Performance analysis of ELM-PSO architectures for modelling surface roughness and power consumption in CNC turning operation

Abstract

The turning operation in the Computer Numerical Control (CNC) needs optimal machining parameters to achieve higher machining efficiency. The selection of machining parameters is very important to find the best performances in machining process. In this study, two different architectures of particle swarm optimization based extreme learning machine were analyzed for modelling inputs parameters: feed rate, cutting speed and depth of cut to output parameters: surface roughness and power consumption. The data were collected from 15 experiments using carbon steel AISI 1045 which were separated into training and testing dataset. Our experimental results shows that Architecture II is the most outstanding model with mean absolute percentage error (MAPE) of 0.0469 for predicting the training data and 0.204 for predicting the testing data.