Session: Student Poster Competition

Submission Number: 185662

How Semi-Empirical Methods Stack-Up Against Surrogate Models - the Case of Nox Prediction 

This work compares the predictive ability and accuracy of empirical and semi-empirical methods with three surrogate models for estimating the value of the Nitrogen Oxide Emissions Index (EINOx) during the operation of the CFM56-7B26 engine under the four operating points defined by the ICAO Landing and Take-off cycle. The use of empirical and semi-empirical methods, such as correlation equations and Fuel Flow methods, to estimate the value of the emissions index of an exhaust gas often leads to unreliable and inaccurate results. Data driven surrogate models have demonstrated the potential to offer better predictions. Thus, three surrogates are developed: a second-degree Polynomial Regressor, a Gradient Boosting Regressor, and an Artificial Neural Network, all trained and validated on the ICAO Emissions Databank data for the CFM56 engine family. Using available experimental and, model derived, thermodynamic data on the CFM56-7B26 engine variant, the emissions index value, for the same operating points, was estimated, for both the surrogates and the considered empirical approaches, respectively. Relative error results, between the predicted and experimental values, indicate that the greatest improvement is observed at operating points of low power, where the maximum relative error was reduced from 1470%, for the correlation equations, to 133%, for the polynomial regressor. At higher power settings, the relative error resulting from the three surrogate models ranges from 2.68% to 29%, significantly lower than that of the empirical and semi-empirical approaches, which ranges from 20% to 220%.

Presenting Author: Panagiotis Antoniadis Aristotle University of Thessaloniki

Presenting Author Biography: Panagiotis (Panos) Antoniadis is a mechanical engineering student with a great interest in aircraft and rocket engine design and development. Having been affiliated with several student lead initiatives in the field of aerospace, he was the lead software developer for the Aristotle Space and Aeronautics Team (ASAT), the aerospace team of his university, tasked with the development of a trajectory simulation software for sounding rockets. He was later appointed head of the rocketry department and led the team towards the development of its first bi-liquid rocket engine. He is currently working on methodologies for emissions estimations and is greatly interested in the fields of combustion and emissions in aircraft and rocket engines while also being affiliated with engine performance assessment and modelling in the past.

Authors:

Panagiotis Antoniadis Aristotle University of Thessaloniki
Vasilis Gkoutzamanis Aristotle University of Thessaloniki
Konstantinos Papadopoulos Aristotle University of Thessaloniki
Konstantinos Bollas Aristotle University of Thessaloniki
Anestis Kalfas Aristotle Unversity of Thessaloniki