牵引柴油机是列车运行的关键部件。因此,开发出状态监测系统(CMS)在早期阶段来检测柴油机故障以防止运行中断。CMS监测柴油机振动、用机器学习算法检测异常振动。在CMS中进行倍频带分析,从振动数据中提取特征向量,进行运行试验来验证CMS的性能。试验结果表明用CMS可以从正常振动中明显区分出模拟的异常振动。 Traction diesel engine is a key component of train operation. Therefore, a condition monitoring system (CMS) was developed to detect diesel faults at an early stage to prevent operational disruption. The CMS monitors diesel vibrations and detects abnormal vibrations using machine learning algorithms. Frequency-band analysis was performed in the CMS, eigenvectors were extracted from the vibration data and run tests were performed to verify the performance of the CMS. The experimental results show that the CMS can clearly distinguish the simulated abnormal vibration from the normal vibration.