Estudo da aplicação de redes neurais artificiais e de regressões para a predição de forças de usinagem e rugosidade em processos de torneamento

Abstract

Due to the high competitiveness of the market and high demands from consumers, understanding of manufacturing processes becomes increasingly important for increasing production and obtaining products with characteristics within acceptable tolerance limits.Turning is one of the manufacturing processes that has the greatest relevance in industries and the prediction of characteristics such as machining forces and the roughness of the machined material is of fundamental importance for the planning of production and for the choice of the best parameters that affect these characteristics. With the estimates of the machining forces it is possible to effectively dimension the lathes motors and the importance of studying the roughness in these machining operations is related to factors such as friction and aesthetics of the final part. Thus, this work aims to study the application of mathematical models such as artificialneural networks and regression models for various materials and cutting tools in turning processes to find functions that can provide predictions of the forces and roughness of the part. A literature search was made to find experimental data that could beused in the models and a Python program was also created that applies the mathematical models and displays all the results. It was possible to conclude that the linear regression presented the worst performance in relation to the training data and there was a tendency of overfitting the training data with the increase of the degree of regression.