Scholarlink

<p>Cone forgings are widely used in aviation, aerospace, nuclear power, and other industries due to their high reliability and strong bearing capacity. In the process of thermal manufacturing, the conical cylinder forging is prone to causing the deformation of the upper and lower bell-mouths due to plastic instability. However, the existing methods are limited by the technical principle, and it is difficult to achieve accurate detection. Therefore, this paper proposes PHoRNet: a bell-mouth deformation detection method for cone forgings based on a 6D probabilistic voting Hough space point cloud registration method. The innovations of PHoRNet are as follows. (1) Multi-scale feature fusion: Fusion of a dynamic graph convolutional network, curvature uncertainty quantization, and a triangular weighted feature extractor to enhance feature robustness. (2) Hough space sampling optimization: Optimize the Hough space using farthest point sampling, and enhance the solution neighborhood coverage by screening quaternion candidate transformations with distance constraints. (3) Optimization of registration results: edge relaxation block and thinning network are introduced, and 6D sparse attention convolution is used to optimize registration and eliminate quantization error..&#xa0;PHoRNet can accurately detect the bell-mouth deformation of cone forgings of 1% and below. The outer diameter detection error is ± 0.15&#xa0;mm, and the roundness is 0.1&#xa0;mm. The error index is better than the high-precision forging detection requirements.</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

PHoRNet: Bell-mouth deformation detection method of cone forgings based on 6D probability voting in Hough space point cloud registration method

  • 期刊论文

  • 发表日期：

    2025年11月26日

查看全文
  • Xiang Chen,
  • Yun-Gang Zhang,
  • Yu-Cun Zhang

摘要

Cone forgings are widely used in aviation, aerospace, nuclear power, and other industries due to their high reliability and strong bearing capacity. In the process of thermal manufacturing, the conical cylinder forging is prone to causing the deformation of the upper and lower bell-mouths due to plastic instability. However, the existing methods are limited by the technical principle, and it is difficult to achieve accurate detection. Therefore, this paper proposes PHoRNet: a bell-mouth deformation detection method for cone forgings based on a 6D probabilistic voting Hough space point cloud registration method. The innovations of PHoRNet are as follows. (1) Multi-scale feature fusion: Fusion of a dynamic graph convolutional network, curvature uncertainty quantization, and a triangular weighted feature extractor to enhance feature robustness. (2) Hough space sampling optimization: Optimize the Hough space using farthest point sampling, and enhance the solution neighborhood coverage by screening quaternion candidate transformations with distance constraints. (3) Optimization of registration results: edge relaxation block and thinning network are introduced, and 6D sparse attention convolution is used to optimize registration and eliminate quantization error.. PHoRNet can accurately detect the bell-mouth deformation of cone forgings of 1% and below. The outer diameter detection error is ± 0.15 mm, and the roundness is 0.1 mm. The error index is better than the high-precision forging detection requirements.