Driven by the shift toward large-scale integrated die-casting in new energy vehicle (NEV) manufacturing, industrial robots are being explored for high-speed drilling to overcome the flexibility limitations of conventional machine tools. However, their low structural stiffness severely limits performance. At the moment of drill breakthrough, the step-unloading of the axial cutting force triggers a rapid release of the stored elastic energy, causing a millisecond-level dynamic response termed the "breakthrough surge", which significantly affects joint loads and hole accuracy. Conventional compensation methods, based on static deformation or steady-state chatter analysis, fail to address this transient behavior. To tackle this, a Stiffness-Inertia Dynamic Synergistic Optimization (SIDSO) strategy is proposed. A transient dynamic model of the surge is developed via a time-domain energy conversion framework, whereby an Energy-Mass Ratio (EMR) metric is introduced to quantify the combined effect of system compliance and operational-space inertia. To avoid local minima in the non-convex configuration space, a Hybrid Multi-Start Sequential Quadratic Programming (HMS-SQP) algorithm is used for global posture optimization under kinematic and singularity constraints. Comprehensive ablation experiments demonstrate the limitations of single-objective stiffness or mass compensation. Furthermore, machining tests show that SIDSO-optimized postures reduce the average surge velocity amplification ratio by 53.5% and the mean hole diameter error by 39.5% compared with default CAM-generated postures, significantly outperforming both stiffness-only and mass-only optimizations. This work establishes a novel, effective framework for robot posture planning, supporting high-quality, high-speed drilling of large integrated thin-walled components in NEV manufacturing.

Industrial robots hold considerable potential in the field of milling operations. However, due to their structural characteristics, significant machining errors often occur during the milling process. For the robotic milling, existing research rarely provides direct control strategies for machining error compensation, which limits the further application of industrial robots in high-precision machining tasks. To enhance the robotic milling accuracy, this paper proposes an error compensation control method based on force sensing. First, to predict the relationship between cutting forces and machining parameters, an improved cutting force model is developed by introducing the material removal parameter Sl. Combining the cutting force signals with the robot's position data, the machining error can be predicted. Furthermore, by considering the stiffness characteristics of both the robot and the workpiece, an error compensation control method is proposed. The initial milling trajectory is generated using the robot's spatial pose and the workpiece model. Based on force sensing and the desired machining accuracy, the cutting parameters are adaptively adjusted. A data-driven adaptive parameter adjustment strategy is further proposed by integrating robot motion data, machining data, and cutting force signals. By adjusting the feed rate in different out-of-tolerance regions, a new compensated milling trajectory is generated to correct machining errors. To validate the effectiveness of the proposed method, robotic milling experiments were conducted on thin-walled light alloy workpieces and feature components. The experimental results demonstrate that the proposed approach significantly reduces machining errors in robotic milling, thereby improving both machining quality and efficiency. These results indicate that the proposed method has strong potential for high-precision robotic milling of complex thin-walled structures.

Thin-walled ring-shaped structures are widely employed in aerospace applications and demand extremely high manufacturing precision. However, during machining, the combined effects of time-varying cutting forces and workpiece stiffness generate non-uniform deformation fields that are challenging to control. To address this, a hybrid mechanism-data digital twin, enhanced with spatial–temporal correlation, is developed for real-time reconstruction of machining deformation fields, achieving virtual-physical synchronization and to support in-process compensation and downstream positioning. Within this digital twin framework, the physical entity is the turning process of the thin-walled ring-shaped part; the virtual model comprises an elastic mechanistic model and a data-driven module. The elastic mechanistic model reconstructs the instantaneous deformation, and by incorporating a substructure method and temporal learning of cutting loads, the computational efficiency is significantly enhanced. Through spatial correlation learning of the deformation field, the global deformation field can be reconstructed using only a minimal number of sampling points. The data-driven module dynamically updates the mechanistic model using real-time cutting signals and historical deformation data, ensuring high-accuracy full-field reconstruction and virtual-physical synchronization. Compared with the traditional mechanistic model, the proposed method improves deformation field reconstruction accuracy by approximately 10.7

In automated machining line operations, on-machine measurement is widely adopted to enhance the form accuracy of workpieces. Due to limitations in response speed and processing efficiency, online measurement systems commonly rely on the least-squares method to process scanned point cloud data; however, the accuracy of such methods is often insufficient for precision measurement tasks. Moreover, complex workpiece surfaces often incorporate diverse geometric features such as holes, slots, and annular structures, which further increase the complexity of measurement and evaluation, thereby reducing accuracy. To address the practical demands for efficient, accurate, and robust identification of surface geometric errors in machined parts, this study develops a precision measurement method for evaluating straightness and flatness errors based on measurement point cloud data. A hybrid iterative method is proposed, combining convex optimization with a point-based control strategy. The method first performs coarse outlier elimination using a convex polyhedral envelope, followed by dimensionality reduction and iterative pruning based on convex hulls. Finally, a point-wise traversal approach is applied to the remaining data to determine the minimum enclosing zone. Adhering to the minimum zone criterion, the proposed method enables more accurate and efficient evaluation of straightness and flatness errors. Experimental validation shows that the proposed method exhibits strong robustness and response speed compatible with production cycle requirements, enabling accurate identification of geometric errors on complex machined surfaces.

The end-effector accuracy of a multi-axis milling system is primarily affected by assembly errors and deformation errors induced by low structural stiffness. This accuracy exhibits spatially nonlinear and non-uniform variations with changes in system pose, especially in robotic milling systems. Therefore, full-space accuracy modeling that accounts for manufacturing and assembly processes is crucial, particularly for machining workspace optimization. However, existing assembly deviation models are generally limited to error fluctuation simulations under fixed poses and lack the capability to analyze accuracy variations across the entire motion space of kinematic mechanisms, often requiring remodeling for different poses. To address this issue, this paper proposes a full-space geometric variation propagation modeling method for multi-axis robotic machining systems, considering local parallel chains. In the proposed model, the effects of manufacturing tolerances of multiple axes and their propagation on the geometric accuracy of a multi-axis milling system are considered in the spatial domain during the milling motion process. Firstly, three-dimensional tolerance expressions of joint and shaft-hole features are defined using small displacement Torsors, which can represent small feature variations within their tolerance ranges. Then, feature-to-feature Jacobian matrices are defined to characterize geometric variation propagation in multi-axis assemblies. Consequently, an overall Jacobian–Torsor-based expression model can be generated through the construction of a dimensional chain diagram. Based on the proposed model, case studies are conducted on a multi-axis robotic milling system to validate its effectiveness in modeling geometric variation propagation. The proposed method provides a comprehensive understanding of the mechanism of geometric variation propagation in robotic milling processes.

Complex cylindrical light-alloy structural castings are widely used in aerospace applications due to their simple forming process and intricate internal cavity geometry. Robotic milling systems are extensively employed to meet the demands of multi-variety, low-volume production of these components. However, under stable machining conditions, the weak stiffness of both the robotic system and the cylindrical workpiece often results in non-uniform force-induced deformation during milling. To address this issue, a spatial system stiffness model is developed based on multi-body small deformation theory and uniformed expression of force-induced deformation in milling process of cylindrical thin-walled parts. A unified expression of equivalent stiffness is established in the machined surface space of the cylindrical thin-walled workpiece, taking into account the weak stiffness of both the robotic milling system and the workpiece. Based on this, a predictive methodology is proposed to estimate the non-uniform distribution field of force-induced deformation. And the accuracy of the model is verified through simulation analysis and milling experiments. This research enables the effective prediction of non-uniform distribution fields and providing a theoretical guidance for deformation error control in the robotic milling of weak stiffness cylindrical thin-walled parts.

Dimensional deviations in battery module length can stem from the internal forces release after the assembly process. This manuscript presents the Dimension-Performance Integrated Assembly Analysis Method, which considers both post-assembly dimensional distributions and post-service residual life. The modified Hill strain energy function is employed to characterize the hyperelastic mechanical behavior of the aerogel-based thermal insulation pads, alongside establishing a process-based model for the battery module's post-welding rebound. Furthermore, an equivalent model is introduced to compute the equilibrium positions of the compression and tension curves to ensure both precision and efficiency. The dimensional measurements validate the accuracy of the process-based model and the equivalent model, and a thorough analysis of the module's rebound mechanism sheds light on the interplay between the equivalent model parameters and the structure mechanical properties or initial dimensions. Accounting for the battery module's post-welding rebound mechanism, the Dimension-Performance Integrated Assembly Analysis Method assesses the impact of key dimensions and process parameters on target dimensional distributions and structural failure probability. This method enables the design of battery thickness tolerance band and process parameter ranges to ensure zero failure probability within the target charge-discharge cycles.

Circumferential topography is a key indicator of the machining quality of ring-shaped workpieces, and it is generally affected by multiple geometric and location error sources arising from the machine tool-fixture-workpiece system during the machining process. To enable effective control of surface quality under these compounded errors, an integrated 3D surface topography prediction model is formulated. In the proposed model, the combined motion errors of the workpiece and tool, the surface topography blank errors of the workpiece, and the alignment deviations in ring positioning using a quick-change clamping system are simultaneously represented within a unified framework. The effects of geometric errors in the machine tool, fixture, tool and workpiece on the machined surface topography are simulated using the equivalent error transmission chain of a multibody machining system. The topography deviations of matching features in the machining system and the workpiece alignment deviations of the clamping system are incorporated via a location deviation simulation algorithm. The circumferential surface topography is then reconstructed over the entire tool trajectory of the turning process. The simulation and experimental results indicate that the proposed model can effectively predict the 3D surface topography in auto-located turning of cylindrical thin-walled parts, which are typically affected by multiple geometric and location deviations, and offer theoretical guidance for surface topography control.

Due to the measurement equipment and human errors, low-quality annotation in real-world industrial datasets is inevitable. Challenges such as sample imbalance, noisy labels, and fault mode scarcity caused by low-quality annotation make zero-shot compound fault diagnosis (ZCFD) a challenging yet valuable task in industrial applications. This article proposes a novel diagnostic framework to address ZCFD challenges under low-quality annotation. Specifically, a distribution-guided channel attention (DCA) module is introduced to prevent the feature weight matrix from falling into local optima. A dilated convolutional network (DCN) is employed as the backbone network for multiscale feature fusion and fault decoupling. A unidirectional labeling bias guided cross-entropy (CE) loss function is designed to mitigate the negative impact of noisy labels during gradient computation. The proposed unidirectional labeling bias guided CE and DCA DCN (USCE-DCADCN) method has been evaluated on both the intershaft bearing fault dataset (IBFD) and the robot compound fault dataset (RCFD) to demonstrate its superior performance on ZCFD tasks under low-quality annotation.

To address the low detection accuracy and high inspection-mold cost of the traditional sheet-metal part detection methods, a fast and accurate non-contact 3D measurement system was designed. For the registration difficulty due to unclear regional features of sheet-metal parts, an ICP algorithm with local boundary neighborhood feature constraints was proposed to fix data slippage and weak convergence during registration. First, the internal point cloud centroid was extracted via bounding-box discretization for downsampling to reduce data stratification. Next, the surface index of downsampled point cloud data was calculated, and the original point cloud boundary-adjacent point set was mapped and extracted. Then, two homologous and heterogeneous point clouds were combined as the input for the point cloud registration algorithm. Considering the distribution and local features of source and target point clouds, the objective function was modified to converge and enhance the algorithm’s focus on feature boundaries. Finally, point cloud data from public datasets and real-world collections were used for verification. The system’s final spatial-position detection accuracy was within ± 0.2 mm, with a mean value of less than 0.05 mm and a standard deviation of less than 0.15 mm, fully meeting the measurement requirements for the body-in-white sheet-metal parts.

Lightweight, high-performance cylindrical components with complex structures are widely used in gas turbines, wind turbines, and new energy vehicles. The complex structure makes single-stage machining challenging, leading to a multi-stage series process. Upstream errors propagate downstream, causing cumulative product errors. The multi-stage process involves multiple uncertainty errors induced by time-varying processes, making error control challenging. A single-stage processing circular topography prediction model was proposed under defined process parameters, mapping global geometric errors in the machine tool-fixture-tool-workpiece system to the processing circular topography. The manufacturing errors of the upstream process and the datum errors of the downstream process were unified using skin models. By integrating inter-stage genetic mechanisms, a multi-stage circular topography prediction model was established, which reveals the mapping relationship from multi-stage process parameters and processing system errors to the final circular topography. Extensive virtual machining using the model evaluated final topography under multi-system geometric errors and parameter uncertainties. Sensitivity analysis using the Sobol method identified key errors affecting processing quality. Based on the results of the sensitivity analysis, the relationships between each Quality Inspection Items for final Processing (QIIFP) and the errors of each process were identified. Additionally, the manufacturing tolerances of the relevant features in the multi-stage processing systems were redesigned to optimize the configuration of resources and process capabilities. Using the multi-stage circumferential topography prediction model, the CP/CPk design requirements were mapped to the process tolerance control limits. In contrast to existing single- stage optimization, the multi-stage circular topography prediction and sensitivity analysis model further improves cost-efficiency. Key factors in the multi-stage process chain were identified, enabling precise quality control within stages and coordinated process optimization, offering theoretical guidance for multi-stage manufacturing of cylindrical components.

Flexible robotic cells are pivotal in flexible and customized manufacturing. An effective scheduling policy for such cells can significantly reduce the makespan and improve the production efficiency. This study introduces an innovative end-to-end real-time scheduling method leveraging deep reinforcement learning (DRL) to minimize the makespan in a flexible robotic cell. We introduce a heterogeneous disjunctive graph model for a nuanced representation of the scheduling problem, which incorporates transportation through specific disjunctive arcs. The DRL utilizes Graph Neural Network (GNN) for model feature extraction and employs Proximal Policy Optimization (PPO) to train the scheduling agent. Our methodology can also better leverage the transport robot capacity to mitigate system blockage and deadlock. Numerical experiments are conducted to demonstrate the effectiveness of the proposed method.

Sheet metal parts account for more than 60

Accurate prediction of cutting forces is crucial for optimizing machining processes, reducing costs, and shortening lead times. Cutter-workpiece engagement (CWE), representing the instantaneous contact geometry between the cutting edges and the in-process workpiece material, is indispensable for defining actual machining conditions and serves as a prerequisite for precise cutting force prediction. However, current approaches often oversimplify cutting-edge distributions by assuming uniform shapes to streamline computational efforts. Such simplifications inadequately capture complex geometries, compromising prediction accuracy and limiting applicability to diverse tool designs. To overcome these limitations, this study proposes a novel cutting force prediction model that accommodates tools with arbitrary cutting-edge geometries. The cutting edges are discretized into points independently of the tool contour to enable precise geometric characterization. An enhanced point-based algorithm is developed to determine CWE by decomposing the machining process into explicit oblique cutting elements. Cutting forces for these elements are predicted using an artificial neural network (ANN) trained on a dataset with labels derived from finite element simulations. The proposed method is validated through two meticulously designed experiments and a practical application in aeroengine blade milling. By bridging the gap between complex cutting-edge distributions and reliable force prediction, this work provides a customized and standardized framework for the efficient simulation of universal five-axis machining and advances the development of robust virtual machining systems capable of optimizing industrial processes.

This manuscript proposes a deviation analysis method based on the Jacobian-Torsor model for the rigid-flexible hybrid deviations in serial assembly. The method establishes the Torsor model to represent the linear and angular deviations induced by the flexible deformation of the hyperelastic aerogel-based thermal insulation pads. Meanwhile, the Jacobian model is constructed to accurately predict the assembly deviations resulting from the combined effects of rigid and flexible deviations. The position deviations of the intermediate plate and the battery side faces are evaluated and utilized to address specific assembly issues in battery stack assembly. The simulation results of the assembly deviations based on this approach align with the production process measurement data, which demonstrates the effectiveness of the proposed method. In addition, the method can be further used to refine the tolerance representation models for various interactions between rigid and flexible components in parallel assembly scenarios. (c) 2025 The Authors. Published by ELSEVIER Ltd. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)

The multi-component stacked assembly of electric vehicle batteries has the characteristic of rigid-flexible hybrids between contact surfaces, such as aerogel thermal insulation pads, which challenges assembly quality control of large-scale and high-speed manufacturing. This article proposes an approach using the weighted objective function of assembly (WFA) to solve the hybrid assembly problem. In order to predict the interface contact state of the rigid-flexible hybrid assembly, the approach considers the distance constraint, the interference constraints, and the equilibrium equation to transform the rigid-flexible hybrid assembly problem into a weighted optimization problem. The target dimension distribution is obtained by leveraging an enhanced genetic algorithm, which combines the elite retention strategy and the targeted gene mutation method. Moreover, the WFA model can be applied not only to consider the dimensional tolerance and the flexible deformation during the assembly process, but also to carry out the coupling analysis under different loading conditions. The accuracy and efficiency of the proposed method are exhibited through an industrial case study of battery stacked assembly. While maintaining computational accuracy, a significant reduction in time costs is achieved, making it applicable for dimensional distribution predictions that rely on Monte Carlo simulations. The proposed WFA method can be applied to support the design and prediction of battery stacked assembly or other rigid-flexible coupled assembly.

In the machining of aero-engine gearbox ring-shaped parts, deterministic positioning is often achieved by general fixtures combined with manual alignment due to various sources of error. However, manual intervention leads to discontinuous and inefficient production. To improve this situation, workpiece alignment is centralized on alignment table equipment, and then the workpieces are transferred to the machine tools by robots for processing. In this case, the accuracy of clamping and positioning on the alignment table is extremely important. In this paper, a simulated model is constructed to predict workpiece clamping and positioning accuracy error distribution by combining manual alignment and the GD&T of process system. Firstly, an error transfer chain for the non-deterministic positioning alignment table is constructed based on assembly layers and features. Secondly, the alignment process is modeled and integrated into the established chain. Through the update of the error transfer chain, the deterministic clamping and positioning model are developed. The model shows excellent accuracy by comparing simulated results with the experimental measurements. Furthermore, an experimental scheme is designed to evaluate the influence of different types of alignment on the positioning accuracy of clamping in the assembly system. Through this experimental evaluation, the key alignment error factors affecting positioning accuracy are identified, and a quantitative metric is provided to assess the impact of workpiece and equipment alignment on the precision of workpiece fixturing and positioning. Based on these findings, the optimization suggestions for tolerance design of alignment are provided.

Assembly geometric error as a part of the machine tool system errors has a significant influence on the machining accuracy of the multi-axis machine tool. And it cannot be eliminated due to the error propagation of components in the assembly process, which is generally non-uniformly distributed in the whole working space. A comprehensive expression model for assembly geometric error is greatly helpful for machining quality control of machine tools to meet the demand for machining accuracy in practice. However, the expression ranges based on the standard quasi-static expression model for assembly geometric errors are far less than those needed in the whole working space of the multi-axis machine tool. To address this issue, a modeling methodology based on the Jacobian-Torsor model is proposed to describe the spatially distributed geometric errors. Firstly, an improved kinematic Jacobian-Torsor model is developed to describe the relative movements such as translation and rotation motion between assembly bodies, respectively. Furthermore, based on the proposed kinematic Jacobian-Torsor model, a spatial expression of geometric errors for the multi-axis machine tool is given. And simulation and experimental verification are taken with the investigation of the spatial distribution of geometric errors on five four-axis machine tools. The results validate the effectiveness of the proposed kinematic Jacobian-Torsor model in dealing with the spatial expression of assembly geometric errors.

The battery is an important part of the new energy electric vehicle, and the control of the flatness of its side plate/bottom plate is the key to quality improvement in mass production. However, there are few pieces of research on the flatness distribution form at present, and the distribution form is often assumed to be a normal distribution, which leads to a significant deviation between the tolerance design and quality control of the flatness and the reality. This paper establishes a statistical model of flatness distribution, its theoretical distribution form is deduced as a normal range distribution, and then the experimental data of the flatness distribution are collected to verify this conclusion. Determining the flatness distribution form has practical effects on improving manufacturing quality and reducing costs in battery manufacturing.

The changing of working conditions, such as effective material removal rate (MRR) and cutting force, is an inevitable phenomenon during practical milling process because of tool wear and variable cutting parameter resulting in the variations of machined surface quality. It is an important indicators of machined surface quality which can be used for process online monitoring. However, for a multi-toothed face milling process the cutting force will be changing along with gradually increasing tool wear and designed variable cutting parameter both in feed direction and circumferential direction, which increases the difficultly for decoupling and identifying the changing portions of working conditions from the nominal curves in time domain. To address this issue, this paper attempts to provide a model- based methodology to identify practical parameters of variable working condition for multi-toothed milling process in real time. Based on a simulation model of instantaneous milling forces functions to the milling parameters and instantaneous MRR, the nominal milling force curves in time domain can be generated with the calculating of designed nominal MRR. Then an inverse problem-solving method is proposed to minimize the corresponding deviation curves between simulated and measured milling force curves, thus to identify the deviation of practical MRR from the nominal values. Therefore, the proposed model-based identification method of variable working condition can be applied to identify the deviations of cutting forces and practical MRR from their nominal values which are directly related with cutting teeth runout or changing tool wear. It provides an effective way and a theoretical guidance to identify the practical deviations from a normally irregular process of variable cutting parameter, and can be used for process monitoring.