The efficient extraction and precise measurement of skin seam features are crucial for surface morphology control and optimization in the manufacturing of next-generation aircraft. Visual measurement methods are efficient but cannot extract complete seam features for subsequent surface morphology analysis. Methods based on 3D point cloud have extremely low feature extraction efficiency and cannot quickly extract seam features from large amounts of redundant data. Therefore, this paper combines the advantages of both methods and proposes an extraction and measurement method for aircraft skin seam based on the correlation mapping between point cloud and image. First, the skin scanning point cloud is converted into a 2D image through point cloud projection, grid division of the projection plane, and pixel value determination. Second, an improved U-Net network is used to detect seam regions in the image, and then the corresponding skin data points for the seam region pixels are found based on the mapping relationship equation between the pixels and the skin projection points, so as to obtain the seam feature point cloud. Finally, measurement points are selected in the seam point cloud based on the seam shape, and the flush and gap for each measurement point are calculated. The validation shows that this method can quickly and accurately extract different types of seam features and achieve comprehensive and precise parameter measurement of the seam. The expanded uncertainties for the seam flush and gap measurements are 0.05048 and 0.06766 mm, respectively, with repeatability of 0.0059 and 0.00728 mm, both of which meet the requirements for practical engineering applications.

In the context of intelligent manufacturing, quality control challenges on manual assembly lines have become increasingly pronounced. Human factors often lead to process omissions and deviations from prescribed workflows, resulting in missing assembly steps and thereby compromising the consistency of product assembly quality. To address these issues, this paper proposes a transient action recognition framework to enable standardized, real-time monitoring of the assembly process on production lines. First, we introduce the Transient Action Dataset (TAD), which was captured in real industrial scenarios to provide a reliable data source for training transient action recognition models. A data preprocessing module was also developed to reduce dataset annotation costs while supplying lightweight data input for the recognition system. Finally, we propose the TANet model for transient action recognition, which integrates the ECA module, MHSA module, and FPN architecture into the ResNet50 backbone. This design facilitates efficient extraction of fine-grained features in transient actions. Experiments conducted in real-world scenarios demonstrate that the model achieves a training accuracy of 92.86%, a recognition accuracy of 99.1% in deployment, and an inference latency of 25.6 ms. These results serve as an effective technical foundation and a robust visual recognition framework for transient action recognition, paving the way for the broader application of action recognition technologies in industrial domains.

Thermal protection systems (TPS) are critical for reusable hypersonic vehicles, but assembly deviations due to manufacturing errors and complex curved surfaces remain challenging. This work proposes a digital twin-based assembly coordination method to precisely control TPS tile alignment. A virtual-real mapping is established via a binocular vision system, synchronizing the physical tile's pose with its virtual model to within 0.05 mm accuracy. A strain-driven compression model for the strain isolation pad dynamically generates pressure compensation commands, reducing the tile-to-body fit gap from 0.53 mm to 0.21 mm (a 60% improvement). Key innovations include fully automated deformation compensation, real-time posture monitoring, and closed-loop feedback optimization-capabilities beyond traditional methods. Experimental results verify that the proposed digital twin framework significantly improves assembly consistency and accuracy. This study thus provides an innovative high-precision assembly solution for TPS in hypersonic spacecraft.

In digital-measurement-assisted assembly of large aircraft components, the spatial layout of Enhanced Reference System (ERS) points determines coordinate transformation accuracy and stability. To address manual layout limitations—specifically low efficiency, occlusion susceptibility, and physical deployment limitations—this paper proposes an adaptive planning method under engineering constraints. First, based on the Guide to the Expression of Uncertainty in Measurement (GUM) and weighted least squares, an analytical transformation sensitivity model is constructed. Subsequently, a multi-scale sample library generated via Monte Carlo sampling trains a high-precision BP neural network surrogate model, enabling millisecond-level sensitivity prediction. Combining this with ray-tracing occlusion detection, a weighted genetic algorithm optimizes transformation sensitivity, spatial uniformity, and station distance within feasible ground and tooling regions. Experimental results indicate that the method effectively avoids occlusion. Specifically, the Registration-Induced Error (RIE) is controlled at approximately 0.002 mm, and the Registration-Induced Loss Ratio (RILR) is maintained at about 10%. Crucially, comparative verification reveals an RIE reduction of approximately 40% compared to a feasible uniform baseline, proving that physics-based data-driven optimization yields superior accuracy over intuitive geometric distribution. By ensuring strict adherence to engineering constraints, this method offers a reliable solution that significantly enhances measurement reliability, providing solid theoretical support for automated digital twin construction.

High-quality aircraft assembly is crucial in modern aviation manufacturing. Assembly simulation methods are used to evaluate the assembly feasibility of components, but the accuracy of simulation models and the reliability of analysis results still need to be improved. Therefore, this paper introduces reconstruction accuracy regression based on the NURBS algorithm and adds iterative updates of the reconstruction base surface to achieve high-precision reconstruction of the simulation model. Then, based on the shape and dimensional parameters of the assembly feature simulation model, an uncertain region of the assembly pose is constructed. The Monte Carlo sampling method is used to sample the pose, generating a model of the pose uncertainty after the monitored feature is assembled, which is used to evaluate the assembly feasibility of the component. Finally, the experimental results show that the proportion of points with reconstruction accuracy within 0.005 mm is 96

In the process of aircraft measurement-assisted assembly, the laser beam path between laser trackers and target points is prone to obstruction by large components and complex tooling, making it difficult to meet assembly requirements in terms of measurement accuracy and efficiency. This paper proposes a spatial accuracy-driven adaptive laser tracker station planning method based on Monte Carlo and genetic algorithm-backpropagation (GA-BP) neural networks to address this issue. First, the measurement uncertainty of laser trackers was analyzed, and a GA-BP neural network-based rapid uncertainty prediction model was established. The analysis efficiency improved by 100 times compared with Monte Carlo methods when processing million-level data points, while maintaining prediction accuracy consistently within 0.001 mm. Subsequently, a spatial accuracy-driven adaptive laser tracker station planning method was developed. Integrated with the rapid uncertainty prediction model and measurement obstruction detection algorithms, this method adaptively plans measurement stations. Compared with traditional station planning approaches, it demonstrates at least double the efficiency while achieving the minimum number of stations and maximum target point coverage. Finally, experimental verification was conducted in a simulated aircraft wing-body docking scenario. Results show that the average difference between simulated and actual measurement uncertainties is less than 0.002 mm, satisfying the requirement of 0.01 mm discrepancy. The fitting residuals of target points under two measurement stations all remain within uncertainty thresholds and below actual precision constraints. The proposed laser tracker station planning method proves effective in achieving high-efficiency and high-precision measurement requirements for aircraft assembly.

When facing large-size targets, for the problem that the target imaging field of monocular camera is small and the target imaging is not clear at a long distance, this paper proposes an uncalibrated visual servoing method using multiple cameras for large targets. On the basis of image-based visual servo control, an uncalibrated visual servo model of an arrayed multiple cameras is constructed, and the nonlinear mapping relationship between image features of multi-view fusion and robot end six-degree-of-freedom motion velocity is constructed by using image moments. Visual servoing operation for large-size targets by predicting robot end-motion velocity using a neural network based on genetic particle swarm optimization without calibrating the camera parameters and hand-eye relationship. The experimental results show that the proposed method can effectively realize multi-view fusion visual servoing operation for large-size targets with high operation accuracy and robustness.

In the digital measurement of aircraft assembly,the accuracy of large-size measurement field construc-tion is highly dependent on the stability of the reference points laid on the tooling.The position of the reference points of large-sized tooling is very susceptible to thermal drift due to changes in ambient temperature,leading to a reduction in the accuracy of the measurement field or even failure.Therefore,this paper takes a combined large-scale tooling as an example to construct a numerical model for predicting the thermal drift of the reference points of large-scale tooling under a non-uniform temperature field;constructs a proxy model for the thermal drift of the tool-ing based on a large amount of thermal drift data obtained from the simulation of the aforementioned model using BP neural network;and formulates a program for improving the accuracy of the measurement field based on the aforementioned proxy model.The temperature and coordinate measurement data collected in the field at the refer-ence points of the tooling are used to verify the validity and correctness of the proxy model,and the temperature-co-ordinate drift data at the reference points obtained from the model are compared and analyzed.The results show that the average relative errors of the simulation results are below 18%,and the average relative errors of the BP neural network results are below 26%,which can effectively improve the measurement field construction accuracy.

The airborne system plays a crucial role in upholding the operational capabilities of an aircraft, and the caliber of its products significantly influences the overall reliability and safety of the aircraft. The foundational task in guaranteeing product quality is metrology, wherein metrological parameters play a pivotal role in shaping the ultimate product quality and determining metrological efficiency. Because the metrology documents are scattered in various departments and the data are highly isolated, manual planning methods are highly subjective, resulting in error-prone and inefficient results. The existing parameter planning methods based on deep learning suffer from the problems of strong training data dependency and poor model interpretability. For this reason, this paper proposes a metrological parameter planning method based on a multi-head sparse graph attention network (MHSGAT) for airborne products. First, an attribute graph structure is designed to represent product metrology information, and a metrology parameter graph dataset is constructed based on historical documents. Then, the proposed MHSGAT assigns sparse attention coefficients to the neighbors of nodes for feature aggregation. Finally, the metrology parameter planning results are output in the dense layer. The experimental results show that the proposed method outperforms the other four baseline methods. The significance of the proposed method in enhancing the metrological accuracy of airborne products and ensuring their quality is demonstrated by application cases.

The thermal protection tile system is one of the most critical structures of reusable launch vehicles, and its assembly quality directly affects the flight performance of the aircraft. The assembly clearance and step-off of thermal protection tiles are essential parameters for assessing the assembly quality of the thermal protection system (TPS). Additionally, digital twin technology provides real-time feedback and the ability to control physical processes virtually. Therefore, this paper proposes a novel digital twin-based evaluation method of clearance and step-off for TPSs. The paper first introduces the automated assembly system for thermal protection tiles used in this study, which consists of three assembly stages. Based on this system, a digital twin model of thermal protection tiles is constructed using a theoretical model and physical entity as inputs through three layers: theoretical data layer, actual measurement data layer, and reconstruction of data layer. This twin model is coupled with the assembly pose dataset of the thermal protection tiles to create an assembly deviation model, which is then used to calculate the assembly clearance and step-off. Finally, the paper also experimentally verifies the proposed measurement method by building an experimental platform using seven thermal protection tile test pieces. The standard values of the clearance and step-off are measured using a laser tracker, and these standard values are compared with the measurements obtained from the proposed method. The results indicate that this method significantly enhances measurement efficiency and realizes the real-time measurement while maintaining accuracy within +/- 0.027 mm.

The parameter traceability chain is essential for maintaining the controllability, consistency, and accuracy of performance parameters throughout the lifecycle of aviation products. Aviation enterprises have amassed substantial parameter traceability data, and enhancing the reuse of this data is critical for achieving efficient digital manufacturing. However, the current approach to establishing metrology parameter traceability relationships relies heavily on the experience of metrology personnel and manual processes. This method often results in redundant work when dealing with similar products and parameters, leading to inefficiencies and increased risk of errors. To this end, this paper proposes an entity matching-based method for reusing parameter traceability chains in aviation products. Initially, a novel entity-matching model utilizing a Siamese neural network is designed. This model transforms text entities into semantic vector representations using word2vec and determines if entity pairs match via a multilayer fully connected neural network. Subsequently, based on the matched parameter entities and the corresponding product information, the historical parameter traceability chains are reused for newly developed products. The experimental results demonstrate that the proposed entity matching model outperforms the five baseline models. Moreover, the case study confirms the method’s effectiveness in reusing aviation product parameter traceability, offering engineers a fast and accurate reference for parameter traceability. This approach is crucial for improving aviation products’ manufacturing efficiency and quality.

To address the low efficiency and lack of precision in existing methods for extracting rivet unevenness features from three-dimensional point clouds,we propose an accurate extraction method based on point cloud dimensionality reduction.First,the surface variant of the point cloud is calculated,and the surface normal variant is color-mapped through mathematical processing.Next,principal component analysis projection technique combined with two-dimensional meshing is used for point cloud dimensionality reduction,generating an image,and the relevant image processing algorithms are used for the coarse segmentation of the point cloud in the rivet region.Finally,hierarchical structure fitting is applied to accurately extract the rivet head and skin point clouds,allowing for the calculation of rivet unevenness.Experimental results show high computational efficiency with large point cloud datasets and a computational error of less than 0.013 mm.

In the automatic assembly of temperature-differential method for the hole-shaft interference fit structure, due to the huge temperature difference between the low temperature and the room temperature environment, the surface of the hole-shaft parts is prone to the formation of a frost layer, which seriously affects the visual measurement accuracy of the assembly pose. To address this problem, this paper draws on the research idea of image dehazing and proposes a Single Image Edge-Enhanced Defrosting network (SIEED) based on the coding-decoding structure to realize efficient defrosting from the image level. SIEED comprises the following key modules: the Edge-Enhanced Convolution Module (EECM), which leverages the sensitivity of convolution operators to edge features, enhancing edge information extraction; the Spatial-Guided Attention Module (SGAM), which employs local sensing techniques to address the non-uniform frost distribution through regional differentiation; the Weight-Based Iterative Fusion Module (WIFM), which dynamically fuses shallow and deep features to mitigate the loss of low-frequency features induced by deep convolution; and the Adversarial Discrimination Module (ADM), which incorporates global and local discriminators to balance the realism of localized defrosting with the overall coherence, using adversarial generation to produce defrosting images closer to reality. In addition, this paper proposes an automated acquisition method for the real dataset of hole-shaft images to guarantee the reliability and practicality of model training. The experimental results show that SIEED exhibits excellent performance in the frost-covered image defrosting task, and the pose measurements of its reconstructed images are highly close to those of the clean images, which fully verifies the validity, and reliability of the method in practical applications.

The new generation of aeroplanes presents a stepwise development of high stealth, long range, and long life, which puts higher requirements on the precision of aeroplane assembly. Therefore, a virtual preassembly technology based on the measured model is investigated to achieve accurate quality control and traceability adjustment in the critical parts of the assembly, which require ultra-high precision, by accurately representing the part manufacturing errors and the cumulative assembly errors throughout the assembly process. However, the reconstruction accuracy of the measured model is crucial for accurately expressing part manufacturing errors. In this paper, a data processing and model reconstruction method is introduced. Through the alignment of measurement data and CAD model, manufacturing and Measurement errors are introduced, unqualified points are filtered, and redundant and mixed points are filtered to obtain high-quality point cloud data; a NURBS surface fitting algorithm based on the asymptotic iteration of the fitting accuracy is proposed to obtain the least-squares base plane through a quadratic polynomial to carry out the initial surface reconstruction, and then add the reconstruction accuracy and smoothness conditions to carry out the secondary surface reconstruction to improve the reconstruction accuracy of the model. Finally, a set of aircraft HUD system simulations is designed to validate the method, which proves its feasibility.

In this paper, a neural-network-based method for decoupling the contact forces between assembly interfaces and motion actuators is explored, aiming at projecting the actual forces on assembly interfaces during the assembly of large aircraft structures. Due to the spatial limitation and geometrical complexity of the assembly interface, it is difficult to directly measure the forces on the assembly interface during pose adjustment and aircraft cabin assembly by traditional methods. Therefore, it is proposed to decouple the assembly surface forces through the relationship model between the data from three-dimensional force sensors and the forces at the assembly interface to ensure the effectiveness of force control and monitoring. Aiming at the errors existing in the force model of the traditional center-of-gravity method, this paper proposes a neural network to construct the force decoupling model, gate-separation multilayer perceptron, and uses the feature-separated gating training to design the input-output relationship to improve the force prediction accuracy. During the training process, the model stability and generalization ability are ensured by the mean square error loss function, network structure screening, and hyperparameter tuning. The model can detect the force condition of the assembly interface in real time, which provides the basis for the force control in the flexible assembly process. Experimental results show that the method can effectively optimize the force prediction accuracy between assembly interfaces during pose adjustment.

In automated aircraft assembly, achieving high-precision alignment is essential due to the presence of multiple coupled error sources that significantly affect final product quality. This study proposes an integrated framework to model multi-source errors via a directed coupling network and to quantify their impact using Monte Carlo simulations. To reduce the complexity of tolerance allocation, Sobol-based global sensitivity analysis is applied to identify dominant contributors to assembly deviations. The most influential parameters are retained for multi-objective optimization using the non-dominated sorting genetic algorithm II (NSGA-II). This framework enables the minimization of key assembly deviations while maintaining computational efficiency. Experimental validation on a typical helicopter ring assembly demonstrates that the proposed optimization approach increases the position pass rate from 67.4% to 100.0% and the coaxiality pass rate from 93.5% to 100.0%. The corresponding process capability indices (CPK) also improve significantly, from 0.31 to 2.19 for position and from 0.62 to 1.06 for coaxiality. These improvements not only satisfy high-precision assembly requirements but also exceed common industry benchmarks, demonstrating the method’s practical effectiveness under multi-source uncertainty.

In the active compliant assembly system of large components, the centroid parameter is an important factor affecting the dynamic model of the attitude adjustment mechanism. This paper proposes a novel automatic centroid measurement method. Firstly, the contact force calculation method between large components during active compliant assembly was described, and the impact of the centroid of large components on the accuracy of the contact force calculation was analyzed. Secondly, based on the principles of torque balance and rigid body rotation, a novel centroid calculation model considering lateral forces is established, and RANSAC is used to optimize the centroid solution. Then, numerical simulation methods were used to analyze the impact of random errors and weighing strategies on the centroid measurement accuracy, and suggestions were given to improve the centroid measurement accuracy. Finally, a compliant assembly system for large component was built in the laboratory, and centroid measurement and accuracy verification experiments were conducted. The experimental results show that the proposed method can significantly improve the accuracy of centroid measurement and can effectively reduce the contact force calculation error during the compliant assembly process of large components.

In existing automated docking assembly solutions for aerospace products, segmented docking surfaces suffer from machining errors and deformation, making assembly difficult. The aim of this study was to reduce the internal stresses in cabin assembly and maintain sufficient assembly accuracy. A cabin assembly optimization method was developed based on force–position coordination analysis. To address the interference problem of the cabin flange surface, a penalty-function-based reweighting iterative approximation method was constructed based on the square-hole characteristics of the docking. In the assembly verification, mechanism calibration and measurement field iterative compensation optimized the mechanism pose, and the accuracy of the pose adjustment component and the cabin assembly stress were tested. The experimental results show that the method proposed can reduce the interference problem during the assembly of a flange face and optimize the assembly stress while achieving assembly accuracy. The results for accuracy and interference force after flange face assembly were obtained; however, no feedback on the analytical method was obtained through the experimental results. Therefore, it is necessary to test further the propositions investigated. The preassembly analysis method for large cabin assemblies proposed optimizes the cabin assembly accuracy and internal stress under the deformation state of the assembly surface.

The accuracy and consistency of metrology data are the cornerstones of the safety and reliability of aircraft throughout aeronautical products’ lifecycles. Due to the heterogeneous nature of metrology data derived from various sources, knowledge silos commonly emerge, complicating the integration and reuse of knowledge. This study introduces an entity alignment model leveraging multi-perspective embedding. It employs a multi-scale graph convolutional network enhanced by a gating mechanism that aggregates multi-hop neighborhood features to capture the structural embeddings of nodes. Additionally, the model utilizes TransD for representing complex relationships and BERT for capturing entity attributes, facilitating more comprehensive entity representations. Entity alignment is then accomplished by integrating structural, relational, and attribute embeddings using a weighted strategy. In this study, we conducted experimental validation on aeronautical metrology data and also assessed our proposed model on five benchmark datasets. The results indicate that our model significantly outperforms comparative models, demonstrating its potential to enhance the management and application of aeronautical metrology data.

With the increasing requirements for aircraft assembly accuracy, pre-assembly analysis technology based on measured data has become one of the most important process precision compensation methods. How to accurately obtain and analyze the measured data of key features of parts according to assembly intention is the key step of pre-assembly analysis technology. Existing researches only explain the importance of the correct selection of measurement datum qualitatively, but none of them analyzed theoretically and provided specific methods when the measurement datum is inconsistent with the design datum, which is easy to misjudge the originally qualified parts. Therefore, this paper proposes a novel assembly-oriented measurement datum transformation and tolerance reallocation method for aircraft assembly. Firstly, based on the assembly type and geometric feature classification, the measurement datum of parts without Geometric Dimensioning and Tolerancing (GD&T) is identified, and the importance of correct selection of measurement datum is proved by a case theoretically. Secondly, when the measurement datum is not coincident with the assembly datum, a spatial homogeneous transformation method of measured data is proposed, which transforms the measured data from a temporary measurement coordinate system (TMCS) to an assembly datum coordinate system (ADCS). Finally, the tolerance redistribution of the measurement target features relative to the TMCS is carried out directly on the basis of the original tolerance. The experimental results show that the key features exceeding the upper tolerance limit of 0.042 mm are within the theoretical tolerance again after datum transformation, which solves the problem that the part has been misjudged as unqualified in the measurement process.