Intelligent integrated optimization of mining and ore-dressing grades in metal mines

An intelligent integrated method is proposed for optimizing the head grade and dressing grade in the mining and ore-dressing management of metal mines, beginning with the establishment of a nonlinear constrained optimization model with the objective function of economic benefit, two constraints comprising of the resource utilization rate and the output of concentrate, along with head grade and dressing grade as the decision variables. Particle swarm optimization (PSO) algorithm is then integrated with artificial neural networks to create a PSO–ANN algorithm capable of identifying the optimal grade combination. The outer layer of PSO–ANN uses the PSO algorithm to carry out a global search, with the head grade and dressing grade being combined as swarm particles for evolutionary computation. The constraint handling techniques of feasibility-based rules are used to update the historical best location of each particle (pbest) and the global best location of the swarm (gbest) to guide the particles toward the optimum. The inner layer uses regression model, BPNN and RBFNN to calculate the loss rate, ore-dressing metal recovery rate and costs, respectively, to facilitate the further calculation of the resource utilization rate, the concentrate output and the economic benefit of each particle. Finally, the proposed method is tested by carrying out a case study based upon Daye Iron Mine to indicate its effectiveness and reliability.