排灌机械工程学报2025，Vol.43Issue(8)：786-794,9.DOI:10.3969/j.issn.1674-8530.23.0270
基于RIME-VMD-LSTM法的流固耦合作用下叶片裂纹特征
Blade crack characteristics under fluid-structure interaction based on RIME-VMD-LSTM method
摘要
Abstract
To enable timely crack detection in hydraulic turbine runner blades,ensure unit health mo-nitoring,and enhance operational safety,an intelligent fault diagnosis method was proposed based on computational fluid dynamics(CFD),rime optimization algorithm(RIME),variational mode decom-position(VMD),and long short-term memory(LSTM)neural network.First,the flow field was simu-lated using CFD,and the results were imported into finite element analysis(FEA)software through fluid-structure interaction to obtain time-domain vibration signals for both healthy and cracked runner blades.Subsequently,the modal component number(K)and penalty factor(α)of VMD were opti-mized by RIME.The optimized VMD was then employed to decompose the vibration signals into multi-ple intrinsic mode components.Finally,these components were fed into the LSTM neural network for feature extraction and fault identification.The results demonstrate that the proposed method eliminates the economic costs associated with physical crack sample acquisition,significantly reduces the develop-ment cycle,and achieves highly accurate blade crack detection.For radial and axial vibration signals,the overall fault recognition accuracy reaches 93.033 0%and 92.893 9%,respectively.关键词
叶片裂纹检测/CFD/流固耦合/RIME-VMD/LSTM神经网络Key words
blade crack detection/CFD/fluid-structure interaction/RIME-VMD/LSTMneural network分类
农业科技引用本文复制引用
朱敏,段娟,钱晶,曾云,单蓉..基于RIME-VMD-LSTM法的流固耦合作用下叶片裂纹特征[J].排灌机械工程学报,2025,43(8):786-794,9.基金项目
国家自然科学基金资助项目(52309113,52269020) （52309113,52269020）
云南省自然科学基金资助项目(202201AU070114,202204BW050001) （202201AU070114,202204BW050001）
