Digital Twin (DT) technology, which integrates multi-source information, is extensively applied for comprehensive monitoring, predicting, and optimizing manufacturing processes. The core of this technology is the Digital Twin Virtual Model (DTVM), which acts as a virtual mirror reflecting the real-world physical processes within a digital environment. In processes like tube bending, constructing a real-time DTVM capable of capturing full-field plastic deformation is essential for monitoring and analyzing plastic behavior. However, existing DTVMs often simplify spatial resolution and suffer from temporal delays, impeding the accurate real-time depiction of the complete state of the real physical processes. To address this issue, a real-time DTVM construction method based on spatio-temporal cascade reconstruction was proposed for full-field plastic deformation monitoring in metal tube bending. Initially, a joint-section driven predefined bending tube coordinate representation method was introduced to comprehensively capture the entire plastic deformation area in bending tubes. Subsequently, through a physics-derived model integrating limited real-time data and plastic forming theory, a low-fidelity model with complete but low accuracy was obtained. This model was subsequently refined into a high-fidelity model with both completeness and high accuracy using the proposed FPDR-Net. To eliminate temporal lags, the concept of compensation for time-delay through prediction was introduced. The newly developed TSCR-Net was applied to leverage past data to predict the present state, thereby achieving real-time synchronization mapping between the physical process and the DTVM. Finally, the proposed real-time reconstruction method for monitoring was validated through a case study on the bending of a 6061-T6 tube. The accuracy of full-field plastic deformation reconstruction was compared to traditional algorithms and finite element methods. The experimental results demonstrated that the proposed approach is highly efficient for real-time and full-field plastic deformation monitoring.

中文翻译：

---

通过时空级联重建的数字孪生虚拟模型实时构建，用于金属管弯曲制造中的全场塑性变形监测


数字孪生（DT）技术集成了多源信息，广泛应用于制造过程的全面监控、预测和优化。该技术的核心是数字孪生虚拟模型（DTVM），它充当反映数字环境中现实世界物理过程的虚拟镜子。在弯管等过程中，构建能够捕获全场塑性变形的实时 DTVM 对于监测和分析塑性行为至关重要。然而，现有的 DTVM 通常会简化空间分辨率并受到时间延迟的影响，从而阻碍了对真实物理过程完整状态的准确实时描述。针对这一问题，提出了一种基于时空级联重建的实时DTVM构建方法，用于金属管弯曲过程中的全场塑性变形监测。最初，引入了关节截面驱动的预定义弯管坐标表示方法来全面捕获弯管中的整个塑性变形区域。随后，通过整合有限实时数据和塑性成形理论的物理推导模型，获得了完整但精度较低的低保真模型。随后使用所提出的 FPDR-Net将该模型细化为具有完整性和高精度的高保真模型。为了消除时间滞后，引入了通过预测补偿时间延迟的概念。新开发的TSCR-Net用于利用过去的数据来预测当前状态，从而实现物理过程和DTVM之间的实时同步映射。 最后，通过 6061-T6 管弯曲的案例研究验证了所提出的实时重建监测方法。将全场塑性变形重建的精度与传统算法和有限元方法进行了比较。实验结果表明，该方法对于实时、全场塑性变形监测是高效的。