Pilot Demonstration of a Strengthening Method for Steel-Bolted Connections Using Pre-Formable Carbon Fiber Cloth with VaRTM

Matsui, Tetsuya; Suzuki, Kohei; Sato, Sota; Kubokawa, Yuki; Nakamoto, Daiki; Davaakhishig, Shijir; Matsumoto, Yukihiro

doi:10.3390/ma14092184

Cited by 3 publications

(2 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These gates can learn the optimal way to remember useful information from previous states and thus decide upon the current state. Equations ( 5)- (7) shows the procedure of computing different gates and hidden states, where i t , f t , and o t are the input, forget, and output gates, respectively, Equation ( 9) while c t denotes the cell's internal state and Equation (10) h t is the hidden state.…”

Section: 𝑧 = 𝜎(𝑤 ℎ + 𝑤 𝑥 + 𝑏 )mentioning

confidence: 99%

“…External forces, including gravity, wind, and collisions with objects or the environment, are then applied to the cloth simulation, which affects the motion and deformation of the cloth [7]. These forces are accurately computed and then exerted upon the cloth's particles or vertices, thus resulting in an accurate simulation of their movement and interplay.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Supervised Video Cloth Simulation: Exploring Softness and Stiffness Variations on Fabric Types Using Deep Learning

Mao,

Va,

Lee

et al. 2023

Applied Sciences

View full text Add to dashboard Cite

Physically based cloth simulation requires a model that represents cloth as a collection of nodes connected by different types of constraints. In this paper, we present a coefficient prediction framework using a Deep Learning (DL) technique to enhance video summarization for such simulations. Our proposed model represents virtual cloth as interconnected nodes that are subject to various constraints. To ensure temporal consistency, we train the video coefficient prediction using Gated Recurrent Unit (GRU), Long-Short Term Memory (LSTM), and Transformer models. Our lightweight video coefficient network combines Convolutional Neural Networks (CNN) and a Transformer to capture both local and global contexts, thus enabling highly efficient prediction of keyframe importance scores for short-length videos. We evaluated our proposed model and found that it achieved an average accuracy of 99.01%. Specifically, the accuracy for the coefficient prediction of GRU was 20%, while LSTM achieved an accuracy of 59%. Our methodology leverages various cloth simulations that utilize a mass-spring model to generate datasets representing cloth movement, thus allowing for the accurate prediction of the coefficients for virtual cloth within physically based simulations. By taking specific material parameters as input, our model successfully outputs a comprehensive set of geometric and physical properties for each cloth instance. This innovative approach seamlessly integrates DL techniques with physically based simulations, and it therefore has a high potential for use in modeling complex systems.

show abstract

Section: 𝑧 = 𝜎(𝑤 ℎ + 𝑤 𝑥 + 𝑏 )mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%