| 王阿久1 ，蔡开龙 2 ，何森3 ，戴郎杰 1 ，郑球4.基于BO-BiLSTM-RBF的航空发动机气路故障诊断[J].航空发动机,2025,51(6):101-107 |
| 基于BO-BiLSTM-RBF的航空发动机气路故障诊断 |
| Aeroengine Gas Path Fault Diagnosis Based on BO-BiLSTM-RBF |
| |
| DOI： |
| 中文关键词: 故障诊断 气路故障 双向长短时记忆网络 贝叶斯优化 径向基神经网络 特征提取 航空发动机 |
| 英文关键词:fault diagnosis gas path faults bidirectional long short-term memory network (BiLSTM) Bayesian optimization (BO) radial basis function (RBF) neural networks feature extraction aeroengine |
| 基金项目:江西省双千计划（JXSQ2018106057）资助 |
|
| 摘要点击次数: 2008 |
| 全文下载次数: 426 |
| 中文摘要: |
| 针对航空发动机气路故障诊断中机器学习算法数据分析能力不足、深度学习算法收敛速度和效果不佳易陷入过拟合的
问题，建立了一种基于贝叶斯算法（BO）优化双向长短时记忆（BiLSTM）网络结合径向基（RBF）神经网络的发动机气路故障诊断模
型。采用BO对BiLSTM网络的初始学习率和L2正则化系数进行寻优，利用优化后的BiLSTM网络对发动机气路数据进行特征提
取，使用RBF神经网络对提取到的抽象特征进行识别分类，开展故障诊断。结果表明：BO-BiLSTM-RBF模型准确率达95.88%，且
其泛化能力、抗干扰能力和鲁棒性等均优于BiLSTM-RBF网络、BiLSTM网络、支持向量机（SVM）、RBF神经网络以及反向传播
（BP）神经网络等模型，可有效地诊断发动机的单一或复合故障，为发动机故障诊断提供新的思路和方法。 |
| 英文摘要: |
| To address the issues of insufficient data analysis capabilities of machine learning algorithms and poor convergence speed
and ineffectiveness of deep learning algorithms prone to overfitting in aeroengine gas path fault diagnosis, a gas path fault diagnosis model
based on bayesian optimization(BO) optimised bidirectional long and short-term memory (BiLSTM) network combined with radial basis
function (RBF) neural network was established. BO was used to optimise the initial learning rate and L2 regularisation coefficient of the
BiLSTM network, and the optimised BiLSTM network was employed to extract features from the aeroengine gas path data, and the RBF
neural network was used to identify and classify the extracted abstract features for fault diagnosis. The results show that the BO-BiLSTM-
RBF model achieves an accuracy rate of 95.88%, and its generalization ability, interference resistance, and robustness outperform models
such as BiLSTM-RBF network, BiLSTM network, support vector machine (SVM), RBF neural network, and Back Propagation (BP) neural
network. It can effectively diagnose both single and compound faults in the aeroengine, providing a novel approach for aeroengine fault
diagnosis. |
| 查看全文 HTML 查看/发表评论 下载PDF阅读器
|