基于CAWOA-BP的船舶凝给水系统故障诊断

Fault diagnosis of ship condensate feed water system based on CAWOA-BP

为克服传统专家经验在故障诊断方面的不足,实现船舶凝给水系统的智能诊断,在标准BP神经网络基础上提出一种优化后的CAWOA-BP故障诊断模型。采用混沌映射以及自适应权重调整策略优化WOA鲸鱼算法,利用优化后的WOA鲸鱼算法改进BP神经网络的权值及阈值矩阵。由于船舶凝给水系统的状态监测数据是复杂多维度数据,利用UMAP降维算法对原始数据进行降维。最后,利用降维处理后的数据训练CAWOA-BP神经网络模型,实现故障诊断。通过对正常及故障数据的学习,发现优化后的CAWOA-BP模型相比于标准BP,WOA-BP,PSO-BP故障诊断模型具有更高的准确率、精确率、召回率及预测误差。研究表明,基于优化后的CAWOA-BP神经网络故障诊断方法能够更加精确实现船舶凝给水系统的故障诊断。

In order to overcome the shortcomings of traditional expert experience in fault diagnosis and realize the intelligent diagnosis of ship condensate feed water system.Based on the standard BP neural network,an optimized CAWOA-BP fault diagnosis model is proposed.Firstly,the WOA whale algorithm was optimized by using chaotic mapping and adaptive weight adjustment strategy.Then,the optimized WOA whale algorithm is used to improve the weight and threshold matrix of BP neural network;Secondly,as the condition monitoring data of the ship condensate feed water system is complex and multi-dimensional data,the UMAP dimension reduction algorithm is used to reduce the dimension of the original data.Finally,the CAWOA-BP neural network model is trained with the reduced dimension data to achieve fault diagnosis.Through learning the normal and fault data,it was found that the optimized CAWOA-BP model had higher accuracy,precision,recall and prediction error than the standard BP,WOA-BP and PSO-BP fault diagnosis models.The research shows that the fault diagnosis method based on the optimized CAWOA-BP neural network can more accurately realize the fault diagnosis of the ship condensate feed water system.