基于粒子群算法的增压柴油机神经网络模型

陈昊天; 王玥; 曹晶; 张继忠; 邓康耀; 崔毅

doi:10.3969/j.issn.1001-2222.2021.03.001

车用发动机 >

2021 , Vol. 0 >Issue 3: 1 - 7

DOI: https://doi.org/10.3969/j.issn.1001-2222.2021.03.001

栏目

基于粒子群算法的增压柴油机神经网络模型

陈昊天 ,
王玥 ,
曹晶 ,
张继忠 ,
邓康耀 ,
崔毅

展开

（1.上海交通大学动力机械与工程教育部重点实验室，上海 200240；2.上海核工程研究设计院有限公司，上海 200233；3.中国北方发动机研究所（天津），天津 300400）

收起

Neural Network Model of Turbocharged Diesel Engine Based on Particle Swarm Algorithm

CHEN Haotian ,
WANG Yue ,
CAO Jing ,
ZHANG Jizhong ,
DENG Kangyao ,
CUI Yi

Expand

(1.Key Laboratory for Power Machinery and Engineering of Ministry of Education,Shanghai Jiao Tong University,Shanghai 200240，China;2.Shanghai Nuclear Engineering Research & Design Institute Co.，Ltd.,Shanghai 200233，China;3.China North Engine Research Institute（Tianjin）,Tianjin 300400，China)

Fold

摘要

神经网络是建立柴油机性能实时计算模型的有效方法。当前柴油机神经网络模型的研究大多是针对稳态工况开展，为了实现对瞬态性能的合理预测，提出了预测全工况稳态及瞬态性能的通用神经网络模型构建方法。此外，为了解决传统BP神经网络无法保证得到全局最优解、泛化能力较差的问题，采用群体智能算法中的粒子群算法（PSO）进行优化。利用某型涡轮增压柴油机的稳态和瞬态数据作为样本对模型进行训练，并与传统的BP神经网络模型进行对比。研究结果表明, PSO-BP神经网络模型可以有效预测发动机的稳态和瞬态性能，稳态预测最大误差4.54%，瞬态预测最大误差4.93%，与传统BP神经网络模型相比，PSOBP模型可以有效实现全局寻优，提升泛化能力。

关键词： 柴油机; 神经网络; 粒子群算法; 数学模型; 泛化能力

本文引用格式

陈昊天 , 王玥 , 曹晶 , 张继忠 , 邓康耀 , 崔毅 . 基于粒子群算法的增压柴油机神经网络模型[J]. 车用发动机, 2021 , 0(3) : 1 -7 . DOI: 10.3969/j.issn.1001-2222.2021.03.001

Abstract

Neural network is an effective method to establish a realtime simulation model of diesel engine performance. Most of the current research on diesel engine neural network models was carried out for steadystate conditions. In order to realize transient performance prediction of diesel engine, a construction method of general neural network model was proposed to predict steadystate and transient performance of diesel engine in all operating conditions. In addition, in order to solve the problem that the traditional backpropagation (BP) neural network could not guarantee the global optimal solution and the poor generalization ability, the particle swarm optimization algorithm (PSO) in the swarm intelligence algorithm was used for optimization. The steadystate and transient test data of a turbocharged diesel engine was used as samples to train the model and was compared with the traditional BP neural network model. The research results show that the PSOBP neural network model can effectively predict the steadystate and transient performance of engine. The maximum error of steadystate and transient prediction is 4.54% and 4.93% respectively. Compared with the traditional BP neural network, PSOBP model can effectively realize global optimization and improve generalization ability.

Key words： diesel engine; neural network; particle swarm algorithm; numerical model; generalization ability

Options

摘要页面

文章导航

模态框（Modal）标题

摘要

本文引用格式

Abstract