A HYBRID ENSEMBLE LEARNING MODEL FOR EVALUATING THE SURFACE ROUGHNESS OF AZ91 ALLOY DURING THE END MILLING OPERATION

In metal-cutting operations, the surface roughness of the end product plays a significant role. It not only affects the aesthetic appearance of the end product but also signifies the product’s performance in the long run. Products with a high surface finish have higher endurance limits with negligible local stresses. On the other hand, products with rough surfaces are subjected to high stresses when they are engaged in various mechanical operations with varying loads. Surface roughness depends on various machining factors such as feed rate, depth of cut, cutting speed, or spindle speed. Therefore, it is required to predict surface roughness for the given machining parameters to reduce the cost and increase the life of the end product. In this work, an attempt has been made to evaluate the surface roughness of AZ91 alloy during the end milling operation. In this regard, various state-of-the-art ensemble learning models have been adopted and compared with the proposed hybrid ensemble model. The proposed hybrid ensemble model is the integration of random forest, gradient boosting, and a deep multi-layered neural network. In order to evaluate the performance of the proposed model, comparative analyses have been made in terms of mean square error, mean average error, and $R^2$ score. The result shows that the proposed hybrid model gives minimum error for surface roughness.