Soft-measuring model for grinding particle size based on the improved GWO-BP network model

The traditional grey wolf optimization(GWO) completes the optimization by simulating the steps of wolf encircling, hunting and attacking. Aiming at the shortcomings of the traditional GWO, such as low accuracy of the solution, weak local search ability, and slow convergence rate at the later stage, an improved GWO based on adaptive search and Gaussian mutation was proposed. In the initial population, using the opposite learning strategy to enhance the average fitness of the initial population and ensure the diversity of the population; In the evolutionary process, using the nonlinear convergence factor to coordinate the search ability of the optimization algorithm; In the position update, adaptive adjustment strategy is proposed to improve the convergence speed and introduce Gaussian mutation operator to avoid falling into local optimum. In this paper, the improved GWO-BP network is applied to the typical two-stage grinding process to established the soft-measuring model for grinding particle size. Through the prediction and simulation of the model, and compared with the basic GWO, Seeker Optimization Algorithm (SOA) and Particle Swarm Optimization (PSO), the simulation results show that the improved GWO has faster convergence speed and higher prediction accuracy, the on-line soft measurement of the grinding particle size can be well achieved for real-time control in the later period.