Abstract: Most of the prevalent modal synthesis methods are referred to as "single-level synthesis". Their application to dynamic analysis of large complex structures may be confined by computer capacity. Two" improved substructural assembling techniques, called"multi-level synthesis"and "successive synthesis , are presented in order to raise computation efficiency and to be available for calculation of large structures on a computer with small interior capacity. In principle, the multi-level synthesis method is extension of the single-level one, but the successive synthesis only involves two substructures in each step of synthesis, one of which is synthesized in its generalized form of motion equation, and another in its physical form. In present paper the "rigid substructure technique" is also employed.A typical example has been calculated with satisfactory results. Numerical calculations show that the accuracies of the two improved assembling methods are trustworthy provided the frequency criterion proposed in this paper is adopted.