Elsevier

Computer-Aided Design

Volume 40, Issue 3, March 2008, Pages 368-380
Computer-Aided Design

Similarity measures for mid-surface quality evaluation

https://doi.org/10.1016/j.cad.2007.11.008Get rights and content

Abstract

Mid-surface models are widely used in engineering analysis to simplify the analysis of thin-walled parts, but it can be difficult to ensure that the mid-surface model is representative of the solid part from which it was generated. This paper proposes two similarity measures that can be used to evaluate the quality of a mid-surface model by comparing it to a solid model of the same part.

Two similarity measures are proposed; firstly a geometric similarity evaluation technique based on the Hausdorff distance and secondly a topological similarity evaluation method which uses geometry graph attributes as the basis for comparison. Both measures are able to provide local and global similarity evaluation for the models.

The proposed methods have been implemented in a software demonstrator and tested on a selection of representative models. They have been found to be effective for identifying geometric and topological errors in mid-surface models and are applicable to a wide range of practical thin-walled designs.

Introduction

Mid-surface models have received a great deal of interest in recent years because they can reduce the computational cost of performing engineering analyses on thin-walled parts. The mid-surface representation of a part is a dimensionally reduced abstraction in which each wall is represented by a surface of zero thickness; for thin-walled objects the mid-surfaces provide a simplified description which retains the main shape characteristics of the original design. Mid-surface models are used extensively for the analysis of thin-walled parts in engineering analysis applications such as finite element analysis and mould filling/cooling analysis [1], [2], [3], [4]. Mid-surface models have also been used as a basis for feature recognition from thin-walled moulded parts [5].

There are a number of automated mid-surface generation techniques that can create a mid-surface abstraction from a CAD solid model, but there are currently no applications that can guarantee to generate a representative mid-surface model for any arbitrary solid part. This is partly due to limitations in the current algorithms, but also because there are many part shapes for which it is not possible to generate a representative mid-surface model [6]. In practice it is often necessary for the analyst to make manual adjustments to an automatically generated mid-surface model before it can be used for engineering analysis. Fig. 1. shows an example of good and poor quality mid-surface models generated using a commercial mid-surface generation tool (UGS I-DEAS NX). In Fig. 1(b) it can be seen that all of the features of the solid part are represented in the mid-surface model, whereas in Fig. 1(c) the boss has not been captured and the main walls have not been correctly connected together.

The motivation for the research presented in this paper is to assist users of mid-surface models who need to judge how well a mid-surface model represents the solid shape from which it has been generated. It should be emphasised that mid-surface representations can only accurately represent parts where the wall thickness is small compared to the other dimensions, and the methods presented in this paper are oriented towards practical applications of mid-surface models. For example in the finite element analysis two-dimensional meshes constructed on the mid-surface geometry are appropriate when “the length in one of the spatial dimensions, for example the material thickness, is much less than the lengths in the other two dimensions” [7]. Similarly parts designed for injection moulding must have a thin and relatively uniform wall thickness. Bralla [8] states that “uniform wall sections will help to produce warp-free and strain-free molded parts…Dimensional variations…are accentuated by uneven or abrupt wall-thickness changes”.

The objective of the research is to develop techniques that can be used to evaluate the accuracy of a mid-surface model by measuring its similarity to the solid model from which it was generated. Two techniques are presented, one to compare the geometric similarity of the two models and the second to compare their topological similarity. The remaining sections of the paper are structured as follows: Section 2 provides a review of the relevant literature relating to the mid-surface model generation and shape similarity measures and Section 3 presents the proposed similarity measures. In Section 4 the demonstrator is described, and in Section 5 case studies are presented. Section 6 discusses the results and conclusions drawn.

Section snippets

Literature review

The automatic generation of mid-surface abstractions from solid models has been a subject of research for many years and there has been significant progress in the field, however there is still no method that can generate an accurate mid-surface model for any arbitrary shape. The following sections provide a brief overview of existing mid-surface generation techniques, and a review of shape similarity measures that have been used to compare computer based geometry models.

Mid-surface model similarity evaluation

An effective mid-surface model similarity evaluation technique must be able to measure the similarity of a mid-surface model to a solid model with respect to both geometry and topology. The geometric evaluation should be sensitive enough to determine whether all the design features of the solid model are represented on the mid-surface model and the topological evaluation must evaluate whether the connectivity between solid model faces is correctly represented on the mid-surface model. The

Geometric similarity

A demonstrator for the geometric similarity evaluation has been implemented using C++ and the CAD system UGS I-DEAS NX. A flow chart of the evaluation process is shown in Fig. 6.

For the demonstrator the grids of points on the mid-surface and solid models have been generated using the automatic finite element mesh generation function in UGS I-DEAS, however a wide range of other techniques could be used. The Hausdorff comparison has been performed using a simple algorithm similar to that

Test cases

The mid-surface quality evaluation techniques have been tested on a range of thin-walled parts. The analysis for one part is presented in detail in Section 5.1, and the results for six other parts are included in Section 5.1.1.

Summary and conclusions

This paper has described two similarity measures for comparing mid-surface and solid models. The literature review highlighted that geometric similarity is not sufficient for evaluating mid-surface model quality, and that in order to obtain useful results it is necessary to use a topological measure of similarity in combination with the geometric similarity evaluation.

The geometric similarity measure proposed in this research is based on the Hausdorff distance and requires the models to be

References (30)

  • B. Li et al.

    Detail feature recognition and decomposition in solid model

    Computer-Aided Design

    (2002)
  • S. Joshi et al.

    Graph-based heuristics for recognition of machined features from a 3D solid model

    Computer Aided Design

    (1988)
  • D.J. Sheehy et al.

    Shape description by medial surface construction

    IEEE Transactions on Visualisation and Computer Graphics

    (1996)
  • C.N. Chu et al.

    Skeleton based design analysis of near net-shape products

    Annals of the CIRP

    (1993)
  • Donaghy RJ, McCune W, Bridgett SJ, Armstrong CG, et al. Dimensional reduction of analysis models. In: Proceedings 5th...
  • Cited by (0)

    View full text