Due to the various requirements of users and large volume of machine tools, most machine tool manufacturers are adopting the small-lot production mode, which largely prolongs the product development cycle. To raise the design efficiency and reduce the complexity, modular design is becoming widely used in serial production of machine tools, especially large gantry machining centers. In this paper, we introduced a function and structure decomposition method to construct the function and structure correlations between components and employed a clustering method based on atomic theory to solve the module clustering problems. The clustering algorithm is modified by integrating the function and structure matrices in order that it applies to modularization of large machine tools. Furthermore, an evaluation method based on the concept of entropy is proposed here to evaluate and optimal select module partition schemes. Finally, we took the large gantry machining center QLMT6300 as an example and compiled the program with Microsoft Visual C++ 6.0 to illustrate the validity of the methods proposed. The presented clustering and evaluation method can be easily applied to different machine tools for modular design, reducing the work amount of designers.