Abstract
Raw ore allocation in mineral processing is crucial for improving the utilization ratio of nonrenewable raw mineral resources. Allocation of raw ore is a nonlinear, multiobjective programming problem. Such a problem is usually modeled under the assumption that the production process is stationary and the model parameters are deterministic. However, in practice, variations in equipment capability and runtime lead to frequent changes of the model parameters including the constraints. To address the above-mentioned issues, this paper formulates a raw ore allocation in mineral processing as a dynamic multiobjective optimization problem. To solve this problem, a new variant of the elitist nondominated sorting genetic algorithm (D-NSGA-II) is proposed, which employs NSGA-II as the basic component assisted by random immigrant scheme, a gradient-based local search strategy, and a mechanism for detecting environmental changes. Simulations are carried out and the results show that the proposed algorithm can efficiently achieve the Pareto front of the multiobjective raw ore allocation dynamic optimization problem. It can simultaneously converge fast while maintaining good population diversity, in particular, in the presence of large environmental changes.