Deep Generative Models in Engineering Design: A Review

Abstract

Automated design synthesis has the potential to revolutionize the modern human design process and improve access to highly optimized and customized products across countless industries. Successfully adapting generative Machine Learning to design engineering may be the key to such automated design synthesis and is a research subject of great importance. We present a review and analysis of Deep Generative Learning models in engineering design. Deep Generative Models (DGMs) typically leverage deep networks to learn from an input dataset and learn to synthesize new designs. Recently, DGMs such as Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), feedforward Neural Networks (NNs) and certain Deep Reinforcement Learning (DRL) frameworks have shown promising results in design applications like structural optimization, materials design, and shape synthesis. The prevalence of DGMs in Engineering Design has skyrocketed since 2016. Anticipating continued growth, we conduct a review of recent advances with the hope of benefitting researchers interested in DGMs for design. We structure our review as an exposition of the algorithms, datasets, representation methods, and applications commonly used in the current literature. In particular, we discuss key works that have introduced new techniques and methods in DGMs, successfully applied DGMs to a design-related domain, or directly supported development of DGMs through datasets or auxiliary methods. We further identify key challenges and limitations currently seen in DGMs across design fields, such as design creativity, handling complex constraints and objectives, and modeling both form and functional performance simultaneously. In our discussion we identify possible solution pathways as key areas on which to target future work.

FRAMED: An AutoML Approach for Structural Performance Prediction of Bicycle Frames

“…Advanced AI-based design frameworks, such as Deep Generative Models (DGMs) have shown promising initial results on a variety of design problems. FRAMED is particularly well positioned to accelerate DGM development since not only do DGMs lack quality data and benchmark problems, most current DGMs do not account for design performance at all [18].…”

Design Target Achievement Index: A Differentiable Metric to Enhance Deep Generative Models in Multi-Objective Inverse Design

“…In a recent review of Deep Generative Models (DGMs), Regenwetter et al [1] discuss the application of DGMs across engineering design fields and analyze key limitations in the current state-of-the-art in DGM methodology. The authors suggest that successfully addressing several key challenges will be essential in the continued development of DGMs for engineering design.…”

“…In this section, we briefly summarize the state of the current research, as well as key drivers behind each of these four challenges. For a more detailed review and discussion, we refer the reader to [1].…”

“…Designs must be high-performing to be competitive, economically viable, and positively impact customers. As Regenwetter et al [1] note, most existing Deep Generative Models fail to account for design performance of any kind, perhaps since most DGMs are adapted from other disciplines such as computer vision, where notions of performance are limited to reconstruction quality or "realism," of generated samples. In general, approaches to incorporate design quality into DGMs fall into three broad categories:…”

“…Design Representation: Designs can be digitized using countless representation schemes, and as Regenwetter et al [1] note, many commonly used parameterizations severely limit the usability of generated designs in downstream tasks. Representations like images and voxelizations are easy to train on due to their spatially-structured properties, but generated designs are difficult to evaluate using computational analysis tools like Finite Element Analysis, and even more challenging to physically fabricate.…”