A New Method for Hierarchical Clustering Analysis of Large-Scale Machine Tools

Abstract

Due to the various requirements of users and the complexity of machine tools, most machine tools were produced in a small-batch production mode, resulting in the largely prolonged design cycle and the reduced competitive ability of enterprise. To improve the design efficiency and reply to the changeful design requirements, the module design has been expected to be used in the rapid design of machine tools. To achieve the module design of machine tools, this paper proposed a method of module partition (one important step of module design) based on the function/means trees method and hierarchical clustering algorithm. The initial unit is firstly obtained by function/means trees method, and then the degree of correlation and its weight between function units is analyzed and calculated. Subsequently, a hierarchical module partition result (a hierarchical diagram) is calculated by hierarchical clustering algorithm. Furthermore, the function-structure mapping of the NC machine was achieved using an axiomatic design method focusing on the design requirement with one order of weight. In addition, this paper taken an example (precision horizontal machining center THM65160) to validate and analyze the characteristics of this method of module partition, and the test and analysis results showed that the size of initial function unit matrix is efficiently reduced by using function /means trees method, and the module result (which calculated by the hierarchical clustering algorithm) was helpful to achieve a rapid retrieval of module partition results for the products with the multi-weight design requirements. Compared to the other clustering algorithms, it can be summarized that this proposed method was beneficial to achieve the rapid clustering calculations and the relevant module design of products (which had the multi-weight design requirements).