Prediction of surface roughness in turning of Ti-6Al-4V using cutting parameters, forces and tool vibration

International Conference on Recent Advances in Materials & Manufacturing Technologies

Sahu, Neelesh Kumar^1 ; Andhare, Atul B.^2 ; Andhale, Sandip^2 ; Abraham, Roja Raju^2

Shri Ramdeobaba College of Engineering and Management, Nagpur, India^1

Visvesvaraya National Institute of Technology, Nagpur, India^2

关键词: Artificial neural network modeling;    Average absolute error;    Backward elimination;    Correlation coefficient;    Cubic boron nitride (cBN);    Cutting parameters;    Full factorial design;    Response surface methodology;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/346/1/012037/pdf
DOI  :  10.1088/1757-899X/346/1/012037

来源: IOP

【 摘 要 】

Present work deals with prediction of surface roughness using cutting parameters along with in-process measured cutting force and tool vibration (acceleration) during turning of Ti-6Al-4V with cubic boron nitride (CBN) inserts. Full factorial design is used for design of experiments using cutting speed, feed rate and depth of cut as design variables. Prediction model for surface roughness is developed using response surface methodology with cutting speed, feed rate, depth of cut, resultant cutting force and acceleration as control variables. Analysis of variance (ANOVA) is performed to find out significant terms in the model. Insignificant terms are removed after performing statistical test using backward elimination approach. Effect of each control variables on surface roughness is also studied. Correlation coefficient (R2pred) of 99.4% shows that model correctly explains the experiment results and it behaves well even when adjustment is made in factors or new factors are added or eliminated. Validation of model is done with five fresh experiments and measured forces and acceleration values. Average absolute error between RSM model and experimental measured surface roughness is found to be 10.2%. Additionally, an artificial neural network model is also developed for prediction of surface roughness. The prediction results of modified regression model are compared with ANN. It is found that RSM model and ANN (average absolute error 7.5%) are predicting roughness with more than 90% accuracy. From the results obtained it is found that including cutting force and vibration for prediction of surface roughness gives better prediction than considering only cutting parameters. Also, ANN gives better prediction over RSM models.