Surface quality is one of the most impellent customer requirements in machining. The main aspect of surface quality on machined parts is probably surface integrity, such as roughness and residual stresses. In particular, residual stresses in the machined surface and subsurface are affected by tool geometry, material and machining parameters. These residual stresses can have significant effects on the service quality and the component life; they can be determined by either empirical or numerical investigations for the selected configurations; however, they are expensive procedures. The problem becomes more difficult if the aim is the inverse determination of the cutting conditions which correspond to a requested residual stress profile, inside the machined material. This paper presents a predictive hybrid model based on the artificial neural networks (ANNs) and finite element method (FEM) that can be used for both forward and inverse prediction. The former is able to determine a residual stresses profile corresponding to a given tool, material and process conditions, the latter is able to determine these conditions when a constraint on the residual stresses distribution is given. Three layer neural networks were trained basing on selected data from numerical investigations on hard machining of 52100 bearing steel, and then validated with data obtained by experiments. Prediction errors range between 4% and 16% for the whole data set, both for forward analysis and inverse process design, showing that this hybrid ANN-FEM based approach provides a robust framework for 52100 machining analysis.
A hybrid FEM-ANN appraoch for predicting residual stresses and the optimal cutting conditions during hard turning of AISI 52100 bearing steel
UMBRELLO, Domenico;AMBROGIO, Giuseppina;FIlice L;
2008-01-01
Abstract
Surface quality is one of the most impellent customer requirements in machining. The main aspect of surface quality on machined parts is probably surface integrity, such as roughness and residual stresses. In particular, residual stresses in the machined surface and subsurface are affected by tool geometry, material and machining parameters. These residual stresses can have significant effects on the service quality and the component life; they can be determined by either empirical or numerical investigations for the selected configurations; however, they are expensive procedures. The problem becomes more difficult if the aim is the inverse determination of the cutting conditions which correspond to a requested residual stress profile, inside the machined material. This paper presents a predictive hybrid model based on the artificial neural networks (ANNs) and finite element method (FEM) that can be used for both forward and inverse prediction. The former is able to determine a residual stresses profile corresponding to a given tool, material and process conditions, the latter is able to determine these conditions when a constraint on the residual stresses distribution is given. Three layer neural networks were trained basing on selected data from numerical investigations on hard machining of 52100 bearing steel, and then validated with data obtained by experiments. Prediction errors range between 4% and 16% for the whole data set, both for forward analysis and inverse process design, showing that this hybrid ANN-FEM based approach provides a robust framework for 52100 machining analysis.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.