Simulation of combustion characteristics and prediction of combustion performance using machine learning in an integrated afterburner

An afterburner design concept that integrated the fuel injection pipe, radial flame stabilizer, and turbine rear frame support plate was suggested as a solution to the problem of increasing the thrust-to-weight ratio of aero-engine. The thermal flow field and combustion characteristics of the afterburner under different kerosene-air ratios and bypass ratios were investigated through large eddy simulation. Distribution characteristics of the temperature, velocity, and pressure in the combustion chamber and their impacts on the combustion process were studied. The distributions of fuel droplets, oxygen, carbon dioxide, and water were analyzed. Using the intake pressure, intake temperature, fuel rate, bypass ratio, and axial distance as input variables, the machine learning method was developed and predictions were made for the total pressure recovery coefficient, temperature uniformity coefficient, and combustion efficiency. The results indicate that the overall combustion performance of the integrated afterburner is high, and only a localized combustion weakening zone is observed below the flame. There are three low-velocity recirculation zones in front of the combustion chamber, located in the strong flame zone, the middle and lower parts of the fuel injection pipe, and the tail of the central cone. As the kerosene-air ratio increases, the combustion efficiency is decreased, while an increase in bypass ratio leads to an improvement in combustion efficiency. The proposed machine learning model has a corrected determination coefficient greater than 0.788 for both the training and testing datasets, indicating good predictive performance.