| 吴 超1 ，陈磊2 ，刘渊1 ，周绮凤 2 ，王奕首 2.基于特征优化和支持向量机的航空发动机气路故障诊断[J].航空发动机,2024,50(4):30-37 |
| 基于特征优化和支持向量机的航空发动机气路故障诊断 |
| Aeroengine Gas-Path Fault Diagnosis Based on Feature Optimization and Support Vector Machine |
| |
| DOI： |
| 中文关键词: 故障诊断 特征优化 支持向量机 主成分分析 深度自编码器 航空发动机 |
| 英文关键词:fault diagnosis feature optimization support vector machine principal component analysis deep autoencoder aeroengine |
| 基金项目:国家级基础加强项目（2019-JCJQ-ZD-339-00）资助 |
|
| 摘要点击次数: 1220 |
| 全文下载次数: 1050 |
| 中文摘要: |
| 针对现有数据驱动的航空发动机故障诊断算法易受飞行监控数据中冗余特征及噪声的干扰，不能及时修正监测数据中
不平衡样本分布对模型泛化性能影响等问题，通过在支持向量机模型中引入特征增维和采用提取算法，提出基于特征优化和支持
向量机的航空发动机气路故障诊断方法，并建立相应模型。将涡桨发动机及CFM56-7B发动机航后数据输入模型，分析与预测实
际故障发生时刻，并将预测结果与真实结果进行比较，同时将其结果与采用随机森林等4种故障诊断方法所得结果进行对比验
证。结果表明：特征优化算法的应用能显著缩短各类故障诊断方法运行时间20%以上；基于特征优化和支持向量机的故障诊断方
法使预测准确率达99.8%；针对非平衡实测数据，特征优化算法和回归预测思想的引入能显著提高算法在不平衡数据集上的性
能，与非回归算法相比故障检测率提高到91.67%。 |
| 英文摘要: |
| Aiming at the problems that existing data-driven aeroengine fault diagnosis algorithms are susceptible to the disruptive
effects of redundant features and noise in flight monitoring data, and unable to timely address the impact of imbalanced sample distribution
in monitoring data on the model's generalization performance, by introducing feature augmentation and using extraction algorithms in sup‐
port vector machine models, an aeroengine gas-path fault diagnosis method was proposed based on feature optimization and the support
vector machine, and the corresponding model was established. A simulation dataset from a turboprop engine and a flight dataset from a
CFM56-7B engine were input into the model to analyze and predict the fault occurrence time. The predicted results were compared with
the actual results, and the latter compared with those obtained by four fault diagnosis methods such as random forest. The results show that
the application of the feature optimization algorithm can significantly shorten the computational time of various fault diagnosing methods by
more than 20%; the fault diagnosis method based on feature optimization and support vector machine achieves a prediction accuracy of
99.8%; for unbalanced measured data, the introduction of feature optimization algorithm and regression prediction can significantly
improve the performance of the algorithm on imbalanced datasets, and the fault detection rate is improved to 91.67% compared with non-
regression algorithms. |
| 查看全文 查看/发表评论 下载PDF阅读器 |