Random Forest 방법을 적용한 초저유량 하이드로싸이클론 설계 및 성능예측

Abstract

A hydrocyclone is a particle separation device. Due to their simple shapes and real-time particle separation functions, hydrocyclones are used in several industrial sites. However, the design of a hydrocyclone through numerical analysis takes prolonged time. In this study, a machine learning method is utilized to reduce the hydrocyclone design time. By using a random forest-based learning algorithm, the following three tasks were accomplished: particle separation efficiency was predicted under given design parameters; design parameters were extracted for a given bid size and the corresponding separation efficiency; finally, an extrapolation-based separation efficiency was investigated. The performance of the proposed learning algorithm-based prediction is demonstrated by comparing the results with numerical analysis data.