Automatically Generating 60,000 CAD Variants for Big Data Applications

Abstract

Machine learning is opening up new ways of optimizing designs but it requires large data sets for training and verification. While such data sets already exist for financial, sales and business applications, this is not the case for engineering product design data. This paper discusses our efforts in curating a large Computer Aided Design (CAD) data set with desired variety and validity for automotive body structural compositions. Manual creation of 60,000 CAD variants is obviously not viable so we examine several approaches that can be automated with commercial CAD systems such as Parametric Design, Feature Based Design, Design Tables/Catalogs of Variants and Macros. We discuss pros and cons of each method and how we devised a combination of these approaches. This hybrid approach was used in association with DOE tables. Since the geometric configurations and characteristics need to be correlated to performance (structural integrity), the paper also demonstrates automated workflows to perform FEA on CAD models generated. Key simulation results can then be associated with CAD geometry and, for example, processes using machine learning algorithms for both supervised and unsupervised learning. The information obtained from the application of such methods to historical CAD models may help to understand the reasoning behind experiential design decisions. With the increase in computing power and network speed, such datasets together with novel machine learning methods, could assist in generating better designs, which could potentially be obtained by a combination of existing ones, or might provide insights into completely new design concepts meeting or exceeding the performance requirements.

CarHoods10k: An Industry-Grade Data Set for Representation Learning and Design Optimization in Engineering Applications

“…We implemented an automated data generation pipeline for automotive hoods in CATIA v5, which results in a large set of variants of hoods that are geometrically valid, manufacturable, exhibit sufficient variability, and have functional properties comparable to real-world designs. In [1], [50]- [52], we provide a detailed description of the CAD data generation, as well as validation and post-processing of generated 3D models and corresponding performance metrics. In the following, we provide a condensed summary of the primary tasks.…”

“…Parameter values describing the placement and characteristics of feature patterns were generated using a sampling scheme and stored in a design table [1]. Examples of design variations for the same base surface and features with different parameter values are shown in Fig.…”

“…We trained the autoencoder with four 1D convolutional encoder layers of sizes [64, 128, 128, 256] followed by one 1 Code available: https://github.com/HRI-EU/GDL4DesignApps. max-pooling layer that generates the latent representation of size L. From the latent representation, three fully-connected decoder layers of sizes [256, 256, N ] reconstruct the input, where N denotes the size of the point cloud.…”

“…In particular, machine learning and other data-driven approaches, nowadays, heavily contribute to the effectiveness of industrial evolutionary design optimization. These approaches provide, for example, improved problem parametrizations or design response prediction models, while relying heavily on the availability of realistic and large-scale benchmark data for methods development and performance assessment [1]. These benchmark data should be designed to closely resemble the data typically encountered in the targeted engineering application and concern geometric properties, such as suitable 2D or 3D shape representations and core features, as well as performance data, such as responses from aerodynamic or mechanical simulations of considered shapes.…”

Rapid feasibility assessment of components to be formed through hot stamping: A deep learning approach

“…To build the simulation models, the tool and blank CAD variant generation was automated using the VBA programming language in SolidWorks. Filtering rules as well as upper and lower bounds were applied to preserve geometric integrity during the automation process, as described by Ramnath et al [36,37]. To mesh the CAD geometries for the FE simulations, the commercial FE pre-processing software HyperMesh was used.…”