Evaluating the impact of solid particle temperature on high-speed gas flow using Eulerian–Lagrangian method and machine learning approach

The interaction of high-speed flows and solid particles is one of the complex studies among multiphase flows that has been less investigated due to the complexity of the simulation. In this study, the effects of injecting solid particles with different temperatures on the characteristics of high-speed gas flow inside a three-dimensional cylinder are investigated. For this purpose, the Eulerian–Lagrangian method is used to study the interaction of solid particles and gas flow with each other. Also, the accuracy of the machine learning model in predicting the properties of gas flow in interaction with the solid particles with different temperatures is investigated. For this purpose, the optimized random forest model is used to predict the changes in the total flow temperature along the cylinder. The results show significant effects of the temperature of solid particles on the characteristics of high-speed gas flow inside the cylinder. Injecting solid particles with a temperature equivalent to the temperature of the total gas flow increases the temperature of the total gas flow along the cylinder and reduces the temperature of the solid particles. Also, injecting solid particles with a lower temperature reduces the temperature of the total flow and increases the temperature of the solid particles. The temperature of solid particles with [Formula: see text][Formula: see text]K increases from 300[Formula: see text]K to 550[Formula: see text]K in a very short time of about 0.00074[Formula: see text]s. At the same time, the temperature of particles in another case with [Formula: see text][Formula: see text]K decreases from 800[Formula: see text]K to about 700[Formula: see text]K. After the injection of solid particles, their velocity increases. This variation occurs very quickly, so the particle speed increases from 100[Formula: see text]m/s to 550[Formula: see text]m/s in just 0.0002[Formula: see text]s for both cases. Also, the results of the random forest model indicate the appropriate accuracy of this model in predicting changes in the total temperature of the flow along the cylinder. The results show that the maximum error between CFD results and random forest prediction is about 2%.