The problem of grouping parts into families and machines into cells (manufacturing cell formation problem) is considered with the objective of maximizing the grouping efficacy. A non-specialized and non-hybridized evolutionary strategy (ES) is proposed and analyzed for solving this problem. The algorithm uses a modified permutation with separators encoding scheme and unique concept of separators' movements during mutation. The set of experiments confirmed that the proposed approach outperforms any known heuristics. For 7 out of 34 problems considered, better solutions have been found (with the same number of cells as in literature), as compared to the previously published results. Moreover, there are 5 problems for which ES has produced the best known solutions with different number of cells than from the literature.