Search for: All records

Data-driven generative design (DDGD) methods utilize deep neural networks to create novel designs based on existing data. The structure-aware DDGD method can handle complex geometries and automate the assembly of separate components into systems, showing promise in facilitating creative designs. However, determining the appropriate vectorized design representation (VDR) to evaluate 3D shapes generated from the structure-aware DDGD model remains largely unexplored. To that end, we conducted a comparative analysis of surrogate models’ performance in predicting the engineering performance of 3D shapes using VDRs from two sources: the trained latent space of structure-aware DDGD models encoding structural and geometric information and an embedding method encoding only geometric information. We conducted two case studies: one involving 3D car models focusing on drag coefficients and the other involving 3D aircraft models considering both drag and lift coefficients. Our results demonstrate that using latent vectors as VDRs can significantly deteriorate surrogate models’ predictions. Moreover, increasing the dimensionality of the VDRs in the embedding method may not necessarily improve the prediction, especially when the VDRs contain more information irrelevant to the engineering performance. Therefore, when selecting VDRs for surrogate modeling, the latent vectors obtained from training structure-aware DDGD models must be used with caution, although they are more accessible once training is complete. The underlying physics associated with the engineering performance should be paid attention. This paper provides empirical evidence for the effectiveness of different types of VDRs of structure-aware DDGD for surrogate modeling, thus facilitating the construction of better surrogate models for AI-generated designs.

 

Abstract In this paper, we present a predictive and generative design approach for supporting the conceptual design of product shapes in 3D meshes. We develop a target-embedding variational autoencoder (TEVAE) neural network architecture, which consists of two modules: (1) a training module with two encoders and one decoder (E2D network) and (2) an application module performing the generative design of new 3D shapes and the prediction of a 3D shape from its silhouette. We demonstrate the utility and effectiveness of the proposed approach in the design of 3D car body and mugs. The results show that our approach can generate a large number of novel 3D shapes and successfully predict a 3D shape based on a single silhouette sketch. The resulting 3D shapes are watertight polygon meshes with high-quality surface details, which have better visualization than voxels and point clouds, and are ready for downstream engineering evaluation (e.g., drag coefficient) and prototyping (e.g., 3D printing). 

Video analysis tools such as Tracker are used to study mechanical motion captured by photography. One can also imagine a similar tool for tracking thermal motion captured by thermography. Since its introduction to physics education, thermal imaging has been used to visualize phenomena that are invisible to the naked eye and teach a variety of physics concepts across different educational settings. But thermal cameras are still scarce in schools. Hence, videos recorded using thermal cameras such as those featured in “YouTube Physics” are suggested as alternatives. The downside is that students do not have interaction opportunities beyond playing those videos. 

Learning analytics, referring to the measurement, collection, analysis, and reporting of data about learners and their contexts in order to optimize learning and the environments in which it occurs, is proving to be a powerful approach for understanding and improving science learning. However, few studies focused on leveraging learning analytics to assess hands-on laboratory skills in K-12 science classrooms. This study demonstrated the feasibility of gauging laboratory skills based on students’ process data logged by a mobile augmented reality (AR) application for conducting science experiments. Students can use the mobile AR technology to investigate a variety of science phenomena that involve concepts central to physics understanding. Seventy-two students from a suburban middle school in the Northeastern United States participated in this study. They conducted experiments in pairs. Mining process data using Bayesian networks showed that most students who participated in this study demonstrated some degree of proficiency in laboratory skills. Also, findings indicated a positive correlation between laboratory skills and conceptual learning. The results suggested that learning analytics provides a possible solution to measure hands-on laboratory learning in real-time and at scale. 

During the COVID-19 pandemic, many students lost opportunities to explore science in labs due to school closures. Remote labs provide a possible solution to mitigate this loss. However, most remote labs to date are based on a somehow centralized model in which experts design and conduct certain types of experiments in well-equipped facilities, with a few options of manipulation provided to remote users. In this paper, we propose a distributed framework, dubbed remote labs 2.0, that offers the flexibility needed to build an open platform to support educators to create, operate, and share their own remote labs. Similar to the transformation of the Web from 1.0 to 2.0, remote labs 2.0 can greatly enrich experimental science on the Internet by allowing users to choose and contribute their subjects and topics. As a reference implementation, we developed a platform branded as Telelab. In collaboration with a high school chemistry teacher, we conducted remote chemical reaction experiments on the Telelab platform with two online classes. Pre/post-test results showed that these high school students attained significant gains (t(26)=8.76, p<0.00001) in evidence-based reasoning abilities. Student surveys revealed three key affordances of Telelab: live experiments, scientific instruments, and social interactions. All 31 respondents were engaged by one or more of these affordances. Students behaviors were characterized by analyzing their interaction data logged by the platform. These findings suggest that appropriate applications of remote labs 2.0 in distance education can, to some extent, reproduce critical effects of their local counterparts on promoting science learning.