Establishment and optimization of prediction model for recovery rate of alloying elements

Steel quality depends on the alloying model precision. While the precision is mainly dependent on the recovery rate of alloying elements calculation, the prediction model for recovery rate of alloying elements was established based on the BP neural network. The simulation shows that using POS algorithm to optimize the model is still easy to fall into local minimum, so a simulated annealing (SA) thought was introduced to improve it. By the comparison we can see that SA-PSO algorithm can overcome above shortcomings. This algorithm strengthens the global convergence ability. It can optimize the model while ensuring high precision and improve the training convergence rate at the same time. The simulation results proved that this model is effective.