Prediction and optimization of CuCrZr microhardness based on PSO-ensemble learning model

Cu-Cr-Zr alloy is widely used in electrical equipment manufacturing, industrial machinery manufacturing and other fields by virtue of its excellent performance. Since the traditional trial-and-error method is limited by the time consumption and the huge unexplored component space, in this study, a machine learning method is proposed to introduce the trained ensemble learning model into the PSO algorithm with asymmetric learning factor, with the purpose of discovering Cu-Cr-Zr alloy with excellent performance in the unexplored component space. The features of the existing data set were filtered by correlation, recursive elimination and exhaustive selection, and a set of key features affecting the properties of the alloy were finally obtained. The micro-hardness of the alloy was improved by designing the key features. It is proved that the key characteristics affecting microhardness are valence electron number, bulk modulus, milling time, vial speed and sintering temperature under the condition of fixed ball powder weight ratio (BPR) and sintering time. On this basis, the composition and process parameters for preparing the highest microhardness alloy are further confirmed. It provides a new idea for improving the mechanical properties of Cu-Cu-Zr alloy and expanding its application range.