An Explainable Prediction Model for Aerodynamic Noise of an Engine Turbocharger Compressor Using an Ensemble Learning and Shapley Additive Explanations Approach

Rong Huang, Jimin Ni, Pengli Qiao, Qiwei Wang (), Xiuyong Shi and Qi Yin
Additional contact information
Rong Huang: School of Automotive Studies, Tongji University, Shanghai 201804, China
Jimin Ni: School of Automotive Studies, Tongji University, Shanghai 201804, China
Pengli Qiao: School of Automotive Studies, Tongji University, Shanghai 201804, China
Qiwei Wang: School of Automotive Studies, Tongji University, Shanghai 201804, China
Xiuyong Shi: School of Automotive Studies, Tongji University, Shanghai 201804, China
Qi Yin: SAIC Motor, General Institute of Innovation Research and Development, Shanghai 201804, China

Sustainability, 2023, vol. 15, issue 18, 1-22

Abstract: In the fields of environment and transportation, the aerodynamic noise emissions emitted from heavy-duty diesel engine turbocharger compressors are of great harm to the environment and human health, which needs to be addressed urgently. However, for the study of compressor aerodynamic noise, particularly at the full operating range, experimental or numerical simulation methods are costly or long-period, which do not match engineering requirements. To fill this gap, a method based on ensemble learning is proposed to predict aerodynamic noise. In this study, 10,773 datasets were collected to establish and normalize an aerodynamic noise dataset. Four ensemble learning algorithms (random forest, extreme gradient boosting, categorical boosting (CatBoost) and light gradient boosting machine) were applied to establish the mapping functions between the total sound pressure level (SPL) of the aerodynamic noise and the speed, mass flow rate, pressure ratio and frequency of the compressor. The results showed that, among the four models, the CatBoost model had the best prediction performance with a correlation coefficient and root mean square error of 0.984798 and 0.000628, respectively. In addition, the error between the predicted total SPL and the observed value was the smallest, at only 0.37%. Therefore, the method based on the CatBoost algorithm to predict aerodynamic noise is proposed. For different operating points of the compressor, the CatBoost model had high prediction accuracy. The noise contour cloud in the predicted MAP from the CatBoost model was better at characterizing the variation in the total SPL. The maximum and minimum total SPLs were 122.53 dB and 115.42 dB, respectively. To further interpret the model, an analysis conducted by applying the Shapley Additive Explanation algorithm showed that frequency significantly affected the SPL, while the speed, mass flow rate and pressure ratio had little effect on the SPL. Therefore, the proposed method based on the CatBoost algorithm could well predict aerodynamic noise emissions from a turbocharger compressor.

Keywords: turbocharger compressor; aerodynamic noise; ensemble learning; emission prediction model; Shapley Additive Explanation (search for similar items in EconPapers)
JEL-codes: O13 Q Q0 Q2 Q3 Q5 Q56 (search for similar items in EconPapers)
Date: 2023
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/2071-1050/15/18/13405/pdf (application/pdf)
https://www.mdpi.com/2071-1050/15/18/13405/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jsusta:v:15:y:2023:i:18:p:13405-:d:1234890

Access Statistics for this article

Sustainability is currently edited by Ms. Alexandra Wu

More articles in Sustainability from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().