柴油机故障的集成诊断方法研究
发布时间:2018-01-15 06:09
本文关键词:柴油机故障的集成诊断方法研究 出处:《西安石油大学》2015年硕士论文 论文类型:学位论文
更多相关文章: 集成学习 支持向量机 遗传算法 柴油机 故障诊断
【摘要】:柴油机作为一种动力设备,在车辆、船舶、电站、工程机械和农用机械等领域都有着广泛的应用。柴油机具有零部件多、运动复杂、工作环境恶劣等特点,因此其出现故障的概率相对较高,故障诊断的研究在保证整个设备正常运行中具有非常重要的意义。随着智能化技术的发展,故障诊断方法也日益完善。论文在学习和总结现有方法和技术的基础上,基于集成学习理论,对柴油机故障诊断方法进行深入探讨。以支持向量机为子学习器的集成学习理论应用于柴油机故障模式识别中,并综合采用BP神经网络、遗传算法和单个支持向量机就柴油机故障诊断中运行状态的分类识别这一关键环节进行系统的对比分析研究。针对柴油机正常状态以及出油阀磨损、供油多等六种故障状态下的振动信号提取了时域信号特征和子带能量特征,结合不同方法在故障模式识别上的优势,论文在重点研究单个支持向量机和支持向量机集成学习算法故障诊断技术的基础上,研究了BP神经网络和遗传算法在柴油机故障诊断中的应用。建立了支持向量机故障诊断模型,并利用交叉验证方法得出最优参数;基于Boosting算法建立支持向量机集成故障诊断模型;同时分析研究了BP神经网络和遗传算法优化BP神经网络故障诊断模型,并与基于集成学习构建的强学习器诊断模型对比分析。研究结果表明,基于集成学习的支持向量机模型诊断结果正确率高于单个支持向量机方法和支持向量机参数优化以后故障诊断结果正确率,且明显高于BP神经网络和遗传优化BP神经网络在柴油机故障诊断中的应用结果。成功验证了基于集成学习构建强可学习的构想,即集成学习应用于柴油机故障诊断的可行性。
[Abstract]:As a kind of power equipment, diesel engine has been widely used in the fields of vehicle, ship, power station, construction machinery and agricultural machinery. Therefore, the probability of failure is relatively high, the study of fault diagnosis is very important to ensure the normal operation of the whole equipment. With the development of intelligent technology. The method of fault diagnosis is becoming more and more perfect. This paper studies and summarizes the existing methods and techniques, and based on the integrated learning theory. The integrated learning theory based on support vector machine (SVM) is applied to diesel engine fault pattern recognition, and BP neural network is used comprehensively. Genetic algorithm (GA) and single support vector machine (SVM) are used to compare and analyze the key link of diesel engine fault diagnosis in classification and recognition, aiming at the normal state of diesel engine and the wear of oil delivery valve. The vibration signals in six kinds of fault states, such as oil supply, extracted the time domain signal characteristics and sub-band energy characteristics, and combined with the advantages of different methods in fault pattern recognition. This paper focuses on the single support vector machine and support vector machine integrated learning algorithm fault diagnosis technology. The application of BP neural network and genetic algorithm in diesel engine fault diagnosis is studied. The support vector machine integrated fault diagnosis model is established based on Boosting algorithm. At the same time, BP neural network and genetic algorithm optimized BP neural network fault diagnosis model are analyzed and compared with the strong learning model based on integrated learning. The diagnostic accuracy of SVM model based on ensemble learning is higher than that of single SVM method and SVM parameters optimization. The application results of BP neural network and genetic optimization BP neural network in diesel engine fault diagnosis are obviously higher than that of BP neural network. That is, integrated learning is feasible for diesel engine fault diagnosis.
【学位授予单位】:西安石油大学
【学位级别】:硕士
【学位授予年份】:2015
【分类号】:TK428;TP18
【参考文献】
相关期刊论文 前5条
1 曹龙汉,曹长修,孙颖楷,景有泉,郭振;柴油机故障诊断技术的现状及展望[J];重庆大学学报(自然科学版);2001年06期
2 李业波;李秋红;黄向华;赵永平;;航空发动机传感器故障与部件故障诊断技术[J];北京航空航天大学学报;2013年09期
3 吕锋;李翔;杜文霞;;基于MultiBoost的集成支持向量机分类方法及其应用[J];控制与决策;2015年01期
4 刘世元,杜润生,杨叔子;柴油机缸盖振动信号的小波包分解与诊断方法研究[J];振动工程学报;2000年04期
5 赵宏,夏哲雷;优化遗传神经网络及其在机械故障诊断中的应用[J];中国计量学院学报;2004年02期
相关博士学位论文 前1条
1 李敏通;柴油机振动信号特征提取与故障诊断方法研究[D];西北农林科技大学;2012年
相关硕士学位论文 前1条
1 张北鸥;基于小波包变换和Elman人工神经网络的电机故障诊断系统的研究[D];武汉理工大学;2010年
,本文编号:1427100
本文链接:https://www.wllwen.com/kejilunwen/dongligc/1427100.html