改进变分模态分解在发动机气门间隙异常故障诊断中的应用

基于振动监测信号的故障诊断技术，对于船舶、油气田、核电等关键领域中大型高速柴油机的健康管理和智能运维具有重要意义。针对柴油发动机气门间隙异常故障，提出了基于振动数据驱动的故障诊断方法。首先，提出模态数量和惩罚因子均为自动优选的改进变分模态分解（VMD）方法，克服了传统VMD方法中上述参数需凭经验预设的缺点；进一步，对VMD分量进行多域特征提取，利用核密度估计方法进行特征敏感性的排序和选择；最后，构建全连接网络分类模型，将优选后的故障敏感特征通过分类模型进行故障识别。利用故障模拟实验台验证了不同工况下的气门间隙异常数据，结果表明本文所提的基于改进变分模态分解的气门间隙异常诊断方法故障识别率超过86%，具有良好的应用效果。

Application of improved variational mode decomposition in fault diagnosis of diesel engine valve clearance

Fault diagnosis technology based on vibration monitoring signals is of great significance in the health management and intelligent operation and maintenance of diesel engines in key fields such as shipping, oil and gas fields and nuclear power. This paper presents a fault diagnosis method based on vibration data for abnormal valve clearance faults of diesel engines. Firstly, an improved variational modal decomposition method is proposed to automatically optimize the number of modes and the penalty factor, which overcomes the shortcomings of the traditional variational mode decomposition (VMD) method in which the above parameters need to be predetermined by experience. Furthermore, multi-domain feature extraction was carried out for the VMD components. A kernel density estimation method was used to sort and select feature sensitivity. Finally, a fully connected network classification model was constructed, and the optimized fault sensitive features were identified by the classification model. The valve clearance anomaly data under different working conditions were verified using a fault simulation experimental platform. The accuracy of the valve clearance anomaly diagnosis method based on our improved variational mode decomposition was more than 86%, showing that it can be used in practical application.