A Machine learning Approach for the prediction of Surface Roughness Using The Tool Vibration Data In Turning Operation

Surface roughness is a key factor in different machining processes and plays an important role in ergonomics, assembly, wear and fatigue life of components. Functionality, performance and durability of parts are also affected by surface roughness. Although maintaining an optimum surface roughness is a major challenge in many manufacturing industries. Surface roughness during machining depends upon machining parameters such as cutting speed, feed rate and depth of cut. Also, tool vibrations during machining have significant influence in surface roughness. In this work an attempt is made to predict the surface roughness of machined components during the turning process by using machine learning techniques with vibration signals. By varying different machining parameters and keeping other tooling and material properties same, a range of surface roughness values can be obtained. For each condition, corresponding tool vibration signals were recorded. Our experimental setup involves collecting this vibration data during the turning operation using a vibration data collector, followed by preprocessing the data and splitting it into training and test sets. We tried to predict the surface roughness by using the vibration data employing various machine learning regression techniques including Linear Regression, Ridge Regression, Support Vector Regression (SVR), Decision Tree Regression, Random Forest Regression, Gradient Boosting Regression, K-Nearest Neighbors Regression (KNN), and Neural Network Regression. The results indicate that the Random Forest Regression model demonstrates the highest performance with a Mean Squared Error (MSE) of 0.183 and an R2 score of 0.835, followed by Gradient Boosting Regression with an MSE of 0.214 and an R2 score of 0.808. The SVR model also shows competitive performance with an MSE of 0.244 and an R2 score of 0.780. In contrast, simpler models like Linear Regression and Ridge Regression exhibit lower R2 scores of approximately 0.676 and higher MSE values. The study highlights the importance of feature extraction and model selection in achieving accurate and reliable surface roughness predictions, ultimately contributing to enhanced machining process control and product quality.