Special Issue: Machine Learning and Representation Issues in CAD/CAM

This is the authors' abstract. We don't add key points for this paper.

Machine learning (ML), a sub-field of artificial intelligence (AI), is profoundly reshaping various aspects of human life. Its application in engineering systems promises to address long-standing challenges, although it also introduces new questions. While the potential of ML is undeniable, integrating existing ML methods into computer-aided design and manufacturing (CAD/CAM) presents distinctive challenges. These challenges encompass representation, adaptation, and the development of novel ML techniques to enhance CAD/CAM systems for diverse design and manufacturing solutions.This special issue aims to explore and resolve issues related to effective engineering model representation for ML, the integration of neural networks and deep generative models into CAD/CAM, mathematical frameworks combining ML with CAD/CAM in geometry and topology, data interpretation, and physics-based learning. The issue features ten papers that delve into various topics, including material prediction for assemblies, robotics mechanism design, multi-modal ML in engineering design, data-driven component segmentation in engineering drawings, evaluating assembly-part semantic knowledge in language models, three-dimensional slice reconstruction from high-resolution 2D images, physics-informed neural networks to expedite thermal simulations in additive manufacturing, transfer learning for defect detection in steel strips, AI-aided hull form design for energy-efficient unmanned underwater vehicles, and real-time high-precision calibration of quadruped robots using machine vision and artificial neural networks. Below, we provide concise summaries of each of the ten papers published in this special issue.The representation of mechanisms and their attributes for ML applications within engineering has hitherto remained unexplored. In their paper entitled “An Invariant Representation of Coupler Curves Using a Variational AutoEncoder: Application to Path Synthesis of Four-Bar Mechanisms,” Nurizada and Purwar address this void by investigating the representation and synthesis of coupler curves in planar mechanisms using a deep neural network. They emphasize that the effective representation of coupler curves in neural networks remains an underexplored domain. The study conducts a comparative analysis of different representations of four-bar coupler curves and illustrates the capacity of a variational autoencoder (VAE) to provide a unified representation for coupler curves, irrespective of the initial input representation. Furthermore, the paper introduces an innovative approach that combines VAE with a fully connected neural network to generate dimensional parameters for four-bar linkage mechanisms, offering potential for automated mechanism design in robotics and machinery.In a similar vein, a distinct challenge persists in ML-based representation when predicting material properties for assembly bodies. In the paper titled “HG-CAD: Hierarchical Graph Learning for Material Prediction and Recommendation in Computer-Aided Design,” authored by Bian et al., the authors introduce HG-CAD, a machine learning architecture utilizing a hierarchical graph representation for automated material prediction in assembly bodies. This prediction is based on recommendations derived from previous designs. The study underscores that a learning approach that combines geometry with topology through graphs to capture neighborhood context provides enhanced predictions compared to methods relying solely on geometric attributes.Additionally, Song et al. present a comprehensive review paper titled “Multi-Modal Machine Learning in Engineering Design: A Review and Future Directions,” discussing the integration of multi-modal machine learning (MMML) into engineering design. This involves the amalgamation of various data modalities, such as text, 2D images, and 3D shapes, along with their associated processing methods. The review covers a wide spectrum of aspects, including methodologies for handling multi-modal data and the potential application of MMML in design tasks encompassing knowledge retrieval, evaluation, and synthesis. The paper also underscores the potential benefits and challenges entailed in the utilization of MMML, such as the imperative need for large, labeled multi-modal design datasets, robust algorithms, incorporation of domain knowledge, and the management of data heterogeneity and noise.Moreover, Zhang et al. delve into the challenges associated with interpreting raster-based part drawings while automating the computerized interpretation of design and manufacturing information. Their paper, “A Data Augmentation Method for Data-Driven Component Segmentation of Engineering Drawings,” introduces a data augmentation approach aimed at enhancing the utility of trainable computer vision techniques for automatic machine interpretations. The approach can be applied to various applications, including similarity search, drawing classification, drawing indexing, and manufacturing process identification. In a related paper titled “What’s in a Name? Evaluating Assembly-Part Semantic Knowledge in Language Models Through User-Provided Names in Computer Aided Design Files,” authored by Meltzer et al., the authors propose utilizing natural language names used by designers in CAD software as a source of design and manufacturing knowledge. They demonstrate that large language models contain valuable domain-specific information for handling CAD data and engineering-related tasks. Nurizada and Purwar in another paper titled “Transforming Hand-Drawn Sketches of Linkage Mechanisms into Their Digital Representation” present a novel method utilizing deep neural networks for digitally transforming and simulating n-bar planar linkages, which consist of revolute and prismatic joints, based on hand-drawn sketches. It combines topological knowledge of linkage mechanisms with convolutional deep neural networks to recognize sketches. They create a dataset of synthetic images resembling hand-drawn sketches and fine-tune a deep neural network to detect objects representing joints and links in sketches. While the algorithm is effective in converting hand-drawn sketches into digital representations, this work has practical implications for enhancing communication and analysis of planar mechanisms.The paper titled “Three-Dimensional-Slice-Super-Resolution-Net: A Fast Few Shooting Learning Model for 3D Super-Resolution Using Slice-Up and Slice-Reconstruction” by Lin et al. introduces an approach for enhancing the resolution of 3D models with minimal data acquisition resources and storage space. This involves the utilization of a pre-trained deep 2D convolutional neural network to enhance the resolution of 2D images and reconstruct them into high-resolution 3D models. In their paper titled “Accelerating Thermal Simulations in Additive Manufacturing by Training Physics-Informed Neural Networks with Randomly-Synthesized Data,” Chen et al. underscore the significance of comprehending the temperature history in additively-manufactured parts within fusion-based additive manufacturing. They investigate the use of physics-informed neural networks to predict the time-varying temperature distribution during laser powder bed fusion manufacturing.Chen et al., in their paper titled “Cross-Domain Transfer Learning for Galvanized Steel Strips Defect Detection and Recognition,” focus on utilizing deep learning for defect detection in strip steel production, an essential aspect of industrial quality inspection. They leverage transfer learning to overcome data limitations in industrial settings and develop a deep learning model that combines defect region detection and feature extraction for precise defect classification. In a paper authored by Ao et al., titled “Artificial Intelligence Aided Design of Hull Form of Unmanned Underwater Vehicles for Minimization of Energy Consumption,” the authors emphasize the significance of designing efficient hulls for unmanned underwater vehicles (UUVs) to enhance their energy endurance. They introduce a data-driven approach to predict UUV resistance with different hull shapes under varying conditions.Furthermore, Li et al. in their paper titled “An Automatic High-Precision Calibration Method of Legs and Feet for Quadruped Robots Using Machine Vision and Artificial Neural Networks” introduce an online intelligent calibration method utilizing machine vision and artificial neural networks to simplify the calibration process and enhance accuracy. This method encompasses the identification of markers on the robot’s legs through target detection, the calculation of marker coordinates, and the construction of an error model based on neural networks to rectify joint angle errors.In conclusion, this special issue features innovative approaches for adequate representation and techniques applicable to a diverse range of ML applications in engineering design and manufacturing challenges. It is our hope that readers will find these papers insightful and thought-provoking, fostering further research in this field. The guest editors of this special issue extend their sincere gratitude to all the authors and reviewers for their remarkable contributions. Additionally, we acknowledge the unwavering support of Professor S. K. Gupta, the former Editor-in-Chief of JCISE, and Professor Yan Wang, the current Editor-in-Chief of JCISE. Our profound appreciation goes to Ms. Amy Suski for her timely reminders and administrative support.