Technical gazette, Vol. 27 No. 3, 2020.
Original scientific paper
https://doi.org/10.17559/TV-20141209182117
Artificial Neural Networks Model for Springback Prediction in the Bending Operations
Florica Mioara Serban
; Faculty of Electrical Engineering, Technical University of Cluj-Napoca, Str. G. Bariţiu nr. 26-28, 400027 Cluj-Napoca, Romania
Sorin Grozav
; Faculty of Machine Building, Technical University of Cluj-Napoca, B-dul Muncii 103-105, 400641 Cluj-Napoca, Romania
Vasile Ceclan
; Faculty of Machine Building, Technical University of Cluj-Napoca, B-dul Muncii 103-105, 400641 Cluj-Napoca, Romania
Antoniu Turcu
; Faculty of Electrical Engineering, Technical University of Cluj-Napoca, Str. G. Bariţiu nr. 26-28, 400027 Cluj-Napoca, Romania
Abstract
The aim of this paper is to develop an Artificial Neural Network (ANN) model for springback prediction in the free cylindrical bending of metallic sheets. The proposed ANN model was developed and tested using the Matlab software. The input parameters of the proposed ANN model were the sheet thickness, punch radius, and coefficient of friction. The resulting data is represented by the springback coefficient. Preparation, assessing and confirmation of the model were achieved using 126 data series obtained by Finite element analysis (FEA). ANN was trained by Levenberg - Marquardt back - propagation algorithm. The performance of the ANN model was evaluated using statistic measurements. The predictions of the ANN model, regarding FEA, had quite low root mean squared error (RMSE) values and the model performed well with the coefficient of determination values. This shows that the developed ANN model leads to the idea of being used as an instrument for springback prediction.
Keywords
artificial neural networks; finite element simulation; springback prediction
Hrčak ID:
239096
URI
Publication date:
14.6.2020.
Visits: 1.734 *