Modeling of Exhaust Gas Temperature at the Turbine Outlet Using Neural Networks and a Physical Expansion Model

Alessandro Brusa (), Alice Grossi, Mirco Lenzi, Fenil Panalal Shethia, Nicolò Cavina and Ioannis Kitsopanidis
Additional contact information
Alessandro Brusa: Department of Industrial Engineering, University of Bologna, 40136 Bologna, Italy
Alice Grossi: Department of Industrial Engineering, University of Bologna, 40136 Bologna, Italy
Mirco Lenzi: Department of Industrial Engineering, University of Bologna, 40136 Bologna, Italy
Fenil Panalal Shethia: Department of Industrial Engineering, University of Bologna, 40136 Bologna, Italy
Nicolò Cavina: Department of Industrial Engineering, University of Bologna, 40136 Bologna, Italy
Ioannis Kitsopanidis: Ferrari S.p.A., 41053 Maranello, Italy

Energies, 2025, vol. 18, issue 7, 1-16

Abstract: The accurate estimation of exhaust gas temperature across the turbine is always more important for the optimization of engine performance, ensuring durability of the turbine impeller and catalyst, and reducing and calculating emissions concentration. Traditional physical modeling approaches, based on thermodynamic and fluid dynamics features of gas expansion, can be used for this purpose. However, recent advancements in machine learning, particularly neural networks, offer a data-driven alternative that may enhance prediction accuracy and computational efficiency. This study presents a methodology that integrates a semi-physical turbine model for estimating the exhaust gas temperature at the turbine outlet with a neural network-based approach for predicting the pressure at the turbine inlet. The model utilizes the exhaust gas temperature upstream of the turbine, a model for which was developed in a previous work of the authors. The models are calibrated with steady-state data and then evaluated based on accuracy and robustness under transient operating conditions on six driving cycles with different features. In this way, robust and reliable validation of the models is presented, since the testing is performed on various conditions not used for model development and calibration. Results show an average root mean square error of 14%, including the initial portions of driving cycles performed with a cold engine. Thus, the developed approach that includes multiple modeling methods shows a good predictivity, which is the main objective of this research activity.

Keywords: neural networks; engine modeling; exhaust gas temperature; turbine; physical approach; simulation (search for similar items in EconPapers)
JEL-codes: Q Q0 Q4 Q40 Q41 Q42 Q43 Q47 Q48 Q49 (search for similar items in EconPapers)
Date: 2025
References: View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/1996-1073/18/7/1721/pdf (application/pdf)
https://www.mdpi.com/1996-1073/18/7/1721/ (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:jeners:v:18:y:2025:i:7:p:1721-:d:1623821

Access Statistics for this article

Energies is currently edited by Ms. Agatha Cao

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