基于模型群预测法对汽油机稳态原排的预测

doi:10.13229/j.cnki.jdxbgxb20200603

摘要/Abstract

摘要：

针对通过数值模拟减少汽油机台架试验环节的工作量、提高试验效率的问题，采用模型群预测法（优化后的人工神经网络方法）对汽油机台架试验过程中的NOx、CO、HC等稳态原排进行建模及预测分析。结果表明：与传统的单个模型预测方法相比较，模型群预测法具有较高的可靠性，能较好地提升预测结果的准确度。采用隔点取点法适当减少神经网络建模的训练数据集，仍能保持较好的预测能力，在项目开发过程中只需进行30%的测试量，将试验结果用于神经网络模型训练，可较好地预测剩余工况排放。通过对其他机型的验证分析，模型群预测法在内燃机稳态原排的预测过程中具有较好的普适性。

关键词: 汽油, 神经网络, 预测, 发动机排放, 台架试验

Abstract:

This paper aims to reduce the workload of gasoline engine bench test and improve the experimental efficiency by numerical simulation technology. The model group prediction method （optimized artificial neural network method） was used to model and predict the steady-state emissions of NOx， CO and HC in the process of gasoline engine bench test. The results showed that the model group prediction method has high reliability and may improve the accuracy of prediction results compared with the traditional single model prediction method. The training data set of neural network modeling may be appropriately reduced by using the method of separating points， and the better prediction ability may still be maintained. In the process of project development， only 30% of the test quantities were needed， and the test results may be used for the training of neural network model， which may better predict the emission of the remaining working conditions. Through the validation analysis of other models， the model group prediction method had a good universality in the prediction process of engine steady-state exhaust.

Key words: gasoline, neural networks, prediction, engine exhausts, bench test

中图分类号:

TK411

陈涛,秦静,赵华,苏庆鹏,吕永,钟凯,王膺博,裴毅强. 基于模型群预测法对汽油机稳态原排的预测[J]. 吉林大学学报(工学版), 2021, 51(5): 1565-1574.

Tao CHEN,Jing QIN,Hua ZHAO,Qing-peng SU,Yong LYU,Kai ZHONG,Ying-bo WANG,Yi-qiang PEI. Prediction of gasoline engine steady state exhaust based on model group prediction method[J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(5): 1565-1574.

图/表 14

表1

表2

图1

图2

图3

表3

表4

表5

表6

图4

表7

图5

表8

表9

模型群预测法在各机型排放预测的普适性验证结果"

项目	NOx	CO	HC
汽油机A建模 $R t r a i n 2$	0.9~0.927 4	0.9~0.995 5	0.9~0.967 5
汽油机A预测 $R p r e d i c t 2$	0.827 89	0.989 34	0.868 72
该预测结果优于单个模型预测结果的概率（汽油机A）/%	95.70	95.70	95.20
汽油机B建模 $R t r a i n 2$	0.8~0.851 4	0.8~0.898 6	0.75~0.792 1
汽油机B预测 $R p r e d i c t 2$	0.806 62	0.861 88	0.534 50
该预测结果优于单个模型预测结果的概率（汽油机B）/%	99.65	97.10	75.45
汽油机C建模 $R t r a i n 2$	0.9~0.993 7	0.9~0.990 1	0.9~0.984 5
汽油机C预测 $R p r e d i c t 2$	0.842 11	0.635 77	0.893 86
该预测结果优于单个模型预测结果的概率（汽油机C）/%	79.10	100.00	98.95
柴油机A建模 $R t r a i n 2$	0.9~0.999 9	0.9~1.0	0.9~1.0
柴油机A预测 $R p r e d i c t 2$	0.972 61	0.955 17	0.992 18
该预测结果优于单个模型预测结果的概率（柴油机A）/%	98.65	90.70	99.68
柴油机B建模 $R t r a i n 2$	0.9~0.991 3	—	0.9~0.996 9
柴油机B预测 $R p r e d i c t 2$	0.969 19	—	0.945 68
该预测结果优于单个模型预测结果的概率（柴油机B）/%	83.15	—	86.75

表9

参考文献 23

1	裴毅强, 张建业, 秦静, 等. 增压直喷汽油机起动怠速及混合气浓度对微粒排放的影响[J]. 天津大学学报: 自然科学与工程技术版, 2014, 47(10): 892-897.
	Pei Yi-qiang, Zhang Jian-ye, Qin Jing, et al. Effect of starting idling condition and mixture concentration of a turbocharged GDI engine on particle emission [J]. Journal of Tianjin University(Science and Technology), 2014, 47(10): 892-897.
2	谢宗法, 付文超, 魏枫展, 等. 进气门早关对汽油机进气过程及燃烧性能的影响[J]. 吉林大学学报: 工学版, 2020, 50(3): 850-858.
	Xie Zong-fa, Fu Wen-chao, Wei Feng-zhan, et al. Effect of early intake valve closing on intake process and combustion performance of SI engine [J]. Journal of Jilin University(Engineering and Technology Edition), 2020, 50(3): 850-858.
3	于秀敏, 商震, 张岳韬, 等. 废气再循环对直喷汽油机燃烧及排放影响的仿真[J]. 吉林大学学报: 工学版, 2016, 46(4): 1109-1117.
	Yu Xiu-min, Shang Zhen, Zhang Yue-tao, et al. Simulation of EGR on combustion and emission of gasline direct-injection engine[J]. Journal of Jilin University(Engineering and Technology Edition), 2016, 46(4): 1109-1117.
4	孙平, 曹智, 于秀敏, 等. 直喷时刻对复合喷射汽油机暖机过程燃烧和排放的影响[J]. 吉林大学学报: 工学版, 2020, 50(6): 1941-1949.
	Sun Ping, Cao Zhi, Yu Xiu-min, et al. Experimental study on effect of compound injection on cold start combustion and emission of gasoline engine[J]. Journal of Jilin University(Engineering and Technology Edition), 2020, 50(6): 1941-1949.
5	卫海桥, 裴自刚, 冯登全, 等. 压电喷油器多次喷射对GDI汽油机颗粒物排放的影响[J]. 吉林大学学报: 工学版, 2018, 48(1): 166-173.
	Wei Hai-qiao, Pei Zi-gang, Feng Deng-quan,et al. Effect of multi-injection piezo injector on particulate emission in gasoline direct injection engine[J]. Journal of Jilin University(Engineering and Technology Edition), 2018, 48(1): 166-173.
6	Heinz-Jakob N, Jörn K, Richard D, et al. Die euro-6-motoren des modularen dieselbaukastens von volkswagen[J]. MTZ-Motortechnische Zeitschrift, 2013, 74(6): 440-447.
7	Raidt B. Local gaussian process regression in order to model air charge of turbocharged gasoline SI engines[C]∥SAE Technical Paper, 2016-01-0624.
8	Gutjahr T, Kleinegraeber H, Huber T, et al. Commercial vehicle engineering congress[J]. SAE Technical Paper, 2015-01-2796.
9	Lasko T A. Efficient inference of gaussian process modulated renewal processes with application to medical event data[J]. Uncertain Artif Intell, 2014(2014): 469-476.
10	Bernhard S. Introduction to Gaussian Processes[M]. Massachusetts: MIT Press, 2008.
11	Mackay D J C. Bayesian interpolation[J]. Neural Computation, 1992, 4(3): 415-447.
12	Kuzin D, Yang L, Isupova O, et al. Ensemble kalman filtering for online gaussian process regression and learning[C]∥21st International Conference on Information Fusion, Cambridge, UK, 2018: 39-46.
13	Kauermann G, Bove D, Sabanes H L. Objective bayesian model selection in generalized additive models with penalized splines[J]. Journal of Computational and Graphical Statistics, 2015, 24(2): 394-415.
14	Baek J, Adams A, Dolson J. Lattice-based high-dimensional gaussian filtering and the permutohedral lattice[J]. Journal of Mathematical Imaging & Vision, 2013, 46(2): 211-237.
15	Rot I, Daniel F P, Rinderknecht S. Investigation of black box modeling approaches for representation of transient gearshift processes in automotive powertrains with automatic transmission[C]∥SAE Technical Papers, 2015-01-1143.
16	Mao G, Zhang C, Shi K, et al. Prediction of the performance and exhaust emissions of ethanol-diesel engine using different neural network[J]. Energy Sources Part A Recovery Utilization and Environmental Effects, 2019, 104: 1656307.
17	Nestor L S Y. Modelling the infiltration process with a multi-layer perceptron artificial neural network[J]. Hydrological Sciences Journal, 2006, 51(1): 3-20.
18	李健. 基于深度学习的变循环发动机气路故障诊断[D]. 上海: 上海交通大学电子信息与电气工程学院, 2019.
	Li Jian. Fault diagnosis for variable cycle engine gas path: a deep learning approach[D]. Shanghai: School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, 2019.
19	肖勇, 赵云, 涂治东, 等. 基于改进的皮尔逊相关系数的低压配电网拓扑结构校验方法[J]. 电力系统保护与控制, 2019, 47(11): 37-43.
	Xiao Yong, Zhao Yun, Tu Zhi-dong, et al. Topology checking method for low voltage distribution network based on improved pearson correlation coefficient[J]. Power System Protection and Control, 2019, 47(11): 37-43.
20	龚麒鉴, 郭亚宾, 陈焕新, 等. 基于粒子群优化算法和BP神经网络的变频压缩机功率预测[J]. 制冷学报, 2020, 41(1): 89-95.
	Gong Qi-jian, Guo Ya-bin, Chen Huan-xin, et al. Prediction of variable-speed compressor power based on particle swarm optimization and back propagation neural network[J]. Journal of Refrigeration, 2020, 41(1): 89-95.
21	谢岩, 廖松地, 朱曼妮, 等. 轻型汽油车稳态工况下的尾气排放特征[J]. 环境科学, 2020, 41(7): 3112-3120.
	Xie Yan, Liao Song-di, Zhu Man-ni, et al. Emission characteristics of light-duty gasoline vehicle exhaust based on acceleration simulation mode[J]. Environmental Science, 2020, 41(7): 3112-3120.
22	Jeeragal R, Subramanian K A. Experimental investigation for NOx emission reduction in hydrogen fueled spark ignition engine using spark timing retardation[J]. Journal of Thermal Science, 2019, 28(4): 789-800.
23	Kosmadakis G M, Rakopoulos D C, Rakopoulos C D. Methane/hydrogen fueling a spark-ignition engine for studying NO, CO and HC emissions with a research CFD code[J]. Fuel, 2016, 185(1): 903-915.

相关文章 15

[1]	卢凯,吴蔚,林观荣,田鑫,徐建闽. 基于KNN回归的客运枢纽聚集人数组合预测方法[J]. 吉林大学学报(工学版), 2021, 51(4): 1241-1250.
[2]	王德兴,吴若有,袁红春,宫鹏,王越. 基于多尺度注意力融合和卷积神经网络的水下图像恢复[J]. 吉林大学学报(工学版), 2021, 51(4): 1396-1404.
[3]	董延华,刘靓葳,赵靖华,李亮,解方喜. 基于BPNN在线学习预测模型的扭矩实时跟踪控制[J]. 吉林大学学报(工学版), 2021, 51(4): 1405-1413.
[4]	徐卓君,杨雯婷,杨承志,田彦涛,王晓军. 雷达脉内调制识别的改进残差神经网络算法[J]. 吉林大学学报(工学版), 2021, 51(4): 1454-1460.
[5]	钱榕,张茹,张克君,金鑫,葛诗靓,江晟. 融合全局和局部特征的胶囊图神经网络[J]. 吉林大学学报(工学版), 2021, 51(3): 1048-1054.
[6]	宋大凤,杨丽丽,曾小华,王星琦,梁伟智,杨南南. 基于行驶工况合成的混合动力汽车电池寿命优化[J]. 吉林大学学报(工学版), 2021, 51(3): 781-791.
[7]	宋震,李俊良,刘贵强. 基于深度学习和限幅模糊的变转速液压动力源恒流量预测方法[J]. 吉林大学学报(工学版), 2021, 51(3): 1106-1110.
[8]	尚福华,曹茂俊,王才志. 基于人工智能技术的局部离群数据挖掘方法[J]. 吉林大学学报(工学版), 2021, 51(2): 692-696.
[9]	魏晓辉,周长宝,沈笑先,刘圆圆,童群超. 机器学习加速CALYPSO结构预测的可行性[J]. 吉林大学学报(工学版), 2021, 51(2): 667-676.
[10]	李志军,刘浩,张立鹏,李振国,邵元凯,李智洋. 过滤壁结构对颗粒捕集器深床过滤影响的模拟[J]. 吉林大学学报(工学版), 2021, 51(2): 422-434.
[11]	刘纯国,于晓彤,岳韬,李东来,张明哲. 双曲率筋条壁板铣削回弹预测[J]. 吉林大学学报(工学版), 2021, 51(1): 188-199.
[12]	何德峰,罗捷,舒晓翔. 自主网联车辆时滞反馈预测巡航控制[J]. 吉林大学学报(工学版), 2021, 51(1): 349-357.
[13]	慕文龙,那景新,谭伟,王广彬,申浩,栾建泽. 基于FTIR分析的CFRP-铝合金粘接接头剩余强度预测[J]. 吉林大学学报(工学版), 2021, 51(1): 139-146.
[14]	魏晓辉,孙冰怡,崔佳旭. 基于图神经网络的兴趣活动推荐算法[J]. 吉林大学学报(工学版), 2021, 51(1): 278-284.
[15]	王柯俨,王迪,赵熹,陈静怡,李云松. 基于卷积神经网络的联合估计图像去雾算法[J]. 吉林大学学报(工学版), 2020, 50(5): 1771-1777.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

类型	排量/L	压缩比	缸数	缸径×行程/mm
汽油机A	1.60	9.50	4	79.7×81.1
汽油机B	1.80	9.60	4	82.5×84.1
汽油机C	2.00	11.65	4	82.5×92.8
柴油机A	7.79	17.50	6	115×125
柴油机B	9.00	17.00	6	123×132

类型	最大功率/kW	最大功率转速/（r·min^-1）	最大转矩/（N·m）	最大扭矩转速/（r·min^-1）
汽油机A	140	5600	240	2400~5200
汽油机B	118	5000~6200	250	1500~4500
汽油机C	147	5100~6000	280	1700~5000
柴油机A	219	2400	998	1400
柴油机B	294	1900	1900	1100~1450

隔点取点法	数据分类	转速/（r·min^-1）
间隔一组转速选取训练数据	训练组转速选取	1000、1500、2000、2800、 3600、4500、5500
间隔一组转速选取训练数据	预测组转速选取	1250、1750、2400、3200、4000、4500、5000
间隔两组转速选取训练数据	训练组转速选取	1000、1750、2800、4000、5500
间隔两组转速选取训练数据	预测组转速选取	1250、1500、2000、2400、3200、3600、4500、5000
间隔三组转速选取训练数据	训练组转速选取	1000、2000、3600、5500
间隔三组转速选取训练数据	预测组转速选取	1250、1500、1750、2400、2800、3200、4000、4500、5000

项目	传统BP神经网络方法	训练最优模型预测	模型群预测法
建模	0.988 23	0.989 79	0.95~0.983 7
预测	0.953 75	0.968 53	0.983 59

项目	NOx	CO	HC
模型群中模型个数	2000	200	309
优于模型群预测法预测结果的模型个数	2	0	0
优于模型群预测法预测结果的模型占比/%	0.10	0	0