离心泵叶轮动应力的长短期记忆网络预测与疲劳分析  

Dynamic stress prediction and fatigue analysis of centrifugal pump impeller based on long short-term memory network

作　　者：乔凤权 孙毅 吴松泉 张鹏昌 方铭坤 张方芳 陶然[2] QIAO Fengquan;SUN Yi;WU Songquan;ZHANG Pengchang;FANG Mingkun;ZHANG Fangfang;TAO Ran(Jiangsu Pumping Station Technology Co.,Ltd.of South-to-North Water Diversion Project,Yangzhou 225000,China;College of Water Resources and Civil Engineering,China Agricultural University,Beijing 100083,China)

机构地区：[1]南水北调江苏泵站技术有限公司,江苏扬州225000 [2]中国农业大学水利与土木工程学院,北京100083

出　　处：《水力发电学报》2024年第11期92-102,共11页Journal of Hydroelectric Engineering

基　　金：国家自然科学基金(U22A20238)。

摘　　要：低比转速离心泵在运行时会伴随着应力集中问题,应力集中和动应力变化可能导致低比转速离心泵发生疲劳损伤。传统的有限元方法在分析应力疲劳问题时受限于预测能力不足。本文提出了一种基于长短期记忆网络(LSTM)的离心泵疲劳寿命预测方法,通过利用长短期记忆网络对额定工况和小流量工况下动应力时间序列信号进行训练及预测,并结合疲劳寿命模型对不同运行条件下的离心泵运行寿命进行预测。结果表明:LSTM模型能够准确捕捉动应力的变化趋势,额定工况下的动应力拟合效果明显优于小流量工况。额定工况下动应力预测集MAPE值相较于小流量工况下降了21.33%。预测动应力为2.28×10^(7)Pa时,其运行寿命可达3×10^(8)的循环次数,满足高强度运行的寿命要求。本文所提出的方法能够有效提高离心泵的运行寿命预测精度,为离心泵的管理和维护提供更可靠的支持。Low-speed centrifugal pumps are prone to stress concentration issues during operation,which,along with dynamic stress variations,may lead to fatigue damage.Traditional finite element methods are limited in predicting fatigue life accurately.This paper describes a new fatigue life prediction method based on Long Short-Term Memory(LSTM)networks.It predicts the pump's operational life under different conditions,using LSTM networks to train and predict dynamic stress time series signals under rated and low-flow conditions,and integrating fatigue life models.Results show that this LSTM model captures stress variations accurately and can achieve fitting performance under rated conditions superior to that under low-flow conditions.The prediction set under rated conditions has a Mean Absolute Percentage Error(MAPE)of 21.3%lower than that under low-flow conditions.The operational life achieves 3×10^(8)cycles under predicted stress of 2.28×10^(7)Pa,meeting the life requirements of high-intensity operation.As a reliable,useful tool for pump management and maintenance,our method improves significantly the accuracy of centrifugal pump operational life prediction.

关 键 词：低比转速离心泵 长短期记忆网络 动应力 疲劳寿命 

分 类 号：TN929.5[电子电信—通信与信息系统]