| 董飞,杨建国,范玉,等.基于迁移学习的船用低速机燃油喷射系统故障诊断方法研究[J].内燃机工程,2023,44(1):17-26. |
| 基于迁移学习的船用低速机燃油喷射系统故障诊断方法研究 |
| Research on Fault Diagnosis Method of Injection System of Marine Low-Speed Diesel Engine Based on Transfer Learning |
| DOI：10.13949/j.cnki.nrjgc.2023.01.003 |
| 关键词:低速机 燃油喷射系统 故障仿真 故障诊断 迁移学习 |
| Key Words:low speed engine fuel injection system fault simulation fault diagnosis transfer learning |
| 基金项目:智能中速柴油机关键技术研究项目（工信部装函［2019］360号） |
|
| 摘要点击次数: 1187 |
| 全文下载次数: 766 |
| 摘要:针对低速机故障模拟试验成本高和故障样本数据获取难度大的问题，通过建立船用低速机喷油系统模型和一维工作过程模型，采用仿真的方法获取了喷油器喷孔磨损、喷油量减少及喷油正时变化等故障样本，分析了低速机燃油喷射系统故障时的参数变化规律。基于故障参数的变化规律提取了7个故障特征参数，用主元分析法检验了所提取的特征参数对故障诊断的有效性和故障可分类性。针对传统的机器学习故障诊断算法要求数据独立同分布等问题，提出了基于TrAdaBoost迁移学习算法的低速机燃油喷射系统诊断模型，通过3 560个仿真故障样本的验证结果表明其诊断准确率在85%以上，诊断模型可在低速机不同负荷之间实现诊断知识的迁移。 |
| Abstract:To solve the problem that the low speed diesel engine fault simulation tests cost is high and its sample data are hard to get， a marine low speed diesel engine injection system model and a one-dimensional diesel engine model were established. Based on simulation， the samples of the injector nozzle wear， the volume of injection reducing and the injection timing fault were simulated and analyzed. Seven fault characteristic parameters were selected based on the percentage change of fault parameters. The effectiveness of the characteristic parameters selected for the fault diagnosis and the fault classification was verified by the principal component analysis. Considering that traditional machine learning fault diagnosis algorithm requires the data is independently and identically distributed， a diagnosis model of the marine low speed diesel engine fuel injection system based on TrAdaBoost transfer learning algorithm was proposed. The 3 560 simulation fault samples and the verification result indicated that the accuracy of diagnosis was above 85%. The results show that the model can transfer diagnosis knowledge between different loads of low speed diesel engine. |
|
查看全文 HTML
查看/发表评论 下载PDF阅读器 |