In this paper, we present multiple methods to detect fasteners (bolts, screws, and nuts) from tessellated mechanical assembly models. There is a need to detect these geometries in tessellated formats because of features that are lost during the conversions from other geometry representations to tessellation. Two geometry-based algorithms, projected thread detector (PTD) and helix detector (HD), and four machine learning classifiers, voted perceptron (VP), Naïve Bayes (NB), linear discriminant analysis, and Gaussian process (GP), are implemented to detect fasteners. These six methods are compared and contrasted to arrive at an understanding of how to best perform this detection in practice on large assemblies. Furthermore, the degree of certainty of the automatic detection is also developed and examined so that a user may be queried when the automatic detection leads to a low certainty in the classification. This certainty measure is developed with three probabilistic classifier approaches and one fuzzy logic-based method. Finally, once the fasteners are detected, the authors show how the thread angle, the number of threads, the length, and major and root diameters can be determined. All of the mentioned methods are implemented and compared in this paper. A proposed combination of methods leads to an accurate and robust approach of performing fastener detection.
Automatic Detection of Fasteners From Tessellated Mechanical Assembly Models
Contributed by the Computers and Information Division of ASME for publication in the JOURNAL OF COMPUTING AND INFORMATION SCIENCE IN ENGINEERING. Manuscript received November 1, 2016; final manuscript received October 18, 2017; published online November 28, 2017. Assoc. Editor: Bahram Ravani.
- Views Icon Views
- Share Icon Share
- Cite Icon Cite
- Search Site
Rafibakhsh, N., Huang, W., and Campbell, M. I. (November 28, 2017). "Automatic Detection of Fasteners From Tessellated Mechanical Assembly Models." ASME. J. Comput. Inf. Sci. Eng. March 2018; 18(1): 011005. https://doi.org/10.1115/1.4038292
Download citation file:
- Ris (Zotero)
- Reference Manager