Affiliation:
1. George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 30332
Abstract
Abstract
The ability to classify the capabilities of different manufacturing processes based on computer-aided design (CAD) models of parts is a key missing link in cybermanufacturing. In this paper, we present a one-step approach for automatically classifying the capabilities of three discrete manufacturing processes—milling, turning, and casting—based on part shape, quality, and material property attributes. Specifically, our approach utilizes machine learning to classify manufacturing process capabilities of these processes in terms of part shape attributes such as curvature, rotational symmetry, and pairwise surface point distance (D2) histogram computed from CAD models, as well as part quality (surface finish and size tolerance) and material property attributes of parts. In this manner, historical data can be utilized to classify the capabilities of manufacturing processes. We show that it is possible to achieve high classification accuracies—88% and 83% for the training and test data sets, respectively—using this approach. In addition, a key insight gained from this work is that part shape attributes alone are inadequate for discriminating between the capabilities of the manufacturing processes considered. Specifically, the inclusion of material property and part quality attributes enables the classifier to predict viable manufacturing processes that would otherwise be ignored using shape attributes alone. Future extensions of this work will include enriching the classification process with additional attributes such as production cost, as well as alternative classification methods.
Subject
Industrial and Manufacturing Engineering,Computer Graphics and Computer-Aided Design,Computer Science Applications,Software
Cited by
14 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献