in aircraft manufacturing to produce modern high-performance aircraft, which makes fastener-hole-making more difficult. Helical milling is an advanced machining process to improve the quality of hole-making. This paper seeks to control the hole diameter error and guarantee fastener-hole fitting performance. The setup of automated helical milling for assembling composite/metal aerostructures is first presented. Meanwhile, the helical milling procedure in aircraft wing manufacturing is discussed. Based on the designed helical milling end-effector, a length-gauge-based feedback control mechanism of the orbital eccentricity is designed to reduce the diameter error of an intended fastener hole. For accurate measurement, a least-squares method is used to find the calibration parameters of the length gauge. Then the mathematical model of hole diameter variation extraction based on B-spline curve construction is given for further variation compensation. After that, the feedback control system of the radial feed mechanism is established. Besides, the modeling and compensation system of the hole diameter variation is developed and used. In helical milling of CFRP/Ti stacks in an aircraft assembly, fastener holes tend to achieve the target diameter of 7.9400 mm without extending the range of hole entrance and exit diameters.

降低车辆碳排放是实现碳达峰和碳中和的重要任务,轻量化设计是实现这一目标的关键技术手段.为了减少典型工程机械叉车的C02排放量,原创性地将尺寸优化和自由形状优化技术进行融合,提出一种面向高维设计空间的叉车货架轻量化设计.建立叉车货架有限元仿真模型,通过静力学分析得到应力和位移数值,并将其作为联合优化过程中应力和位移约束条件的设置基准;采用基于试验设计的灵敏度分析方法,研究叉车货架质量、位移和应力等试验指标对各设计变量变化的影响;结合所提出的联合优化模型与灵敏度分析,构建设计变量的高维设计空间,综合应力和位移约束,进行叉车货架轻量化设计.优化结果表明,与常用的逐步优化技术相比,尺寸和自由形状联合优化技术具有更大的优化求解空间,同时可以以更小的刚度和强度性能损失,实现更优的轻量化效果.

Aircraft assembly demands efficient and accurate drilling and fastening of a large number of complex thin-walled aircraft parts. In this paper, a robotic assembly system is developed. The machine design and component functions of the robotic system are first discussed. Subsequently, the formation of the relative positioning error between the end-effector and an aircraft part is analyzed and modeled. Measuring instruments are used for relative error measurement to compensate for drilling positions. To achieve high positioning accuracy, we theoretically compare relative positioning errors across instrument configurations from a tolerance management perspective to recommend a good hand-eye configuration. Besides, the impact of hand-eye offset on positioning accuracy is explored based on positioning error Jacobian to guide the end-effector design and the setting of the vision coordinate system. After that, the on-machine hand-eye calibration and system positioning methods are proposed and modeled, eliminating the effect of the dissimilarity of the robot's absolute positioning error. For positioning and processing, the control system of the multifunctional robotic assembly system is built. Finally, experiments for accurate positioning are conducted on the robotic assembly system developed to assemble an aircraft. With the recommended hand-eye configuration, the minimized hand-eye offset, and the on-machine hand-eye calibration, the maximal positioning error is 0.08 mm, which can adequately meet the accuracy standard of the aircraft assembly.

Composite/metal wing-boxes are hard to drill by the conventional drilling method. In this paper, an automated machine-tool-based machining system equipped with a multifunctional end-effector is developed for helical milling, as well as circular and ellipsoidal dimple drilling of composite/metal wing-boxes. The system constitution, the control system, and the application software of the machining system are demonstrated. In aircraft manufacturing, the out-of-tolerance pose and shape of fastener holes deteriorate an aircraft's strength and fatigue performances. Thus, a multi-sensor information fusion and integration architecture is built for enhanced pose and shape accuracy of machined fastener holes. The relative pose error of the end-effector and an aerostructure is analyzed. A coarse-to-fine control method by the touch probe and LDSs is applied for controlling the error. The fastener holes' shape errors are controlled through the elaborate structure design and multi-sensor-based closed-loop motion control of the end-effector. Experiments performed on the machining system show that the positioning errors, hole diameter errors, dimple depth errors, and ellipsoidal dimple shape can fulfill the wing-box assembly requirements.

To reduce downstream rework and design changes, variation modeling and analysis are indispensable in the assembly of complex products. In this paper, a rigid-compliant hybrid variation analysis method using the Monte Carlo interval approach is developed to assembly ladder structures, such as the skeleton of a horizontal stabilizer or a wing box. We first present the classical locating scheme of a low-rigidity aeronautical structure, and the contributors to the assembly variation of a ladder structure comprising locating errors and part geometric errors. Assembly variations induced by rigid-body locating errors and part geometric errors are mathematically modeled with rigid-body kinematics and the mechanistic method based on the Finite Element Analysis, respectively. And then, the two types of assembly variations are integrated into a rigid-compliant hybrid variation model. Probability distributions of the contributors are often unknown, especially in aircraft manufacturing with low production volume. Therefore, a novel variation analysis method using the Monte Carlo interval approach is proposed to compute the assembly variation, represented in the form of interval structural parameters. The assembly case of a scale wing skeleton shows the proposed rigid-compliant hybrid variation analysis method is efficient in the assembly variation analysis for low-rigidity aircraft structure.

Strict quality requirements in aircraft manufacturing demand high-accuracy pose adjustment systems. However, the pose alignment process of a large complex structure is also affected by thermal and gravity deformations to a great extent. Even though the pose adjustment system passes accuracy verification, the pose of the large complex structure remains challenging to smoothly and efficiently converge to the desired pose. To solve this problem, we developed a pose adjustment system enhanced by integrating physical simulation for the wing-box assembly of a large aircraft. First, the development of the pose adjustment system, which is the base of the digital pose alignment of a large aircraft's outer wing panel, is demonstrated. Then, pose alignment principles of duplex and multiple assembly objects based on the best-fit strategy are successively explored. After that, contribution analysis is conducted for nonideal pose alignment. Immediately following, influences of thermal and gravity deformations simultaneously coexisting for the pose alignment are discussed. Finally, a physical simulation-assisted pose alignment method is developed considering multisource errors, which uses the finite element analysis to integrate temperature fluctuation and gravity field effects. Compared with a conventional digital pose adjustment system driven by the classical best-fit, deviations of the key characteristic points significantly decreased despite the impacts of thermal and gravity deformations. The enhanced pose adjustment system has been applied to large aircraft wing-box assembly. It provides an improved understanding of the pose alignment of large-scale complex structures.

Thickness control is a critical process of automated polishing of large and thin Si wafers in the semiconductor industry. In this paper, an elaborate double-side polishing (DSP) system is demonstrated, which has a polishing unit with feedback control of wafer thickness based on the scan data of a laser probe. Firstly, the mechanical structure, as well as the signal transmission and control of the DSP system, are discussed, in which the thickness feedback control is emphasized. Then, the precise positioning of the laser probe is explored to obtain the continuous and valid scan data of the wafer thickness. After that, a B-spline model is applied for the characterization of the wafer thickness function to provide the thickness control system with credible thickness deviation information. Finally, experiments of wafer-thickness evaluation and control are conducted on the presented DSP system. With the advisable number of control points in B-spline fitting, the thickness variation can be effectively controlled in wafer polishing with the DSP system, according to the experimental results of curve fitting and the statistical analysis of the experimental data.

Modern aircraft assembly demands assembly cells or machines with higher machining efficiency and accuracy. Thus, a dual-machine drilling and riveting cell is developed in this paper. We firstly discuss its physical design, as well as the automatic drilling and riveting process. With the automatic drilling and riveting cell, drilling and riveting production line of aircraft panels can be expected. The frame chain of the drilling and riveting cell is constructed to link the assembly cell to its task space, which is the kinematics base. System calibrations, including task space calibration, the sensor calibration of an orientation alignment unit, the floating calibration of the implicit hand-eye relationship, are explored. For high positioning accuracy, a multi-sensor servoing method is proposed for cell positioning. An orientation-based laser servoing strategy, which uses the feedback of the orientation errors measured by laser displacement sensors, is used to align drilling direction and camera shooting direction. Besides, A single-camera-based visual servoing is applied to align the tool center point (TCP) to reference holes, to obtain their coordinates for drilling position modification. Experiments of multi-sensor servoing for cell positioning are performed on an automatic drilling and riveting machine developed for the panel assembly of an aircraft in China. With the cell positioning method, the automatic drilling and riveting cell can approximately achieve an accuracy of 0.05 mm, which can adequately fulfill the requirement for the assembly of the aircraft.

Tight position tolerance is required for fastener holes in wing manufacturing. Automated drilling system with high positioning accuracy is the key to achieve the requirement. The paper seeks to determine allowable values of variation sources and guarantee the hole position tolerance. The process of reference hole positioning and the compensation of drilling positions are firstly explored and formalized for an automated drilling system integrated with an industrial camera. Based on this, a positioning variation model for automated drilling considering positioning error measurement and compensation is built. After that, positioning variation synthesis being imposed engineering constraints on is mathematically modeled based on the theory of mathematical statistics. In the positioning variation synthesis, imperfect camera installation, nonideal measurement conditions, equipment positioning error, etc. are included. The positioning variation model and involving synthesis strategy have been used to develop an automated drilling system for wing assembly. Experiments conducted on the developed drilling system show that the fastener holes' desired position tolerance 0.3 mm will not be exceeded, which is a necessary condition of the satisfactory drilling quality of the aircraft wing.