A novel differential evolution algorithm for global search and sensor selection

Differential evolution (DE) algorithm is a simple yet powerful population-based stochastic search technique for solving optimization problems in the continuous search domain. However, the performance of the canonical DE algorithm crucially depends on appropriately choosing mutation strategies and their associated parameter settings. Unsuitable choice of trial vector generation manners and control parameter values may deteriorate the search process. In this paper, a new version of the differential evolution algorithm is reported, in which both diverse mutation operators and mutation rates are heuristically assigned to various individuals. During the iteration process, the whole populations are classified into subgroups by sufficiently analyzed the individuals' state. Multiple population parallel search policy can effectively expedite the convergence of the proposed algorithm. Diverse mutation operators with distinct characters are assigned to relative subgroups, which are considered to be a better balance between exploration and exploitation. The empirical values and negative feedback technique are used in parameters selection, which relieve the burden of specifying the parameters values. The experimental study of the new approach is test on a set of standard benchmark functions and a practical sensor selection problem for turbofan engine health estimation. The simulation results suggest that it outperforms to other state-of-the-art techniques referred to in this paper in terms of the quality of the final solutions.