A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Prediction of average surface roughness and formability in single point incremental forming using artificial neural network
Single point incremental forming (SPIF) is a flexible, innovative, and cheap process for rapid manufacturing of complex sheet metal parts. It is a crucial task for engineers to predict a process when many independent parameters are affecting simultaneously its performance. An artificial neural network (ANN) based prediction model was developed to evaluate average surface roughness (Ra) and maximum forming angle (Ømax) while SPIF forming of AA5052-H32 material. A feedforward backpropagation network with Levenberg—Marquardt algorithm was employed to build ANN model. The ANNs (4-n-1, 4-n-2) were generated by introducing different combinations of transfer functions and a number of neurons. The confirmation runs were performed to verify the agreement between the ANN predicted and the experimental results. The developed ANN model (4-n-1) was capable of predicting the process response with an excellent accuracy and resulted in overall R-value, MSE, and MAPE of 0.99807, 0.0209, and 5.96% for Ra 0.99913, 0.0281, and 0.003 for Ømax. The optimum 4-n-2 model was built with overall R-value, MSE of 0.99999 and 0.057194, respectively. Hence, it was found that the engineering efforts may be reduced in the SPIF process with successful ANN model implementation.
Prediction of average surface roughness and formability in single point incremental forming using artificial neural network
Single point incremental forming (SPIF) is a flexible, innovative, and cheap process for rapid manufacturing of complex sheet metal parts. It is a crucial task for engineers to predict a process when many independent parameters are affecting simultaneously its performance. An artificial neural network (ANN) based prediction model was developed to evaluate average surface roughness (Ra) and maximum forming angle (Ømax) while SPIF forming of AA5052-H32 material. A feedforward backpropagation network with Levenberg—Marquardt algorithm was employed to build ANN model. The ANNs (4-n-1, 4-n-2) were generated by introducing different combinations of transfer functions and a number of neurons. The confirmation runs were performed to verify the agreement between the ANN predicted and the experimental results. The developed ANN model (4-n-1) was capable of predicting the process response with an excellent accuracy and resulted in overall R-value, MSE, and MAPE of 0.99807, 0.0209, and 5.96% for Ra 0.99913, 0.0281, and 0.003 for Ømax. The optimum 4-n-2 model was built with overall R-value, MSE of 0.99999 and 0.057194, respectively. Hence, it was found that the engineering efforts may be reduced in the SPIF process with successful ANN model implementation.
Prediction of average surface roughness and formability in single point incremental forming using artificial neural network
Archiv.Civ.Mech.Eng
Mulay, Amrut (author) / Ben, B. Satish (author) / Ismail, Syed (author) / Kocanda, Andrzej (author)
Archives of Civil and Mechanical Engineering ; 19 ; 1135-1149
2019-12-01
15 pages
Article (Journal)
Electronic Resource
English
Single-point incremental forming and formability-failure diagrams
| British Library Online Contents | 2008
Formability and effect of hole bridge in the single point incremental forming
| British Library Online Contents | 2017
Application of Neural Network Technique to Predict the Formability in Incremental Forming Process
| British Library Online Contents | 2009
Towards Fracture Prediction in Single Point Incremental Forming
| British Library Online Contents | 2013