Abstract: To improve the global search ability of a harmony search algorithm， a modified one with global crossover (MHSgc) was proposed. In the MHSgc， the collaborative improvisation of multipleharmonymemory was applied. A neighborlearning strategy replaced the intrinsic pitch adjusting，and thus the population diversity was increased. Moreover， global crossover operation was introduced into the MHSgc algorithm to avoid getting stuck into local minima. Simulations were carried out based on several benchmark functions. The results show that the proposed algorithm outperforms eight intelligent algorithms (IHS， GHS， NGHS， EHS， ITHS， MPSO， DE， ABC) reported in earlier literatures and has better optimization potential.

Key words: harmony search algorithm, multipleharmonymemory, neighborlearning, global crossover, optimization