首页|基于自注意力优化隐藏细胞状态的行星齿轮系统可靠性分析

基于自注意力优化隐藏细胞状态的行星齿轮系统可靠性分析

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
针对部分随机变量呈现高度非线性以及变量相互之间存在相关性难以建立可靠性分析模型等问题,提出了一种基于自注意力机制优化隐藏细胞状态的长短期记忆神经网络模型.首先,建立长短期记忆神经网络,并在每个单元的隐藏状态输出位置引入了自注意力机制,通过重新分配注意力权重来反映序列随机变量之间的相互关系.然后,通过数学推导,得到响应值与随机变量之间的显式函数关系,并使用一次二阶矩法计算可靠度.最后,构建了三级行星齿轮三维有限元模型,研究了在不同影响因素作用下齿轮的破坏过程和强度退化速度.结果表明:模型的拟合均方根误差保持在0.036以下,表明该模型对于高度非线性和数据之间有相关性的问题,具有一定的指导意义.
Reliability Analysis of Hidden Cell State Optimization Based on Self-attention
Aiming at the problems that some random variables are highly nonlinear and the correlation between variables is difficult to establish a reliability analysis model,a long-term and short-term memory neural network model based on self-attention mechanism is proposed to optimize the hidden cell state.Firstly,a long-term and short-term memory neural network is established,and a self-attention mechanism is introduced at the hidden state output position of each unit.The relation between sequence random variables is reflected by redistributing the attention weight.Then,through mathematical derivation,the explicit functional relation between the response value and the random variable is obtained,and the first-order second-moment method is used to calculate the reliability.Finally,a three-dimensional finite element model of a three-stage planetary gear is constructed to study the failure process and strength degradation rate of the gear under different influencing factors.The results show that the fitting root mean square error of the model is kept below 0.036,which indicates that the model has certain guiding significance for the problem of highnonlinearity and correlation between data.

high nonlinearityvariable correlationself-attention mechanismhidden cell statelong short-term memory neural networksplanetary gear

闵世成、高鹏、谢里阳

展开 >

辽宁石油化工大学 机械工程学院,辽宁抚顺 113001

东北大学 现代设计与分析研究所,沈阳 110819

东北大学 航空动力装备振动及控制教育部重点实验室,沈阳 110819

高度非线性 变量相关性 自注意力机制 隐藏细胞状态 长短期记忆神经网络 行星齿轮

抚顺市"抚顺英才计划"资助项目辽宁省高等学校创新人才支持计划资助辽宁省教育厅科学研究经费项目

FSYC202107014LR2017070L2019019

2024

机械设计与研究
上海交通大学

机械设计与研究

CSTPCD北大核心
影响因子：0.531
ISSN：1006-2343
年,卷(期)：2024.40(4)