High-Speed Computing Method for Powder Mixing Using Machine Learning

Abstract

The discrete element method (DEM) is widely used to analyze powder mixing processes, however DEM simulation is limited by the computing capability. Hence, we proposed an original model, namely Recurrent Neural Network with Stochastically calculated Random motion (RNNSR), to simulate the powder mixing with low computational cost. RNNSR combines recurrent neural network and stochastic model, adapting to powder mixing simulation in rotating drum mixer. The performance of RNNSR was evaluated in terms of degree of powder mixing and effective computation speed. As a result, RNNSR was able to simulate the powder mixing with ultra-fast speed.