Prediction of Cutting Forces Using ANNs Approach in Hard Turning of AISI 52100 steel

Abstract

In this study, artificial neural networks (ANNs) was used to predict cutting forces in the case of machining the hard turning of AISI 52100 bearing steel using cBN cutting tool. Cutting forces evolution is considered as the key factors which affect machining. Predicting cutting forces evolution allows optimizing machining by an adaptation of cutting conditions. In this context, it seems interesting to study the contribution that could have artificial neural networks (ANNs) on the machining forces prediction in both numerical and experiment studies. Feed-forward multi-layer neural networks trained by the error back-propagation (BP) algorithm are used. Levenberg-Marquardt (LM) optimization algorithm was used for finding out weights. The training of the network is carried out with experimental machining data. The input dataset used are cutting speed, feed rate, cutting depth and hardness of the material. The output dataset used are cutting forces (F t -cutting force, F a -feed force and F r -radial force). Results of the neural networks approach, in comparison with experimental data are discussed in last part of this paper.

Comparative analysis and optimization of thermodynamic behavior of combined gas-steam power plant using grey-taguchi and artificial neural network

“…Finally, the neural network simulating methodology is opted to verify and validate the optimal response obtained by the Grey-Taguchi method. Numerous researchers [31,[49][50][51] use this technique for the validation of output results of engineering applications. With the aid of the 'nntool' command and importing input and target parameters from Table 2, the ANN model is created, trained, and simulated using the feed-forward backprop technique with the number of neurons layers to be 3 namely, input layers…”

Determination of Stress Concentration Factor for a Stepped Bar Under Bending Loading: An Artificial Neural Networks Approach

“…The output response (s) using non-linear function is calculated as follows (eq. 7), [7]: (7) During training, weights (w) and biases (b) are initialized to small random values to avoid sharp saturation in activation functions (f).…”

“…It is a good practice to use the Sigmoid function the in hidden layer and the Linear function ( ) in the output layer (S.L.) [7].…”

Mini symposium on cutting and machining: 25 years of ESAFORM activity

“…1 Finite elements output results used by industries to predict the machining performance experimentally evaluated (Arrazola et al 2013a) major topics. With regards to materials, steels are by far the most studied (Makhfi et al 2011, Marouvo et al 2021, followed by Ti alloys, Ni-based alloys (Inconel 718, Waspalloy) and composites.…”