Development of an Automated CAD Database and Application on Aluminum Wheel

As data science technologies are being widely applied on various industries, the importance of data itself increased. A typical manufacturer company has a vast data set of products as 2D&3D drawing formats, but a common problem was that building a database from the 2D&3D drawings costs much, and it is hard to update the database after it once built. Also, it is high-cost job when the new factor researched and necessary to investigate the new factors on previously fixed or uploaded drawings. As new products are developed with time, these problems are getting more difficult. In this paper, an automated database building method using CATIA introduced and future probabilities are suggested. An aluminum wheel part was used as an example. An automated logic used CATIA V5’s VBA functions and was handled by python programming language. Product database was established by using the automated logic for extracting engineering design features, and data mining process was deployed based on the extracted data. The database established from part drawings consists of tabular features and 2D images of projection views. CNN and DNN fusion models were used for wheel weight regression, and Auto-Encoder was used for searching similarity in z-space. Using z-space distance among wheel images, it was able to find similar wheel designs. By using the methods of this paper, establishing and analyzing database were efficiently performed with low cost. This paper shows that constructing database on currently existing 2D&3D drawings can be done with low cost and suggests that inserting the method into a product life cycle as a middle process can build a fully automated data science cycle as future work.