Research on dynamic weighing of rare earth oxide based on Kalman-PSO-BP neural network

Molten salt electrolysis is currently a commonly used method for the preparation of rare earth metals, and in order to improve the electrolysis efficiency and product purity, automated equipment is gradually replacing manual operation. The molten salt electrolysis process requires strict control of the temperature in the electrolysis tank, so it is necessary to control the quality of each rare earth oxide discharging. However , the vibration and noise interference during the operation of the equipment leads to dynamic weighing errors. To improve the dynamic weighing accuracy, Kalman filtering and Particle Swarm Optimization(PSO) were combined with Back propagation Neural Network (BP) to obtain the dynamic weighing data of six working conditions by the rare earth oxide automatic discharging equipment. The Kalman-PSO-BP model was constructed and trained by Matlab, and simulations showed that the model error was reduced by 55.9%, 52.6%, 47.2%, 52.9%, 53.7%, and 53.2% compared to the original error.