车轴应力预测的物理信息神经网络方法
An Physics-informed Neural Network for Axle Stress Prediction
岳晨语 1彭岳星1
作者信息
- 1. 北京邮电大学信息与通信工程学院北京 100876
- 折叠
摘要
作为工业生产中最常见，最重要的结构之一,轴结构具备承担载荷、传递运动等功能，广泛应用于绝大多数工业机械中。轴结构通常长期承受交变载荷,容易累积金属疲劳损伤,并在累积损伤达到阈值后产生裂纹极大降低性能，严重时甚至发生瞬间断裂从而引发重大事故。然而,由于轴结构在承受载荷时应力分布不可见,且微小的疲劳裂纹难以进行可靠检测，所以对于轴结构工作状态的可靠检测，是保证轴结构安全运行的可靠手段。本文针对一种重载列车车轴的应力预测问题，在机理建模与有限元仿真的基础上提出了一种机理与数据联合驱动的应力预测物理信息神经网络模型,实现对车轴变径截面应力分布数值预测。
Abstract
As one of the most common and crucial structures in industrial production, the shaft structure serves functions such as bearing loads and transmitting motion, and is widely utilized in the vast majority of industrial machinery. Axles typically endure long-term alternating loads, and may accumulate metal fatigue damage. Once the cumulative damage reaches a threshold, cracks emerge, which significantly reduce performance, and might leads to instantaneous fracture and then results in accidents in some cases. Due to the invisible stress distribution and the difficulty in reliably detecting minor fatigue cracks, reliable detection of the working state of the shaft structure is a dependable means to ensure its safe operation. To address the stress prediction issue of a heavy-duty train axle, a physics-informed neural network is proposed for stress prediction, which can reliably predict the stress distribution across the varying diameter sections of the axle.关键词
重载列车车轴/应力分布/有限元建模/物理信息神经网络Key words
Heavy-haul train axle/stress distribution/finite element modeling/physics-informed neural network引用本文复制引用
岳晨语,彭岳星.车轴应力预测的物理信息神经网络方法[EB/OL].(2026-01-13)[2026-01-18].http://www.paper.edu.cn/releasepaper/content/202601-16.学科分类
铁路运输工程
评论