Presentation is loading. Please wait.

Presentation is loading. Please wait.

An Evolutionary Strategy for Free Form Feature Identification in 3D CAD Models WSCG’07 conference Thomas R. Langerak, Joris. S.M. Vergeest, Huaxin Wang,

Similar presentations


Presentation on theme: "An Evolutionary Strategy for Free Form Feature Identification in 3D CAD Models WSCG’07 conference Thomas R. Langerak, Joris. S.M. Vergeest, Huaxin Wang,"— Presentation transcript:

1 An Evolutionary Strategy for Free Form Feature Identification in 3D CAD Models WSCG’07 conference Thomas R. Langerak, Joris. S.M. Vergeest, Huaxin Wang, Yu Song January, 31, 2007

2 2 Contents Introduction Method outline Underlying theoretical model Evolutionary feature recognition An application example Results & Analysis Conclusions and future work

3 January, 31, 2007 3 Introduction Modern CAD modellers make use of features

4 January, 31, 2007 4 Introduction Modern CAD modellers make use of features

5 January, 31, 2007 5 Introduction Modern CAD modellers make use of features Han (2000)

6 January, 31, 2007 6 Introduction Modern CAD modellers make use of features Han (2000) Laakko & Mäntylä (2000)

7 January, 31, 2007 7 Introduction Modern CAD modellers make use of features Han (2000) Laakko & Mäntylä (2000) Cavendish (1995)

8 January, 31, 2007 8 Introduction Modern CAD modellers make use of features Han (2000) Laakko & Mäntylä (2000) Cavendish (1995) Pernot (2003)

9 January, 31, 2007 9 Introduction Modern CAD modellers make use of features Han (2000) Laakko & Mäntylä (2000) Cavendish (1995) Pernot (2003) Vergeest (2001)

10 January, 31, 2007 10 Introduction Modern CAD modellers make use of features Feature recognition is needed when: High-order information is not available Translation of the high-order information to another domain is required

11 January, 31, 2007 11 Method outline Underlying theoretical model Features have a nurbs representation Parameter control occurs through displacement of the control points

12 January, 31, 2007 12 Method outline Underlying theoretical model Features have a nurbs representation Parameter control occurs through displacement of the control points

13 January, 31, 2007 13 Method outline Underlying theoretical model Features have a nurbs representation Parameter control occurs through displacement of the control points

14 January, 31, 2007 14 Method outline Underlying theoretical model Features have a nurbs representation Parameter control occurs through displacement of the control points

15 January, 31, 2007 15 Method outline Evolutionary feature recognition Feature in a CAD model are recognized using template features A succesful recognition occurs when the ‘distance’ between a template feature and the target shape is minimal

16 January, 31, 2007 16 Method outline Evolutionary feature recognition Evolutionary computation: treating computational problems as cases of natural evolution “Survival of the fittest” Features can be viewed as organisms, with parameters as ‘genes’

17 January, 31, 2007 17 Method outline Evolutionary feature recognition Consecutive generations of a feature population are computed, using the shape similarity as a fitness indicator. The populations ‘evolve’ to an optimal solution

18 January, 31, 2007 18 An application example Bottle Gen #0

19 January, 31, 2007 19 An application example Bottle Gen #1

20 January, 31, 2007 20 An application example Bottle Gen #2

21 January, 31, 2007 21 An application example Bottle Gen #3

22 January, 31, 2007 22 An application example Bottle Gen #4

23 January, 31, 2007 23 An application example Bottle Gen #5

24 January, 31, 2007 24 Results and analysis 2500 tests conducted Target shapes were simulated by instantiating features on a flat surface and distorting them with Gaussian noise. Different values for the feature population size were used

25 January, 31, 2007 25 Results and analysis Feature type (# of params) Correctly identified Incorrectly identified Not identifiedTotal Plane (6)308 (96%)13 (4%)1 (0%)322 Bump (8)291 (94%)14 (5%)3 (1%)308 Ridge (9)274 (88%)32 (10%)5 (2%)311 Cross (9)324 (96%)8 (2%)4 (1%)336 Step (8)296 (95%)14 (5%)1(0%)311 Wave (10)212 (69%)73 (24%)23 (7%)308 Blend (8)301 (96%)9 (3%)2 (1%)312 Crown (10)284 (97%)4 (1%) 292 Total2290 (92%)167 (7%)43 (2%)2500 Table 1: Result for different feature types

26 January, 31, 2007 26 Results and analysis Population size/ Selection size 10001500200025003000Total 5%789196100 463 10%8396100 479 20%82949599100470 30%7688 97100449 40%62838993100429 Total381 (76.2%)452 (90.4%)468 (93.6%)489 (98.8%)500 (100%)2290 Table 2: Result for different population sizes

27 January, 31, 2007 27 Conclusions and future work The method is able to identify features in a reasonable time (approx. 2 minutes) The method needs to be validated for more complicated situations. Work is being done to use the method for feature recognition and automatic feature type construction.

28 January, 31, 2007 28 Thanks and acknowledgements The research project DIT.6240 is supported by the Technology Foundation STW, applied science division of NWO and the technology program of the Ministry of Economic Affairs, The Netherlands. Thank you for listening!


Download ppt "An Evolutionary Strategy for Free Form Feature Identification in 3D CAD Models WSCG’07 conference Thomas R. Langerak, Joris. S.M. Vergeest, Huaxin Wang,"

Similar presentations


Ads by Google