This study addresses a multi-objective programming model for optimizing periodic maintenance scheduling of department assets with a specified set of machines and instruments under a given planning time period. An aim is to minimize the overall variance of human resources and maintenance costs. The principle of the desirability function is incorporated into the optimization model while considering diminishing marginal utility concepts. The classic krill herd algorithm is modified via three swap mechanisms. After conducting a case study in the large retailer company in Thailand, one of the modified krill herd algorithms is proved to be highly effective in providing high quality solutions. This method is powerful to find out the desired degree of desirability and shows superior in learning preference structures with respect to the alternative solutions examined.