摘要
针对内燃机车用柴油机机油稀释状态难以评估的问题,提出一种基于布谷鸟搜索算法和贝叶斯正则算法的改进高斯过程回归模型的柴油机机油黏度变化趋势预测方法。首先,采用理化检验手段分析机油黏度数据,将其用于模型验证。其次,基于布谷鸟搜索算法和贝叶斯正则算法集成优化高斯过程回归模型并对不同工况下机油黏度变化规律进行拟合,以协方差函数为优化目标提升模型的评估精度。最后,通过HXN3B型内燃机车柴油机机油黏度数据验证所提模型的有效性和实用性。现场测试验证与检修复核表明,所提模型不仅能够预测一定时间内机油黏度变化的趋势,有效评估机油稀释状态,还能在故障早期提出预测性维护建议并动态优化检修排程,提升了内燃机车的安全运用保障能力,进一步验证了该方法的工程应用价值。
Focusing on the difficulties of accessing the dilution state of engine oil,a hybrid algorithm of Cuckoo search(CS)and Bayesian regular(BR)algorithm is proposed to improve the Gaussian process regression(GPR)model for predicting the viscosity variation trend of diesel engine oil.First,bunch of oil viscosity data generated from physicochemical tests are used for modeling,the CS-BR integrated GPR model is regularized to fit the characteristics of the oil viscosity under different operation conditions,and the covariance function is further optimized to improve the evaluation accuracy.Furthermore,the effectiveness of the proposed model is verified by the oil viscosity data of 12 HXN3B locomotive diesel engines undertaking a 6-month field test.Consequently,the proposed model thrives in predicting the trend of oil viscosity change within a certain period of time and shows its capabilities predictive maintenance in terms of early-stage failure prognostics and dynam⁃ic maintenance scheduling.This contributes to the safety operation of a fleet of diesel locomotives in overall.Additionally,the proposed method is validated the engineering application value.
作者
陈艳伶
刘旭鸣
郑福印
李雷
陆航
CHEN Yanling;LIU Xuming;ZHENG Fuyin;LI Lei;LU Hang(Tianjin Railway Technical and Vocational College,Tianjin 300240,China;China Railway Beijing Bureau Group Co.,Ltd.,Tianjin Locomotive Depot,Tianjin 300230,China;Locomotive&Car Research Institute,China Academy of Railway Sciences Co.,Ltd.,Beijing 100081,China)
出处
《大连交通大学学报》
CAS
2024年第5期61-68,90,共9页
Journal of Dalian Jiaotong University
基金
中国国家铁路集团有限公司科技研究开发计划项目(P2023J007)
中国铁道科学研究院集团有限公司科技研究开发计划项目(2021YJ020)。
关键词
内燃机车
机油稀释
高斯过程回归
故障分析
维修策略
diesel locomotives
oil dilution
Gaussian process regression
fault analysis
maintenance strategy
