Any practical implementation of any multi-objective evolutionary algorithm (MOEA) must include a secondary population composed of all Pareto-optimal solutions found during its search process. Such an implementation with an active participation of solutions from the secondary population into the generational population of the genetic cycle is expected to improve the effectiveness of the MOEA. In this work, two kinds of secondary population, one with set of non-dominated solutions and another with a set of inferior solutions, accrued out of the generation cycles are constructed, and with different combinations of feeding of solutions from these two secondary populations, seven different implementation schemes are designed with an aim of intensifying the convergence and diversification capabilities of the genetic process of MOEA. All the schemes were implemented in a genetic algorithm-based MOEA designed to solve the scheduling problem with dual objectives for a flexible manufacturing system and tested with common experimental data. The performances of the schemes are compared, and the most appropriate implementation scheme is proposed. © 2011 Springer-Verlag London Limited.