Scholarlink

<p>This paper presents a solidification deformation prediction method for additive manufacturing of thin-walled components based on multi-layer digital twins (MDT). The interaction mechanism between additive manufacturing parameters and structure deformation is firstly forward constructed via finite element method (FEM). Innovatively, to avoid repetitive simulation of diverse geometries, the manufacturing process is virtually twinned in digital space in multiple layers, including the geometry and property twinning layers, videlicet the multi-layer digital twins. For the geometric twinning layer, the virtual entity of the designed manifold is morphologically constructed via tunnel-free voxelization, where the accuracy can be customized by adjusting voxelizing resolution. For the property twinning layer, the neural operator networks are conducted to construct the nonlinear relations between manufacturing parameters and properties based on FEM outcomes. Instead of fitting the deformation in the global structure domain, which may tarnish the scalability of the neural operator about diverse geometries, the solidification deformation of a specific voxel is defined as the function of local states defined by adjacent voxels, which are generated in the geometric twinning layer. The effectiveness of MDT in predicting solidification deformation is verified by comparing with FEM outcomes of thin-walled structures, where the neural operator is trained with evolution data of an 8-blades centrifugal turbine and applied to predict the deformation of a 14-blades one. The physical experiment is conducted via selective laser manufacturing of the mechanical unshrouded turbine. The metallographic diagrams of manufactured components are generated to observe the fabricating quality. The average prediction error of layer normalized deformation is 0.0176 and the precision attenuation ratio is 1.45%. The proposed MDT is especially useful for anti-deformation designing and mechanical performance regulating of thin-walled structures.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Solidification deformation prediction for additive manufacturing of thin-walled components based on multi-layer digital twins

  • 期刊论文

  • 发表日期：

    2025年9月12日

查看全文
  • Linxuan Wang,
  • Jinghua Xu,
  • Shuyou Zhang,
  • Jianrong Tan,
  • Shaomei Fei

摘要

This paper presents a solidification deformation prediction method for additive manufacturing of thin-walled components based on multi-layer digital twins (MDT). The interaction mechanism between additive manufacturing parameters and structure deformation is firstly forward constructed via finite element method (FEM). Innovatively, to avoid repetitive simulation of diverse geometries, the manufacturing process is virtually twinned in digital space in multiple layers, including the geometry and property twinning layers, videlicet the multi-layer digital twins. For the geometric twinning layer, the virtual entity of the designed manifold is morphologically constructed via tunnel-free voxelization, where the accuracy can be customized by adjusting voxelizing resolution. For the property twinning layer, the neural operator networks are conducted to construct the nonlinear relations between manufacturing parameters and properties based on FEM outcomes. Instead of fitting the deformation in the global structure domain, which may tarnish the scalability of the neural operator about diverse geometries, the solidification deformation of a specific voxel is defined as the function of local states defined by adjacent voxels, which are generated in the geometric twinning layer. The effectiveness of MDT in predicting solidification deformation is verified by comparing with FEM outcomes of thin-walled structures, where the neural operator is trained with evolution data of an 8-blades centrifugal turbine and applied to predict the deformation of a 14-blades one. The physical experiment is conducted via selective laser manufacturing of the mechanical unshrouded turbine. The metallographic diagrams of manufactured components are generated to observe the fabricating quality. The average prediction error of layer normalized deformation is 0.0176 and the precision attenuation ratio is 1.45%. The proposed MDT is especially useful for anti-deformation designing and mechanical performance regulating of thin-walled structures.