|
| |
基于RIDLA的卫星承重结构螺栓连接非线性滞回建模 |
Nonlinear Hysteresis Modeling of Bolt Connections in Satellite Load-Carrying Structures Using a Residual Improvement Deep Learning Algorithm |
投稿时间:2024-06-27 修订日期:2024-07-29 |
DOI: |
中文关键词: 承重结构,螺栓连接,滞回模型,残差改进,深度学习 |
英文关键词:Load-carrying structure Bolt connection Hysteresis model residual improvement Deep learning |
基金项目:国家重点研发计划:2021YFA1003501,航空科学基金:2022Z061001资助. |
|
摘要点击次数: 82 |
全文下载次数: 0 |
中文摘要: |
准确构建螺栓连接处的非线性滞回曲线模型对卫星承重结构的减振和安全性能评估至关重要。传统计算模型的时域分析方法需要大量时间成本,典型的数据驱动模型难以构建高精度的滞回模型。针对上述挑战,提出了一种新的残差改进的深度学习算法(Residual Improvement Deep Learning Algorithm, RIDLA),用于构建螺栓连接处位移与力的滞回曲线模型。该算法充分利用长短期记忆(LSTM)神经网络拟合时间序列非线性关系的能力,通过实测响应与计算残差之间的交互迭代,构建了多级别的残差改进深度学习模型,从而实现了对螺栓连接处滞回模型的准确建模。使用某卫星承重结构的子部件循环加载实验数据验证了RIDLA方法的性能。结果表明 RIDLA实现了对螺栓连接处的位移和力滞回曲线高度精确的预测。此外,RIDLA方法有可能应用于预测其他复杂非线性系统的动态响应。 |
英文摘要: |
(1. State Key Laboratory of Structural Analysis, Optimization and CAE Software for Industrial Equipment, Dalian 116024, China;2. Ningbo Research Institute of Dalian University of Technology, Ningbo 315000, China):Accurately constructing the nonlinear hysteresis loop model at the bolt connection is crucial for the vibration reduction and safety performance evaluation of the satellite load-carrying structure. Traditional time-domain analysis methods of computational models require substantial time costs, and typical data-driven models struggle to construct high-precision hysteresis models. To address these challenges, a novel Residual Improvement Deep Learning Algorithm (RIDLA) is proposed for constructing the hysteresis loop model of displacement and force at the bolt connection. The algorithm fully leverages the capacity of Long Short-Term Memory (LSTM) neural networks to fit nonlinear relationships in time series. It adopts an innovative approach by creating a multi-level residual improvement deep learning model that iteratively refines predictions based on measured responses, resulting in highly accurate modeling of hysteresis at bolt connections. The performance of the RIDLA method was validated using experimental data from cyclic loading of a subcomponent of satellite load-carrying structure. The findings demonstrate that RIDLA achieves highly accurate predictions of the displacement and force hysteresis loop at the bolt connection. Additionally, the RIDLA method could be applied to predict the dynamic responses of other complex non-linear systems. |
查看/发表评论 下载PDF阅读器 |