基于改进t-SNE和RBFNN的柴油机故障诊断  

Fault Diagnosis of Diesel Engine Based on Improved t-SNE and RBFNN

作　　者：尚前明[1] 黄兴烨 沈栋 朱仁杰[1] 胡秋芳 邱天 SHANG Qianming;HUANG Xingye;SHEN Dong;ZHU Renjie;HU Qiufang;QIU Tian(School of Naval Architecture,Ocean and Energy Power Engineering,Wuhan University of Technology,Wuhan 430063,China)

机构地区：[1]武汉理工大学船海与能源动力工程学院,武汉430063

出　　处：《船舶工程》2023年第1期91-97,共7页Ship Engineering

基　　金：工信部高技术船舶科研项目(工信部装函[2017]614号);国家自然科学基金浙江两化融合联合基金重点项目(U1709215)。

摘　　要：针对柴油机故障诊断问题,提出一种基于改进t分布的随机邻域嵌入(t-SNE)和径向基函数神经网络(RBFNN)的柴油机故障诊断方法。针对t-SNE算法对振动信号的实际降维效果不够理想的问题,进行自适应加权优化;引入遗传算法(GA)解决果蝇优化算法(FOA)陷入局部最优的问题,将GA-FOA应用于RBFNN参数选取中;采用改进后的RBFNN模型对经自适应加权t-SNE降维的数据进行故障识别。研究结果表明,改进后的算法能明显改善聚类效果,提高故障识别的正确率,具有良好的应用前景。Aiming at the problem of diesel engine fault diagnosis, a fault diagnosis method based on the combination of improved t-distributed stochastic neighbor embedding(t-SNE) and radial basis function neural network(RBFNN) is proposed. Adaptive weighting optimization is carried out to solve the problem that the effect of t-SNE algorithm is not ideal when reducing the dimension of vibration signal;genetic algorithm(GA)algorithm is introduced to improve the problem of fruit fly optimization algorithm(FOA) falling into local optimization, and GA-FOA algorithm is applied to the selection of RBFNN parameters;The data reduced by adaptive weighted t-SNE are used for fault identification with the improved RBFNN model. The results show that the improved algorithm significantly improves the clustering effect, improves the accuracy of fault identification, and has a good application prospect.

关 键 词：柴油机 振动信号 故障诊断 t分布的随机邻域嵌入(t-SNE) 径向基函数神经网络(RBFNN) 

分 类 号：TK428[动力工程及工程热物理—动力机械及工程]