Beijing Research Institute of Automation for Machinery Industry (China)

Abstract Firstly, based on the principle of a six-axis manipulator, a complete machine for the detection of Complex curved surface workpiece surface structure is designed. After modular analysis, the motion parameters and structural parameters of each mechanism are determined, and its three-dimensional drawing is completed in SolidWorks software; Secondly, the ZMC network motion control system is used to complete its trajectory planning, establish the main steps and application technology of surface material detection, and carry out interference detection; Finally, the mechanical properties of the 3D model are analyzed in SolidWorks-Simulation module. The simulation results show that the shape variable of the cantilever beam is not more than 0.1010 m, the amplitude is not more than 0.0611 m, and the maximum deformation of the arc pendulum table is 0.8152 m. The stability of the whole machine is remarkable; The stresses of the left and right cantilevers are 20.9 MPa and 26.5 MPa, respectively, and are evenly distributed, which improves the inherent defects of uneven stress and excessive local load of the traditional six-axis manipulator.

In the automated pig slaughtering industry, to address the urgent need for precise robotic segmentation of pig carcass hind legs, this paper proposes an improved keypoint detection and localization algorithm, YOLOv8-GCSA-WL. To overcome the insufficient accuracy of traditional vision algorithms in complex industrial environments—impaired by factors such as blood stains and water vapor occlusion—this study innovatively utilizes X-ray images as the data source. To enhance the model's global perception and multi-scale representation of key features against complex backgrounds, a Global Channel-Space Attention (GCSA) module is designed and introduced. By capturing global long-range dependencies and local fine-grained spatial information in parallel, the GCSA module significantly improves the model's performance and balance. Furthermore, considering the high-precision requirements of industrial segmentation tasks, the keypoint localization loss function was replaced with Wing Loss, which effectively improves coordinate prediction accuracy due to its high sensitivity to small-range errors. Extensive experiments conducted on a self-constructed X-ray dataset of pig carcass hind legs validate the superior performance of the YOLOv8-GCSA-WL model. Its keypoint detection metric, P<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mAP50-95</inf> reached 0.961, significantly outperforming the baseline model, while also maintaining a high performance of 0.923 in the object detection task, thereby achieving a better balance between accuracy and efficiency. This paper confirms the high precision and robustness of the proposed algorithm in real-world industrial scenarios, providing a valuable reference for the development of automated and intelligent meat segmentation production lines.

Transverse coupled-bunch instabilities, driven by narrow-band impedances, can degrade beam quality in multi-bunch, high-intensity circular accelerators. To mitigate this effect, strict constraints on transverse narrow-band impedance are typically imposed during the accelerator design phase. However, slight deviations in component structures during the construction of accelerators and modifications during subsequent operations may result in impedance deviations from the design value. Therefore, determining the impedance parameters of circular accelerators through experimental beam measurements during operation is of greater practical significance. In this study, by analyzing the mode distribution of coupled-bunch instabilities and growth rates, we proposed two methods to obtain the transverse narrow-band impedance model, which is represented by an RLC resonator. The analysis is limited to the case of short bunches in a storage ring with narrow-band impedance, which is characteristic of the BEPCII accelerator. The corresponding approximation of considering only the lowest mode l = 0 for the growth rate of coupled-bunch mode was applied. Finally, we provided numerical calculation examples with predefined RLC resonators to demonstrate the implementation process of the two methods.

To address the challenges of short allowable motion windows and complex motion planning inherent in dual forging manipulator systems, this study proposes a coordinated motion pattern tailored to dual-manipulator operations, focusing on forging deformation behavior and press control characteristics. First, six representative long-shaft forging materials were classified based on typical industrial applications. Using DEFORM-3D (V11.0) software, the deformation process during the elongation operation was analyzed, and the velocity and displacement characteristics at both ends of the forgings were extracted to clarify the compliant motion requirements of the grippers. Next, a segmented computation method for manipulator allowable motion time was developed based on the motion–time curve of the hydraulic press, significantly improving the time utilization efficiency for coordinated control. Furthermore, experimental tests were carried out to verify the dynamic response performance and motion accuracy of the dual-manipulator system. Finally, the dual-manipulator forging cycle was systematically divided into four stages—pre-forging adjustment, inter-pass compliance, execution phase, and forging completion—resulting in a structured and implementable coordination control framework. This research provides both a theoretical foundation and practical pathway for achieving efficient and precise coordinated motion control in dual forging manipulator systems, offering strong potential for engineering application and industrial deployment.

This paper addresses the challenges of low efficiency and poor reliability in manual inspection of weld defects during ship manufacturing, proposing an improved YOLOv8-based detection method for hull closure weld appearance defects. By constructing a weld dataset containing five typical defects including cracks and pores, we innovatively integrate a Spatial and Channel Reconstruction Convolution (ScConv) module into the YOLOv8 backbone network. The spatial reconstruction unit and channel reconstruction unit effectively reduce feature redundancy while enhancing feature representation capability for micro-defects. Experimental results demonstrate that the improved model achieves 87% mAP50 on our proprietary dataset, with accuracy and recall rates reaching 87.8% and 86.4% respectively. Field deployment tests confirm a detection speed of 10.4s/m, showing 98.6% improvement in efficiency compared to traditional manual inspection (20.7s/m), thereby meeting real-time requirements. This method provides an automated solution for ship weld quality inspection with significant engineering application value.