Prediction of the NOx emissions from experimental heavy-duty engine using optimized Sparrow search algorithm combined with back propagation neural network

It is crucial to control Nitrogen oxide (NOx) emissions from diesel engines as its seriously affects on environment and human health. However, measuring NOx emissions through experiments is not only consumes a huge amount of human, financial and time costs, but also prolongs engine development cycle...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Journal of environmental chemical engineering Ročník 13; číslo 4; s. 117380
Hlavní autoři: Pan, Xiubin, Guan, Wei, Gu, Jinkai, Sang, Hailang, Wang, Hui, Zhou, Shengkai, Liu, Xing
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier Ltd 01.08.2025
Témata:
Back propagation neural network
Heavy-duty engine
NOx emissions
Prediction model
Sparrow search algorithm
Heavy-duty engine
NOx emissions
Back propagation neural network
Sparrow search algorithm
Prediction model
ISSN:2213-3437
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:It is crucial to control Nitrogen oxide (NOx) emissions from diesel engines as its seriously affects on environment and human health. However, measuring NOx emissions through experiments is not only consumes a huge amount of human, financial and time costs, but also prolongs engine development cycle. Therefore, this paper combines experimental data and predictive modeling to study NOx emissions. Firstly, a large amount of engine operating and emissions data was collected through testing on a heavy-duty diesel engine. Secondly, Pearson and Spearman correlation analysis methods are used to select seven most important variables as inputs to reduce the dimension of data. Whereafter, the back propagation neural network (BPNN) model optimized by Sparrow Search Algorithm (SSA) is used to predict the NOx emissions. Results show that the coefficient of determination (R2) of the optimized model for the training and testing sets were 0.99025 and 0.98812, respectively, which verified the good robustness of prediction model. Finally, comparing to six other machine learning models: BPNN, TCN, ELM, GRU, RF, and XGBOOST in predicting NOx emissions, the SSA-BPNN not only exhibited excellent generalization ability but also achieved lower prediction errors. This indicates that SSA-BPNN is a promising method for accurate NOx emissions prediction, and could be an effective means to alleviate the tedious and expensive experimental testing work. •Engine parameters and NOx correlation was analyzed using PCC and SCC.•The performance of the proposed prediction model was evaluated systematically.•Comparison of the prediction performance of seven ML methods for NOx emissions.•The SSA- BPNN model plays a crucial role in NOx emission prediction.
ISSN:2213-3437
DOI:10.1016/j.jece.2025.117380