基于深度学习的柴油机气门健康状态评估  

Evaluation of Diesel Engine Valve Health Status Based on Deep Learning

作　　者：白雲杰 贾希胜 梁庆海 白永生 BAI Yun-jie;JIA Xi-sheng;LIANG Qing-hai;BAI Yong-sheng(Department of Equipment Command and Management, Shijiazhuang Campus of Army Engineering University, Shijiazhuang 050003, China;66029 Unit of the Chinese People's Liberation Army, Xilinguole 011200, China;Hebei Key Laboratory of Condition Monitoring and Assessment of Mechanical Equipment, Shijiazhuang 050003, China)

机构地区：[1]陆军工程大学石家庄校区装备指挥与管理系,石家庄050003 [2]中国人民解放军66029部队,锡林郭勒011200 [3]河北省机械装备状态监测与评估重点实验室,石家庄050003

出　　处：《科学技术与工程》2022年第10期3941-3950,共10页Science Technology and Engineering

基　　金：军内科研项目(LJ20191C050518)。

摘　　要：柴油发动机在运行过程中,其气门间隙会随其性能状态退化发生改变,为了解决传统的健康状态评估方法健康指标确定困难、权重人为经验依赖性大的问题,提出一种基于深度学习的柴油机气门健康状态评估方法。首先通过小波包分解算法对柴油机缸盖振动信号进行分解,对分解得到的节点信号分别提取常见的14个时域特征和小波包分解信号能量比向量,构建多维综合健康评估指标向量。然后基于一维卷积神经网络(one-dimensional convolutional neural network,1DCNN)构建健康状态评估模型,将得到的健康评估指标向量输入模型中进行训练与健康状态评估。通过柴油机实验台开展气门退化模拟实验验证了该方法的有效性,与传统方法相比解决了健康指标的筛选问题以及人为主观经验带来的影响,并具有更好的健康状态评估效果。During the operation of a diesel engine,its valve clearance will change with the degradation of its performance state.In order to solve the problems of the traditional health assessment method,the determination of health indicators is difficult and the weight is artificially dependent on human experience.A method for evaluating the health status of diesel engine valves based on deep learning was proposed.Firstly,the diesel engine cylinder head vibration signal was decomposed by wavelet packet decomposition algorithm,and 14 common time-domain features and wavelet packet decomposition signal energy ratio vector were extracted from the decomposed node signals,and a multi-dimensional comprehensive health assessment index vector was constructed.Then build a health evaluation model based on one-dimensional convolutional neural network(1DCNN),and input the obtained health evaluation index vector into the model for training and health evaluation.The effectiveness of the method is verified through the valve degradation simulation experiment carried out on the diesel engine test bench.In addition,compared with the traditional method,it solves the problem of screening health indicators and the influence of subjective experience of people,and has a better health evaluation effect.

关 键 词：一维卷积神经网络 小波包分解 柴油发动机 气门间隙 健康状态评估 

分 类 号：TH212[机械工程—机械制造及自动化]