An Optimized ANN Approach for Cutting Forces Prediction in AISI 52100 Bearing Steel Hard Turning

Cutting forces are classified among the important technological parameters in machining processes due to their significant impacts on product quality. A large number of interrelated machining parameters have a great influence on cutting forces so it is quite difficult to develop a proper theoretical model to describe efficiently and globally a machining process.In this paper, an artificial neural network (ANN) model is then proposed to predict cutting force components during hard turning of an AISI 52100 bearing steel using CBN cutting tools. This study is based on an experimental dataset of cutting forces measured during hard turning. Cutting speed (Vc, m/min), feed rate (f, mm/rev), cutting depth (ap, mm) and workpiece hardness (HRc, MPa) are taken as input parameters in the ANN model, while the three cutting force components (feed force Fa, radial force Fr and cutting force Ft, in N) are the output data.The ANN model consists of a multi-layer feed-forward, trained by a back-propagation (BP) algorithm. The influence of a double hidden layer (instead of a single hidden layer) is investigated, and a comparison is carried out between Bayesian Regularization associated with Levenberg–Marquardt algorithm (BR/LM) and simple Levenberg–Marquardt algorithm (LM). A various number of neurons in the hidden layer are also tested.The best prediction accuracy is found while using a feed forward single hidden layer ANN trained by BR/LM and using a sigmoid activation function on hidden layer and a linear one on output layer. The best structure uses 11 neurons in the hidden layer and average prediction errors on the testing dataset are given: 11.47% on Fa, 11.47% on Fr and 6.17% on Ft.