Elsevier

Computer-Aided Design

Volume 27, Issue 1, January 1995, Pages 59-64
Computer-Aided Design

Research
Neural network approach to the reconstruction of freeform surfaces for reverse engineering

https://doi.org/10.1016/0010-4485(95)90753-3Get rights and content

Abstract

Reconstruction and manufacturing of existing freeform surfaces are of paramount importance for reverse engineering. These are particularly useful for the situations where the surfaces are partially damaged or the surface models are not available. In order to generate models based on the existing freeform surfaces, the surfaces are first digitized by either a laser scanner or a coordinate measuring machine. The digitized points are then used to construct the models. The paper presents a neural network approach to reconstruction of computer models for existing freeform surfaces, and manufacturing of surfaces. To evaluate the effectiveness of the approach, a mathematically known surface, a nonuniform B-spline surface, was used for generating a number of samples for training the networks. Three 4-layered neural networks were designed and trained using a modified back propagation algorithm. The trained networks then generated a number of new points which were compared with the calculated points using the known surface equations. These points were also used to generate toolpaths for machining the surface. The results show that the approach can be used to reconstruct computer representations of the existing surfaces, and to manufacture these surfaces. An example is included to illustrate the approach.

References (14)

There are more references available in the full text version of this article.

Cited by (125)

  • PDE patch-based surface reconstruction from point clouds

    2022, Journal of Computational Science
    Citation Excerpt :

    Another approach is based on the use of partial differential equations (PDEs) for surface reconstruction; see [9] for an overview on the field. In recent years, different machine learning methods, mostly based on neural networks and deep learning [10,11], metaheuristic techniques such as genetic algorithms and nature-inspired optimization algorithms [12,13], or combinations of both [14] have also been used for 3D reconstruction from point clouds. In this section, we will briefly review some of the recent works, which are closely related to this paper.

  • Novel re-parameterization for shape optimization and comparison with knot-based gradient fitting method

    2018, Computer Methods in Applied Mechanics and Engineering
    Citation Excerpt :

    The same analogy could be used in shape fitting but it entails excessive algorithm complexity which makes it less suitable in shape optimization. Neural networks are employed as the base framework in reconstruction of free-form surfaces for reverse engineering by Gu, P. and X. Yan [30]. Barhak, J. and Fischer, A [31] also used neural network and PDE techniques in parameterization and reconstruction from 3D scattered points.

  • Particle-based meta-model for continuous breakpoint optimization in smooth local-support curve fitting

    2016, Applied Mathematics and Computation
    Citation Excerpt :

    Unfortunately, the optimization problems given by all these approaches are very difficult and cannot be generally solved by traditional mathematical optimization techniques [22]. Interesting research carried out during the last two decades has shown that the application of Artificial Intelligence techniques can achieve remarkable results for optimization problems [4,7,8,40–43]. Most of these methods rely on some kind of neural networks, such as multi-layered neural networks with modified back-propagation algorithms [40], or Kohonen’s SOM (Self-Organizing Maps) nets [41].

  • Industrial Artificial Intelligence Approach for Shape Reconstruction in Quality Assessment of Digital Data from Manufactured Workpieces

    2023, Proceedings - 2023 4th International Conference on Industrial Engineering and Artificial Intelligence, IEAI 2023
View all citing articles on Scopus
View full text