东南大学学报（自然科学版）2025，Vol.55Issue(3)：831-838,8.DOI:10.3969/j.issn.1001-0505.2025.03.025
基于机器学习的高压转子危险点温度预测
Dangerous components temperature prediction of high-pressure rotor based on machine learning
摘要
Abstract
To address the challenges of predicting the critical point temperatures of high-pressure rotors during operation,this study proposes a real-time multi-step rolling prediction method based on machine learning.Firstly,finite element analysis was employed to obtain temperature data for critical points on the rotor,provid-ing a reliable data foundation for subsequent modeling.To enhance data quality and prediction accuracy,data denoising and feature selection algorithms were applied to preprocess the operational data.Then,nine differ-ent machine learning models were designed,including convolutional neural networks(CNN),long short-term memory networks(LSTM),bidirectional LSTM(BiLSTM),and their variants combined with Attention mechanisms,as well as Transformer and temporal convolutional networks(TCN).Finally,the predictive per-formance of these models was evaluated through experimental comparative analysis,revealing that the CNN-BiLSTM-Attention and TCN models demonstrate the best accuracy,adapting to the complex operating conditions of high-pressure rotors.The proposed method provides an efficient predictive tool for temperature forecasting and safety management of high-pressure rotors,contributing to the safe,stable,and economical operations of power units.关键词
汽轮机转子/特征筛选/机器学习/温度预测Key words
turbine rotor/feature screening/machine learning/temperature prediction分类
数理科学引用本文复制引用
潘蕾,陈帅尧,王海涛,张君正..基于机器学习的高压转子危险点温度预测[J].东南大学学报（自然科学版）,2025,55(3):831-838,8.基金项目
国家重点研发计划资助项目(2022YFB4100403). （2022YFB4100403）
