基于堆叠稀疏自编码器的多缸喷油器堵塞定位算法

doi:10.12382/bgxb.2023.0764

摘要/Abstract

摘要：

燃油喷射系统的工作质量直接影响柴油机工作过程及性能,针对多缸机不同喷油器发生堵塞故障且故障程度不一时,传统故障诊断方法难以精准定位故障喷油器的问题,提出一种基于堆叠稀疏自编码器(Stacked Sparse Autoencoder, SSAE)的故障定位算法。通过SSAE提取不同喷油器发生堵塞故障时轨压信号的深层特征,以softmax网络实现故障部件定位。以一维轨压信号为输入,故障喷油器定位为输出,并研究算法超参数对算法精度的影响。研究结果表明,此算法能精准定位发生堵塞故障的喷油器,且精度不受堵塞程度的影响,故障诊断正确率可达96.7%。

关键词: 高压共轨, 不同喷油器堵塞, 堆叠稀疏自编码器, 故障定位

Abstract:

The quality of the fuel injection system directly affects the working process and performance of diesel engines. The traditional fault diagnosis methods are difficult to accurately locate the faulty injectors when the blocking faults occur in different injectors of multi-cylinder engines with varying degrees of fault. A fault location algorithm based on stacked sparse autoencoder (SSAE) is proposed. The deep features of rail pressure signals are extracted by SSAE when the fuel injectors at different positions experience the blocking faults, and softmax network is used to locate the faulty injectors. The influence of algorithm hyperparameters on algorithm accuracy is studied by takin one-dimensional rail pressure signal as input and faulty injector position as output. The final results indicate that this algorithm can accurately locate the fuel injector with blocking faults, and the locating accuracy is not affected by the degree of blockage. The fault diagnosis accuracy can reach 96.7%.

Key words: high-pressure common rail, blockage of different fuel injectors, stacked sparse autoencoder, fault location

中图分类号:

TK428

王健, 黄英, 高晓宇, 王拓, 王绪, 惠嘉赫. 基于堆叠稀疏自编码器的多缸喷油器堵塞定位算法[J]. 兵工学报, 2024, 45(10): 3706-3717.

WANG Jian, HUANG Ying, GAO Xiaoyu, WANG Tuo, WANG Xu, HUI Jiahe. Blockage Location Algorithm of Multi-cylinder Fuel Injectors Based on Stacked Sparse Autoencoder[J]. Acta Armamentarii, 2024, 45(10): 3706-3717.

图/表 23

图1 高压共轨系统仿真模型

Fig.1 Simulation model of high pressure common rail system

表1 高压共轨系统参数

Table 1 High pressure common rail system parameters

部件	参数	数值
	喷孔直径/mm	0.12
	喷孔个数	6
	针阀升程/mm	0.25
喷油器	控制阀升程/mm	0.05
	喷嘴压力室容积/mm³	0.2
	进油孔直径/mm	0.28
	出油孔直径/mm	0.31
	共轨管直径/mm	9.5
共轨管	共轨管长度/mm	250
	高压油管内径/mm	2
燃油泵	柱塞直径/mm	7
	柱塞行程/mm	12

图2 不同轨压下喷油器入口压力校核结果

Fig.2 Verificated results of fuel injector inlet pressure under different rail pressures

图3 堆叠稀疏自编码器结构

Fig.3 Structure diagram of SSAE

图4 基于SSAE喷油器堵塞故障定位流程

Fig.4 Flowchart of locating the injector blocking faults based on SSAE

表2 各缸喷油器堵塞故障模拟参数表

Table 2 Simulation parameters for blocking faults of fuel injectors in each cylinder

故障类别	堵塞程度
正常	正常
1缸喷油器堵塞	堵塞25%,堵塞 50%,堵塞75%
2缸喷油器堵塞
3缸喷油器堵塞
4缸喷油器堵塞

图5 150MPa各缸喷油器堵塞对轨压影响

Fig.5 The impact of blocked fuel injectors on rail pressure in each cylinder at 150MPa

图6 SSAE训练误差随迭代次数变换趋势

Fig.6 Change of SSAE training error with iteration times

图7 各缸喷油器堵塞故障特征可视化

Fig.7 Visualization of fault characteristics of fuel injectors blocking in each cylinder

表3 研究网络层数影响的超参数设置

Table 3 Hyperparameter settings for studying the influence of network layers

试验组	n	ρ	epoch
1	2	0.1	400
2	3	0.1	400
3	4	0.1	400
4	5	0.1	400

图8 不同网络层数特征提取结果对比

Fig.8 Comparison of feature extraction results for different network layers

表4 网络层数影响规律

Table 4 The influence of network layers

试验组	n	轮廓系数	训练时间/s
1	2	0.6695	24.53
2	3	0.7165	39.05
3	4	0.7218	50.19
4	5	0.7242	54.74

表5 研究稀疏因子影响的超参数设置

Table 5 Hyperparameter settings for studying the influence of sparse factors

试验组	n	ρ	epoch
1	3	0.1	400
2	3	0.2	400
3	3	0.3	400
4	3	0.4	400

表6 稀疏因子影响规律

Table 6 The influence of sparse factor

试验组	ρ	轮廓系数	训练时间/s
1	0.1	0.7165	39.05
2	0.2	0.9054	38.75
3	0.3	0.5433	39.24
4	0.4	0.4727	39.61

图9 不同稀疏因子特征提取结果对比

Fig.9 Comparison of feature extraction results for different sparse factors

表7 研究迭代次数影响的超参数设置

Table 7 Hyperparameter settings for studying the influence ofiteration times

试验组	n	ρ	epoch
1	3	0.2	50
2	3	0.2	100
3	3	0.2	200
4	3	0.2	400

表8 迭代次数影响规律

Table 8 The influence of iteration times

试验组	epoch	轮廓系数	训练时间/s
1	50	0.8089	5.42
2	100	0.8400	9.12
3	200	0.9031	12.51
4	400	0.9054	38.75

图10 不同迭代次数特征提取结果对比

Fig.10 Comparison of feature extraction results for different iteration times

表9 SSAE网络超参数设置

Table 9 Hyperparameter setting of SSAE network

N	ρ	epoch
3	0.2	200

图11 SSAE模型训练集和测试集诊断结果

Fig.11 Diagnosis results of SSAE model training and testing sets

图12 轨压90MPa下SSAE模型训练集和测试集诊断结果

Fig.12 Diagnosis results of SSAE model training and testing sets at 90MPa

图13 轨压120MPa下SSAE模型训练集和测试集诊断结果

Fig.13 Diagnosis results of SSAE model training and esting sets at 120MPa

表10 不同算法诊断结果对比

Table 10 Comparison of diagnostic results of different algorithms

算法	训练精度/%	训练时间/s	诊断用时/s	测试精度/%
AE	76.4	13.04	0.73	80.8
SAE	93.1	4.94	0.7	86.7
SSAE	100	14.8	0.93	96.7

参考文献 25

[1]	孙宜权, 王滨, 张英堂, 等. 基于自适应平行因子的柴油机喷油故障诊断研究[J]. 兵工学报, 2013, 34(5): 519-526. doi: 10. 3969/ j. issn. 1000-1093. 2013. 05. 002
	SUN Y Q, WANG B, ZHANG Y T, et al. Study of fault diagnosis of diesel engine fuel injection based on adaptive parallel factor[J]. Acta Armamentarii, 2013, 34(5): 519-526. (in Chinese)
[2]	AGARWAL A K, SINGH A P, MAURYA R K, et al. Combustion characteristics of a common rail direct injection engine using different fuel injection strategies[J]. International Journal of Thermal Sciences, 2018, 134: 475-484.
[3]	杨强, 杨建国. 船用中速柴油机高压共轨系统的现状与发展趋势[J]. 船海工程, 2019, 48(3): 142-146.
	YANG Q, YANG J G. Present situation and development trend of high pressure common rail system of marine medium-speed diesel engine[J]. Ship & Ocean Engineering, 2019, 48(3): 142-146. (in Chinese)
[4]	孙运涛. 共轨泵用燃油计量阀复位弹簧参数对流量的影响[J]. 现代车用动力, 2014, 153(1): 27-29.
	SUN Y T. Impact of return spring parameters on flow for common rail pump fuel metering valve[J]. Morden Vehicle Power, 2014, 153(1): 27-29. (in Chinese)
[5]	XI W K, LI Z X, ZHE T, et al. A feature extraction and visualization method for fault detection of marine diesel engines[J]. Measurement, 2018, 116(1): 429-437.
[6]	JAFARIAN K, DARJANI M, HONARKAR Z. Vibration analysis for fault detection of automobile engine using PCA technique[C]// Proceedings of the 4th international conference on control, instrumentation, and automation. Qazvin, Iran: IEEE, 2016: 372-376.
[7]	FLETT J, BONE G. Fault detection and diagnosis of diesel engine valve trains[J]. Mechanical Systems & Signal Processing, 2016, 72: 316-327.
[8]	ZHANG B, YAN J, TIAN C. Study on fault diagnosis system of diesel engine fuel injection system based on BP neural network[C]// Proceedings of 2008 ISECS International Colloquium on Computing, Communication, Control & Management. Guangzhou, China: IEEE, 2008: 108-112.
[9]	刘建敏, 史玉鹏, 许世永, 等. 基于缸盖振动的燃油喷射系统故障诊断[J]. 装甲兵工程学院学报, 2011, 25(3): 29-34.
	LIU J M, SHI Y P, XU S Y, et al. Fuel injection system fault diagnosis based on cylinder head vibration signal[J]. Journal of Academy of Armored Force Engineering, 2011, 25(3): 29-34. (in Chinese)
[10]	AYATI M, SHIRAZI F A, ANSARI R S, et al. Classification-based fuel injection fault detection of a trainset diesel engine using vibration signature analysis[J]. Journal of Dynamic Systems Measurement and Control, 2020, 142(5): 051003.
[11]	胡志勇, 牛家骅, 郭丽娜, 等. 基于时域能量划分和PSO-SVM的发动机故障诊断[J]. 汽车工程, 2016, 38(1): 86-90.
	HU Z Y, NIU J Y, GUO L N, et al. Engine fault diagnosis based on time-domain energy division and PSO-SVM algorithm[J]. Automotive Engineering, 2016, 38(1): 86-90. (in Chinese)
[12]	柯赟, 宋恩哲, 姚崇, 等. 层次离散熵及其在高压共轨喷油器故障诊断中的应用[J]. 振动与冲击, 2021, 40(2): 72-80.
	KE Y, SONG E Z, YAO C, et al. Hierarchical dispersion entropy and its application in fault diagnosis of high pressure common rail injectors[J]. Journal of vibration and shock, 2021, 40(2): 72-80. (in Chinese)
[13]	李良钰, 苏铁熊, 马富康, 等. 基于集合经验模态分解-支持向量机的高压共轨系统故障诊断方法[J]. 兵工学报, 2022, 43(5): 992-1001. doi: 10.12382/bgxb.2021.0155
	LI L Y, SU T X, MA F K, et al. Fault diagnosis method of high-pressure common rail system based on EEMD-SVM[J]. Acta Armamentarii, 2022, 43(5): 992-1001. (in Chinese) doi: 10.12382/bgxb.2021.0155
[14]	靳莹, 乔新勇, 顾程, 等. 基于Res-CNN和燃油压力波的柴油机喷油器故障诊断方法[J]. 汽车工程, 2021, 43(6): 943-951.
	JIN Y, QIAO X Y, GU C, et al. A method for fault diagnosis of fuel injector of diesel engine based on Res-CNN and fuel pressure wave[J]. Automotive Engineering, 43(6): 943-951. (in Chinese)
[15]	靳莹, 乔新勇. 基于多层双向长短时记忆网络的装甲车辆柴油机喷油器故障诊断[J]. 兵工自动化, 2022, 41(1): 34-38.
	JIN Y, QIAO X Y. Injector fault diagnosis of armored vehicle diesel engine based on multi-layer bidirectional long short term memory network[J]. Ordnance Industry Automation, 2022, 41(1): 34-38. (in Chinese)
[16]	秦慈伟, 金江善, 董晓露, 等. 基于压力信号的船用柴油机共轨系统高压油泵典型故障诊断研究[J]. 柴油机, 2019, 41(3): 16-21.
	QIN C W, JIN J S, DONG X L, et al. Fault diagnosis for high-pressure pump of marine diesel engine with common rail system based on pressure signal[J]. Diesel Engine, 2019, 41(3): 16-21. (in Chinese)
[17]	刘原宾. 柴油机燃油系统的故障诊断与健康管理[D]. 大连: 大连理工大学, 2021.
	LIU Y B. Fault diagnosis and health management of diesel fuel system[D]. Dalian: Dalian University of Technology, 2021. (in Chinese)
[18]	刘国强, 林叶锦, 张志政, 等. 基于粗糙集和优化DAG-SVM的船舶主机故障诊断方法[J]. 中国舰船研究, 2020, 15(1): 68-73.
	LIU G Q, LIN Y J, ZHANG Z Z, et al. Main marine engine fault diagnosis method based on rough set theory and optimized DAG-SVM[J]. Chinese Journal of Ship Research, 2020, 15(1): 68-73. (in Chinese)
[19]	WANG Y L, YANG H B, YUAN X F, et al. Deep learning for fault-relevant feather extraction and fault classification with stacked supervised auto-encoder[J]. Journal of Process Control, 2020, 92(8): 79-89.
[20]	白雲杰, 贾希胜, 梁庆海. 基于DVMD和SSAE的柴油机混合故障诊断[J]. 振动与冲击, 2022, 41(11): 271-277.
	BAI Y J, JIA X S, LIANG Q H. Hybrid fault diagnosis of Diesel engine based on DVMD and SSAE[J]. Journal of vibration and shock, 2022, 41(11): 271-277. (in Chinese)
[21]	王树宇, 袁嫣红, 张建义. 基于对抗自编码模型的高速泵异常检测[J]. 机床与液压, 2022, 50(7): 176-180. doi: 10.3969/j.issn.1001-3881.2022.07.032
	WANG S Y, YUAN Y H, ZHANG J Y. Anomaly detection of high speed pump based on adversarial auto-encoder model[J]. Machine Tool & Hydraulics, 2022, 50(7): 176-180. (in Chinese)
[22]	郑涛. 高压共轨系统喷油器故障诊断与健康评估研究[D]. 哈尔滨: 哈尔滨工程大学, 2021.
	ZHENG T. Research on fault diagnosis and health assessment of injector in common rail system[D]. Harbin: Harbin Engineering University, 2021. (in Chinese)
[23]	OU J Y, LI H K, HUANG G J, et al. A novel order analysis and stacked sparse auto-encoder feature learning method for milling tool wear condition monitoring[J]. Sensors, 2020, 20(10): 2878.
[24]	LI Z C, TIAN L, JIANG Q C, et al. Distributed-ensemble stacked autoencoder model for non-linear process monitoring[J]. Information Sciences, 2021, 542: 302-316.
[25]	徐活耀, 陈里里. 基于堆栈稀疏自编码器的滚动轴承故障诊断[J]. 机床与液压, 2020, 48(14): 190-194. doi: 10.3969/j.issn.1001-3881.2020.14.041
	XU H Y, CHEN L L. Fault diagnosis of rolling bearing based on stacked sparse autoencoder[J]. Machine Tool & Hydraulic, 2020, 48(14): 190-194. (in Chinese)