The machine-part group formation(MPGF) problem arises frequently in FMS planning. By viewing the problem as one of finding good assignments, a powerful solution algorithm is presented. The new algorithm solves the threshold dilemma found in previous solution procedures employing similarity coefficients. It also compared favorably with other existing MPGF algorithms by finding minimum exceptional elements for the tested problems. Furthermore the new algorithm can solve dynamic and more realistic MPGF problems by considering production volumes or costs. Such diverse machine-part relationship values were not considered in previous MPGF studies, which included only 0,1 incidence values. An example problem is solved where production volumes are the elements of MPGF incidence matrix.