基于PSO-ELM算法实现船舶发电机组故障识别

针对船舶发电机组的不同故障类型,通过传感器采集不同故障下柴油机缸盖处的振动信号,构成大量数据集,选取部分数据集作为样本数据。通过EEMD算法对样本数据进行分解降噪,把一维数据分解成能反映柴油机工况信息的多维数据,对分解形成的多维数据使用KICA算法进行特征提取,并对提取后的数据进行训练集、验证集分组。使用PSO-ELM算法搭建故障识别模型,并使用训练集训练模型,使用验证集验证模型,根据验证结果评价模型是否满足故障识别的精确度。

Fault Identification Of Marine Generator Set based on PSO-ELM Algorithm

The abnormality of the rotating machinery during the operation will cause the transient shock signal of the vibration signal, causing the amplitude of the vibration signal to change. Different fault types result in transient shock signals with different strengths. The effective detection and extraction of vibration signals under different working conditions is the key to fault identification. In this paper, according to the different types of faults of the marine generator set, the vibration signals at the cylinder head of the diesel engine under different faults are collected by sensors to form a large number of data sets, and some data sets are selected as sample data. First, the EEMD algorithm is used to decompose the sample data to reduce noise, and the one-dimensional data is decomposed into multi-dimensional data that can reflect the working condition information of the oil recovery machine. Then the KICA algorithm is used to extract the multi-dimensional data formed by decomposition, because feature extraction is an important basis for fault recognition and directly affects the accuracy of recognition. Finally, a fault recognition model is built based on the PSO-ELM algorithm, and the processed data is used to verify the model to achieve accurate fault recognition.