|
|
Performance Prediction of Diesel Engine Based on Machine Learning
WANG Wei
2022, 47(19):
123-129.
DOI: 10.16638/j.cnki.1671-7988.2022.019.023
With the increasingly severe energy situation and stringent emission regulations, the study
of economy and emission performance of diesel engine has become an important topic in recent years.
At present, the research of diesel engine performance is mainly based on various test and simulation
methods. However, the development and application of various new technologies continuously
improve the nonlinear complexity of diesel engine system. The test and simulation methods are
increasingly difficult to meet the research needs and the research cost is also rising sharply. Therefore,
based on the actual bench test data, the representative indexes of diesel engine economy and emission,
brake specific fuel consumption and NOx, are predicted by machine learning method. The suitable
parameters of support vector regression, decision tree, random forest and back propagation neural
networks machine learning algorithms are selected for testing, and four different prediction models are established. The back propagation neural networks model with better prediction effect is screened
by comparison, and the genetic algorithm-back propagation neural networks prediction model is
further improved from back propagation neural networks model by genetic algorithm. The prediction
results show that the prediction errors of the genetic algorithm-back propagation neural networks
model for Brake Specific Fuel Consumption and NOx are 1.78% and 1.86%, respectively, which are
about 15% higher than those of the back propagation neural networks model. So, the genetic
algorithm-back propagation neural networks model has good prediction accuracy and generalization
ability.
References |
Related Articles |
Metrics
|
