Heat transfer optimization in oil shale rotary retorting via DEM and BP neural network

To enhance heat transfer between shale ash and oil shale particles in a rotary retorting furnace, this study coupled the discrete element method (DEM) with a particle heat conduction model to simulate mixing and heat transfer, examining the effects of particle filling ratio, furnace rotational speed, and baffle structures. A backpropagation neural network (BP-NN) model was built from simulation data to map furnace operation time with key parameters, and a genetic algorithm was used to optimize parameters to minimize operation time. The research results show that lower filling degrees and higher rotation speeds significantly strengthen particle mixing and heat exchange, which accelerate the system’s stabilization, improve temperature field uniformity, and reduce the temperature standard deviation. The mixing and heat transfer effect of the straight baffle is between that of the right-angle baffle and the inclined baffle, but it causes the largest temperature standard deviation. In contrast, the right-angle baffle demonstrates stronger advantages in heat transfer uniformity during particle lifting and throwing. The constructed BP-NN prediction model achieves a relative error accuracy within 0.25%, effectively solving the long computation time problem of DEM simulation. The optimized parameter combination provides a theoretical basis for the development of high-efficiency and energy-saving rotary retorting furnaces.