基于非线性目标函数的时间卷积网络RUL预测  

Remaining Useful Life Prediction Based on Temporal Convolutional Network with Nonlinear Target Function

作　　者：刘斌[1] 许靖 霍美玲 崔学英[1] 谢秀峰[1] LIU Bin;XU Jing;HUO Mei-ling;CUI Xue-ying;XIE Xiu-feng(School of Applied Science,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学应用科学学院,太原030024

出　　处：《太原科技大学学报》2024年第2期211-216,共6页Journal of Taiyuan University of Science and Technology

基　　金：国家自然科学基金(11701406,11901134,12061091,11972019)。

摘　　要：机械设备剩余使用寿命(RUL)预测是系统维护策略的重要组成部分。在建立深度学习预测方法的目标函数的过程中,退化模型通常以分段线性函数的形式建立,异常值对预测结果的影响很容易被放大。提出了一种基于分段非线性退化的时间卷积网络回归模型。非线性函数能较好地描述传感器的退化趋势,减少线性模型预测引起的系统偏差。在美国航天局公布的涡扇发动机(C-MAPSS)数据集上验证了该模型的有效性,实验表明该模型比目标函数为分段线性函数的模型具有更低的误差,优于现有的一些预测方法。Remaining Useful Life(RUL)prediction of machinery is an important part of system maintenance strategy.In the process of establishing the target function of the deep learning prediction approach,the RUL degradation model is usually established as a piece wise linear function.The influence of outliers on the predicted results is easily amplified.This paper proposes a temporal convolutional network regression model with piece wise nonlinear degradation.The nonlinear function can better depict the sensor degradation trend and reduce the systematic deviation caused by linear model prediction.The framework is validated in NASA s Company-Modular Aero-Propulsion System Simulation data sets(C-MAPSS)dataset,experiments show that this model has lower error than the model whose target function is piecewise linear function,and is better than some existing prediction methods.

关 键 词：剩余使用寿命预测 深度学习 非线性目标函数 时间卷积网络 

分 类 号：O213.2[理学—概率论与数理统计] V235.13[理学—数学]