An official website of the United States government
Abstract Function is defined as the ensemble of tasks that enable the product to complete the designed purpose. Functional tools, such as functional modeling, offer decision guidance in the early phase of product design, where explicit design decisions are yet to be made. Function-based design data is often sparse and grounded in individual interpretation. As such, function-based design tools can benefit from automatic function classification to increase data fidelity and provide function representation models that enable function-based intelligent design agents. Function-based design data is commonly stored in manually generated design repositories. These design repositories are a collection of expert knowledge and interpretations of function in product design bounded by function-flow and component taxonomies. In this work, we represent a structured taxonomy-based design repository as assembly-flow graphs, then leverage a graph neural network (GNN) model to perform automatic function classification. We support automated function classification by learning from repository data to establish the ground truth of component function assignment. Experimental results show that our GNN model achieves a micro-average F1-score of 0.617 for tier 1 (broad), 0.624 for tier 2, and 0.415 for tier 3 (specific) functions. Given the imbalance of data features and the subjectivity in the definition of product function, the results are encouraging. Our efforts in this paper can be a starting point for more sophisticated applications in knowledge-based CAD systems and Design-for-X consideration in function-based design.
more »
« less
Toward Automated Functional Modeling: An Association Rules Approach for Mining the Relationship between Product Components and Function
Abstract The objective of this research is to support DfX considerations in the early phases of design. In order to do conduct DfX, designers need access to pertinent downstream knowledge that is keyed to early stage design activities and problem knowledge. Product functionality is one such “key” connection between early understanding of the design problem and component choices which dictate product performance and impact, and repositories of design knowledge are one way to archive such design knowledge. However, curation of design knowledge is often a time-consuming activity requiring expertise in product modeling. In this paper, we explore a method to automate the populating of design repositories to support the overall goal of having up-to-date repositories of product design knowledge. To do this, we mine information from an existing repository to better understand the relationships between the components, functions, and flows of products. The resulting knowledge can be applied to automate functional decompositions once a product's components have been entered and thus reliably provide that “key” between early design activities and the later, component dependent characteristics.
more »
« less
- Award ID(s):
- 1826469
- PAR ID:
- 10160481
- Date Published:
- Journal Name:
- Proceedings of the Design Society: International Conference on Engineering Design
- Volume:
- 1
- Issue:
- 1
- ISSN:
- 2220-4342
- Page Range / eLocation ID:
- 1713 to 1722
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Machine learning can be used to automate common or time-consuming engineering tasks for which sufficient data already exist. For instance, design repositories can be used to train deep learning algorithms to assess component manufacturability; however, methods to determine the suitability of a design repository for use with machine learning do not exist. We provide an initial investigation toward identifying such a method using “artificial” design repositories to experimentally test the extent to which altering properties of the dataset impacts the assessment precision and generalizability of neural networks trained on the data. For this experiment, we use a 3D convolutional neural network to estimate quantitative manufacturing metrics directly from voxel-based component geometries. Additive manufacturing (AM) is used as a case study because of the recent growth of AM-focused design repositories such as GrabCAD and Thingiverse that are readily accessible online. In this study, we focus only on material extrusion, the dominant consumer AM process, and investigate three AM build metrics: (1) part mass, (2) support material mass, and (3) build time. Additionally, we compare the convolutional neural network accuracy to that of a baseline multiple linear regression model. Our results suggest that training on design repositories with less standardized orientation and position resulted in more accurate trained neural networks and that orientation-dependent metrics were harder to estimate than orientation-independent metrics. Furthermore, the convolutional neural network was more accurate than the baseline linear regression model for all build metrics.more » « less
-
null (Ed.)Engineering designers currently use downstream information about product and component functions to facilitate ideation and concept generation of analogous products. These processes, often called Function-Based Design, can be reliant on designer definitions of product function, which are inconsistent from designer to designer. In this paper, we employ supervised learning algorithms to reduce the variety of component functions that are available to designers in a design repository, thus enabling designers to focus their function-based design efforts on more accurate, reduced sets of potential functions. To do this, we generate decisions trees and rules that define the functions of components based on the identity of neighboring components. The resultant decision trees and rulesets reduce the number of feasible functions for components within a product, which is of particular interest for use by novice designers, as reducing the feasible functional space can help focus the design activities of the designer. This reduction was evident in both case studies: one exploring a component that is known to the designer, and the other looking at defining function of an unrecognizable component. The work presented here contributes to the recent popularity of using product data in data-driven design methodologies, especially those focused on supplementing designer cognition. Importantly, we found that this methodology is reliant on repository data quality, and the results indicate a need to continue the development of design repository data schemas with improved data consistency and fidelity. This research is a necessary precursor for the development of function-based design tools, including automated functional modeling.more » « less
-
null (Ed.)Populating the different types of data for a design repository is a difficult and time-consuming task. In this work, we report on techniques to automate the population of data related to product function. We explore a preliminary method to automate the generation of the functional chains of components from new products based on hierarchical data from an existing design repos- itory. We use datasets of various scale and specificity to find correlations between functions and flows for components of products in the Design Repos- itory. We use the results to predict the most likely functions and flows for a component, and then verify the accuracy of our algorithm by cross-validating a subsection of the data against the automation results. We apply existing grammar rules to order the functions and flows in a linear functional chain. Ultimately, these findings suggest methods for further automating the process of generating functional models.more » « less
-
This work-in-progress paper shares findings of the early stage of a 3-year research funded by the National Science Foundation. The major aim of the project is to advance engineering and mathematics (EM) education theory and practice related to students’ self-regulation of cognition and motivation skills during problem-solving activities. The self-regulation includes students’ metacognitive knowledge about task (MKT) and self-regulation of cognition (SRC). The motivational component of self-regulation (SRM) includes self-control of the motivation needed to maintain the level of engagement and deliberate practice necessary for scientific thinking and reasoning. To be effective problem-solvers, students must understand the relationship between the MKT, SRC and SRM throughout the problem-solving activities. Four research questions will guide the research: (1) How do students perceive their self-regulation of cognition (SRC) and motivation (SRM) skills for generic problem-solving activities in EM courses; (2) How does students’ metacognitive knowledge about problem-solving tasks (MKT) inform their Task interpretation?; (3) How do students’ SRC and SRM dynamically evolve?; and (4) How do students’ SRC and SRM reflect their perceptions of self-regulation of cognition and motivation for generic EM problemsolving activities? A sequential mixed-methods research design involving quantitative and qualitative methods are used to develop complementary coarse- and fine-grained understandings of undergraduate students’ SRC and SRM during academic problem-solving activities. Two 2nd year EM courses: Engineering Statics, and Ordinary Differential Equations were purposefully selected for the contexts of the study. One hundred forty two students from both courses were invited and participated in quantitative data collection using two validated surveys during spring 2022 semester. Later in the semester, qualitative data will be generated with twenty students in both courses through one-on-one interviews with students and course instructors, think-aloud protocols with students, and classroom observations. Coarse-grained understandings of students’ SRC and SRM are currently developed through analysis of quantitative data collected using self-report surveys (i.e., BRoMS and PMI). Fine-grained understandings of students’ SRC and SRM will be developed through analysis of qualitative data gathered via one-on-one interviews, think-aloud protocols, classroom observations, and course artifacts gathered as students engage in EM problem-solving activities.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1017/dsi.2019.177
