.png){kind=icon}
返回
A Crashworthiness Design Framework based on Temporal-Spatial Feature Extraction and Multi-Target Sequential Modeling
.png){kind=icon}
PRE
AI
DOI：10.1016/j.tws.2024.112694.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Temporal-spatial crashworthiness design remains a challenging issue in engineering applications. Metamodeling techniques have been widely used to improve design efficiency by reducing the need for extensive experiments or simulations. However, these methods often fail to capture the essential information of temporal and spatial during the dynamical procedure. In this study, a novel multi-target modeling and optimization framework is introduced to overcome these limitations. This framework utilizes autocorrelation functions to identify key temporal-spatial segments, ensuring that the most influential factors are captured, and then builds a metamodel using multi-target regression techniques and partial autocorrelation functions, effectively capturing the complex relationships among different time steps. An adaptive sampling strategy is also employed to generate additional training data according to the objective functions, thereby enhancing the accuracy and robustness of the metamodels. These improvements enable a more accurate and interpretable integration of temporal-spatial information compared to popular methods. The effectiveness of the proposed framework is demonstrated through its successful implementation in optimizing crashworthiness across diverse scenarios: a cylindrical tube, a multi-cell energy-absorbing structure, and a B-pillar designed to withstand side impacts. The results show that the proposed method provides reliable predictions for subsequent optimization tasks and has the potential to address complex crashworthiness design challenges by comprehensively considering temporal-spatial information.
Keyword:
Crashworthiness design
Multi-target regression
Machine learning
Temporal sequential modeling
Temporal-spatial feature extraction
Adaptive sampling strategy
期刊
T
IF:
6.6
论文数:
1.1W
被引数:
4.0W
机构
引用论文
.png){kind=icon}
Data-driven spatiotemporal modeling for structural dynamics on irregular domains by stochastic dependency neural estimation通过随机依赖神经估计在不规则区域上对结构动力学进行数据驱动的时空建模
PREAI
分享收藏分享收藏Crashworthiness in preliminary design: Mean crushing force prediction for closed-section thin-walled metallic structures初步设计中的耐撞性: 封闭截面薄壁金属结构的平均压碎力预测
OAAI
OAAI分享收藏Finite-time control of discrete-time semi-Markov jump linear systems: A self-triggered MPC approach离散半马尔可夫跳变线性系统的有限时间控制: 一种自触发MPC方法
PREAI
分享收藏
查看更多内容