摘要详情

ID / 提交时间

37 / 2025-03-28 13:42:54

标题

LNO-PINN: Laplace Neural Operator Based Physically Informed Neural Networks for Dynamic System Fault Detection

关键字

dynamic systems, fault detection, physically informed neural networks, laplace neural operator

主题及专题

S07 系统运行安全智能监测、诊断与控制技术

状态

全文待审

作者

蕊黄 / 重庆大学

可张 / 重庆大学

家喜张 / 重庆大学

摘要

Fault detection in dynamic systems is essential for ensuring safety, reliability, and optimal performance across various engineering domains. However, it faces significant challenges in complex systems exhibiting nonlinearity, high-order dynamics, and time-variant characteristics. Traditional fault detection methods prove inadequate for addressing the complexity of modern systems. This study introduces a novel framework termed the Laplace Neural Operator-Based Physically Informed Neural Network (LNO-PINN). By integrating physical laws with machine learning, the algorithm employs the Laplace neural operator to effectively handle complex spatio-temporal dependencies. This innovative approach enhances its capability to tackle nonlinear characteristics, high-order dynamics, and time-varying behaviors. Extensive numerical simulations validate its effectiveness in fault detection for complex dynamic systems, particularly in systems exhibiting high-order nonlinear dynamics, thereby providing a groundbreaking methodology for this research field.