基于同步提取增强广义S变换的柴油机气门性能退化状态评估

doi:10.12382/bgxb.2023.0363

摘要/Abstract

摘要：

柴油机在运行过程中气门间隙逐渐增大,其状态会随气门性能退化而发生改变。针对传统状态评估方法难以对其气门性能退化状态进行准确评估的问题,提出基于同步提取增强广义S变换(Synchro Extracting Enhanced Generalized S-Transform, SEEGST)的柴油机气门性能退化状态评估方法。通过传感器采集反映柴油机状态的振动信号;为解决传统信号时频分析方法存在时频分辨率低、能量聚集性弱等问题,基于同步提取算法与广义S变换提出SEEGST时频分析方法,将振动信号转换为二维时频图;利用MLP-Mixer模型提取时频图像特征进行训练,实现柴油机状态评估。通过柴油机状态监测实验台开展气门性能退化实验,将所提方法与SSGST-MLPMixer、GST-MLPMixer、SEEGST-ViT、SEEGST-2DCNN、FFT spectrum-1DCNN 5种传统方法对比。实验结果表明:所提方法的整体评估准确率达到98.96%,可有效应用于柴油机气门性能退化状态评估领域,为开展柴油机气门性能退化状态评估提供一种新的思路。

关键词: 柴油机, 状态评估, 同步提取增强广义S变换, MLP-Mixer

Abstract:

The operating status of diesel engine changes with the performance degradation of valve as the valve clearance gradually increases during operation. It is difficult to accurately assessment the performance degradation status of valve by traditional status assessment methods. A diesel engine valve performance degradation status assessment method based on synchroextracting enhanced generalized S-transform (SEEGST) is proposed. The vibration signal reflecting the status of diesel engine is collected by sensors. To solve the problems of low time-frequency resolution and weak energy aggregation in traditional signal time-frequency analysis methods, a SEEGST time-frequency analysis method is proposed based on the synchroextracting algorithm and generalized S-transform to convert the vibration signal into a two-dimensional time-frequency map. MLP-Mixer model is used to extract the time-frequency image features for training,thus realizing the assessment of diesel engine status. The proposed method is compared with five traditional methods, namely SSGST-MLPMixer, GST-MLPMixer, SEEGST-ViT, SEEGST-2DCNN and FFT spectrum-1DCNN, by conducting the valve performance degradation experiment on a diesel engine status monitoring test bench. The experimental results show that the overall assessment accuracy of the proposed method reaches 98.96%, which can be effectively applied to the diesel engine valve performance degradation status assessment and provides a new idea for conducting the diesel engine valve performance degradation status assessment.

Key words: diesel engine, status assessment, synchroextracting enhanced generalized S-transform, MLP-Mixer

中图分类号:

TK428

刘子昌, 白永生, 李思雨, 张坤, 刘敏, 贾希胜. 基于同步提取增强广义S变换的柴油机气门性能退化状态评估[J]. 兵工学报, 2024, 45(6): 2003-2016.

LIU Zichang, BAI Yongsheng, LI Siyu, ZHANG Kun, LIU Min, JIA Xisheng. Assessment of Diesel Engine Valve Performance Degradation Status Based on Synchroextracting Enhanced Generalized S-transform[J]. Acta Armamentarii, 2024, 45(6): 2003-2016.

图/表 20

图1 MLP-Mixer模型结构

Fig.1 MLP-Mixer model structure

图2 柴油机气门性能退化状态评估方法技术路线

Fig.2 Technical route for assessment of degradation status of diesel engine valve performance

图3 柴油机状态监测实验台

Fig.3 Diesel engine status monitoring test bench

图4 传感器安装位置

Fig.4 Sensor installation position

图5 实测信号时域波形与频谱

Fig.5 Time domain waveform and spectrum of measured signal

图6 不同方法得到的实测信号时频谱

Fig.6 Time-frequency spectrum of the measured signal obtained by different methods

表1 实测信号的Renyi熵值

Table 1 Renyi entropy value of measured signal

类别	STFT	CWT	ST	GST	SSGST	SEEGST
Renyi熵	21.16	17.31	21.66	21.31	19.16	12.02

图7 配气机构结构示意图

Fig.7 Schematic diagram of valve structure

图8 气门间隙示意图

Fig.8 Schematic diagram of valve clearance

表2 气门间隙预置情况

Table 2 Valve clearance pre-set condition

状态序号	退化过程	气门间隙/mm	健康状态
1	正常	0.3	健康
2	气门间隙略微增大	0.4	注意
3	气门间隙增大	0.5	异常
4	气门间隙严重增大	0.7	恶化

图9 气门间隙调整过程

Fig.9 Valve clearance adjustment process

图10 气门不同间隙时域波形

Fig.10 Time domain waveforms with different valve clearances

图11 气门不同间隙下信号频谱

Fig.11 Signal spectrum at different valve clearances

图12 不同状态下的时频图

Fig.12 Time-frequency diagrams in different statuses

表3 MLP-Mixer模型结构

Table 3 MLP-Mixer model structure

层	输出
Rearrange-1	[-1, 196, 768]
Linear-2	[-1, 196, 512]
LayerNorm-3	[-1, 196, 512]
Conv1d-4	[-1, 784, 512]
GELU-5	[-1, 784, 512]
Dropout-6	[-1, 784, 512]
Conv1d-7	[-1, 196, 512]
Dropout-8	[-1, 196, 512]
PreNormResidual-9	[-1, 196, 512]
LayerNorm-10	[-1, 196, 512]
Linear-11	[-1, 196, 2048]
GELU-12	[-1, 196, 2048]
Dropout-13	[-1, 196, 2048]
Linear-14	[-1, 196, 512]
Dropout-15	[-1, 196, 512]
PreNormResidual-16	[-1, 196, 512]
︙	︙
LayerNorm-31	[-1, 196, 512]
Reduce-32	[-1, 512]
Linear-33	[-1, 4]

图13 各模型的训练结果对比

Fig.13 Comparison of the training results of each model

表4 各模型的准确率与损失值

Table 4 Accuracy and loss value of each model

模型	准确率/%		损失值
模型	训练集	验证集	训练集	验证集
SEEGST-MLPMixer	100.00	99.75	1.08×10^-4	1.43×10^-2
SSGST-MLPMixer	100.00	97.50	1.49×10^-4	1.00×10^-1
GST-MLPMixer	100.00	97.25	3.60×10^-4	9.27×10^-2
SEEGST-ViT	100.00	96.50	3.24×10^-3	1.32×10^-1
SEEGST-2DCNN	94.73	94.75	2.44×10^-1	2.37×10^-1
FFT spectrum-1DCNN	91.86	92.50	5.37×10^-1	6.33×10^-1

图14 不同评估模型的混淆矩阵

Fig.14 Confusion matrixes for different assessment models

表5 测试集下各模型的精度

Table 5 Accuracy of each model under the test set

模型	精度/%
SEEGST-MLPMixer	98.96
SSGST-MLPMixer	97.40
GST-MLPMixer	95.83
SEEGST-ViT	95.50
SEEGST-2DCNN	93.50
FFT spectrum-1DCNN	89.00

图15 状态评估结果三维立体可视化

Fig.15 Three-dimensional visualization of status assessment results

参考文献 19

[1]	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
[2]	姬亚萌, 张卫正, 原彦鹏, 等. 高强化柴油机活塞加速热疲劳与等效寿命评估方法[J]. 兵工学报, 2022, 43(12): 3008-3019.
	JI Y M, ZHANG W Z, YUAN Y P, et al. Study of equivalent and influence laws of accelerated thermal fatigue life of aluminum alloy pistons in highly strengthened diesel engines[J]. Acta Armamentarii, 2022, 43(12): 3008-3019. (in Chinese)
[3]	毕晓博. 基于振动信号特征的柴油机状态智能识别方法研究[D]. 天津: 天津大学, 2020.
	BI X B. Research on intelligent recognition method for diesel engine states based on vibration signal features[D]. Tianjin: Tianjin University, 2020. (in Chinese)
[4]	LIU Y S, KANG J S, WEN L, et al. Health status assessment of diesel engine valve clearance based on BFA-BOA-VMD adaptive noise reduction and multi-channel information fusion[J]. Sensors, 2022, 22(21): 8129-8153.
[5]	KE Y, HU Y H, SONG E Z, et al. A method for degradation features extraction of diesel engine valve clearance based on modified complete ensemble empirical mode decomposition with adaptive noise and discriminant correlation analysis feature fusion[J]. Journal of Vibration and Control, 2022, 28(19): 2570-2584.
[6]	XU Y G, WANG L, YU G, et al. Generalized S-synchroextracting transform for fault diagnosis in rolling bearing[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 71: 1-14.
[7]	沈虹, 曾锐利, 杨万成, 等. 基于时频图像极坐标增强的柴油机故障诊断[J]. 振动.测试与诊断, 2018, 38(1): 27-33.
	SHEN H, ZENG R L, YANG W C, et al. Diesel engine fault diagnosis based on polar coordinate enhancement of time-frequency diagram[J]. Journal of Vibration, Measurement & Diagnosis, 2018, 38(1): 27-33. (in Chinese)
[8]	牟伟杰, 石林锁, 蔡艳平, 等. 基于振动时频图像全局和局部特征融合的柴油机故障诊断[J]. 振动与冲击, 2018, 37(10): 14-19.
	MOU W J, SHI L S, CAI Y P, et al. Diesel engine fault diagnosis based on the global and local features fusion of time-frequency image[J]. Journal of Vibration and Shock, 2018, 37(10): 14-19. (in Chinese)
[9]	XU Y G, WANG L, HU A J, et al. Time-extracting S-transform algorithm and its application in rolling bearing fault diagnosis[J]. Science China Technological Sciences, 2022, 65(4): 932-942.
[10]	WEI Z N, MAO Y X, YIN Z H, et al. Fault detection based on the generalized Stransform with a variable factor for resonant grounding distribution networks[J]. IEEE Access, 2020, 8: 91351-91367.
[11]	YU G, YU M J, XU C Y. Synchro extracting transform[J]. IEEE Transections on Industrial Electronics, 2017, 64(10): 8042-8054.
[12]	陈育成. 基于瞬时转速的船用中速柴油机健康状态评估方法研究[D]. 武汉: 武汉理工大学, 2020.
	CHEN Y C. Health assessment for marine medium-speed diesel engine based on instantaneous speed[D]. Wuhan: Wuhan University of Technology, 2020. (in Chinese)
[13]	白雲杰, 贾希胜, 梁庆海, 等. 基于深度学习的柴油机气门健康状态评估[J]. 科学技术与工程, 2022, 22(10): 3941-3950.
	BAI Y J, JIA X S, LIANG Q H, et al. Evaluation of diesel engine valve health status based on deep learning[J]. Science Technology and Engineering, 2022, 22(10): 3941-3950. (in Chinese)
[14]	HAN K, WANG Y H, CHEN H T, et al. A survey on vision transformer[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(1): 87-110.
[15]	刘敏, 陈健, 张岩, 等. 基于同步压缩广义S变换的发动机故障诊断[J]. 振动.测试与诊断, 2021, 41(5): 984-990.
	LIU M, CHEN J, ZHANG Y, et al. Engine fault diagnosis based on synchro squeezing generalized S-transform[J]. Journal of Vibration, Measurement & Diagnosis, 2021, 41(5): 984-990. (in Chinese)
[16]	康佳星. 同步提取变换算法的改进研究及其在地震信号分析中的应用[D]. 成都: 成都理工大学, 2018.
	KANG J X. Research on improved synchro extracting transform and its application in seismic signal processing[D]. Chengdu: Chengdu University of Technology, 2018. (in Chinese)
[17]	TOLSTIKHIN I O, HOULSBY N, KOLESNIKOV A, et al. MLP-mixer: an all-MLP architecture for vision[J]. Advances in Neural Information Processing Systems, 2021, 34: 24261-24272.
[18]	征博文. 应用于阿尔兹海默病MRI辅助诊断的人工智能算法研究[D]. 深圳: 中国科学院大学(中国科学院深圳先进技术研究院), 2022.
	ZHENG B W. Research on artificial intelligence algorithm for MRI assisted diagnosis of Alzheimer’s disease[D]. Shenzhen: Shenzhen Institute of Advanced Technology,Chinese Academy of Sciences, 2022. (in Chinese)
[19]	赖岳华. 变工况下柴油机故障在线监测与维修决策优化方法研究与应用[D]. 北京: 北京化工大学, 2018.
	LAI Y H. Research and application of online monitoring and maintenance decision optimization method for diesel engine fault under variable working conditions[D]. Beijing: Beijing University of Chemical Technology, 2018. (in Chinese)