Method of ensemble modeling for abrasive water jet machinability of metal materials

Abstract

As a potential machining tool that can cut almost all kinds of materials, abrasive water jet (AWJ) has been widely used in various fields. However, when facing a new type of material, obtaining the machinability number, which has been used to calculate the cutting head traversing speed, is a big challenge in application. Currently, the machinability number can be obtained through a series of experiments. Although this method is feasible, it is time-consuming and involves high experimental costs. In this paper, an ensemble modeling method was investigated to predict the machinability number of metal materials using material properties that are usually published in the public domain. Two data enhancement algorithms, SMOTER and WERCS, and a quadratic polynomial feature enhancement method are used for small sample data pre-processing. Furthermore, two ensemble models, RandomForest and LightGBM, have been built. With these models, the machinability number could be calculated quickly instead of obtaining it through time-consuming experiments. The data pre-processing method and the ensemble modeling method in this paper could also be used in other small sample data scenarios.