基于t-SNE-VNWOA的船舶柴油机故障诊断

Fault Diagnosis of Marine Diesel Engine Based on t-SNE-VNWOA

文中提出一种基于t-SNE-VNWOA-LSSVM故障诊断模型,并进行了台架试验.试验设置了正常工况、供气不足、燃烧提前和单缸断油四种工况,将各种工况采集的缸盖振动信号进行快速傅里叶变换(FFT),提取了13个时域和频域特征,利用t分布邻域嵌入算法(t-SNE)对数据降维、可视化故障特征.结合鲸鱼优化算法(VNWOA)对分类器(LSSVM)初始参数δ2和γ寻优,搭建其故障识别模型,将遗传算法(GA)和粒子群算法(PSO)的寻优诊断结果与之对比.结果表明:基于t-SNE-VNWOA-LSSVM故障诊断模型精度高达96.57%,且具有良好的稳定性及诊断速度.

A fault diagnosis model based on t-SNE-VNWOA-LSSVM was proposed,and the bench test was carried out.The test set up four working conditions:normal working condition,insufficient air supply,early combustion and single cylinder oil cut-off.The cylinder head vibration signals collected under various working conditions were processed by fast Fourier transform(FFT),and 13 time-domain and frequency-domain features were extracted.t-SNE was used to reduce the dimension of the data and visualize the fault features.The initial parametersδ2 andγof the classifier(LSSVM)were optimized by the whale optimization algorithm(VNWOA),and its fault identification model was established.The diagnosis results of genetic algorithm(GA)and particle swarm optimization(PSO)were compared with them.The results show that the accuracy of fault diagnosis model based on t-SNE-VNWOA-LSSVM is as high as 96.57%,and it has good stability and diagnosis speed.