Surface roughness prediction for CNC-turned C45 steel utilising adaptive neuro-fuzzy inference systems

The surface roughness, Ra of mechanical products is an important parameter commonly used to evaluate the surface quality of a part after machining. Therefore, developing accurate prediction models for surface roughness is essential for optimizing manufacturing processes without multiple and repetitive experiments, leading to significant savings in both cost and time. This study examines the impact of turning parameters, including cutting speed, V, feed rate, f, and depth of cut, d, on Ra and develops a predictive model for the turning of C45 steel. The Taguchi method and the analysis of variance (ANOVA) were used to design the experiments and analyze the effects of machining parameters on Ra. Additionally, a predictive model for Ra was developed using the Adaptive Neuro-Fuzzy Inference System (ANFIS), and the efficiency of this model was evaluated based on the coefficient of determination, R² and the Root Mean Square Error (RMSE). The results of the ANOVA analysis showed that all three cutting parameters had significant effects on Ra. However, the f parameter had the most significant influence on Ra at 90.35%, followed by V and d at 1.9% and 3.24%, respectively. The developed ANFIS prediction model with Gaussian membership functions for Ra achieved R² of 0.989 for training and 0.963 for testing and RMSE of 0.112 and 0.223, respectively. These results indicate that the ANFIS model can predict Ra relatively accurately based on cutting parameters. Thus, the results of this research can be useful in C45 steel turning operations, which further enables assessment and improvement of cutting parameters to reduce Ra, thereby improving product quality.