Estimation of Future Thickness of Carbon Steel Pipe and Curing Time of Adhesive of GRE Pipe by Using Neural Network Models

  • Qasim Mohammed Doos Department of Mechanical Engineering / University of Baghdad
  • Mustafa M. Mansour
Keywords: Neural network, carbon steel pipe, GRE pipe, Alyuda neuroIntelligence, curing time

Abstract

The main objective of this research is to estimate both of (future thickness of carbon steel pipe and curing time of Adhesive of GRE pipe) by using neural network model. Alyuda NeuroIntelligence software has been used to obtain these two models. These models will be based on multi – layer feed forward neural network and by applying two experiments for each case, the best networks have been concluded to estimate these cases. The results shows that the network with a number of hidden neurons 5 and that has been trained by conjugate gradient descent algorithm and with using logistic activation function for hidden and output layer gave good performance indication for estimating the future thickness which gave results of network output that are nearly closer to the targets, with correlation (0.9999) and R-Squared (0.9967), while the network with a number of hidden neurons 6 and that has been trained by Quasi – Newton algorithm and with using Hyperbolic Tangent activation function for hidden and output layer gave good performance indication for estimating the curing time which gave results of network output that are nearly closer to the targets, with correlation (0.9999) and R-Squared (0.9958).

Published
2019-01-15
How to Cite
Doos, Q., & Mansour, M. (2019). Estimation of Future Thickness of Carbon Steel Pipe and Curing Time of Adhesive of GRE Pipe by Using Neural Network Models. Association of Arab Universities Journal of Engineering Sciences, 25(5), 579-597. Retrieved from https://jaaru.org/index.php/auisseng/article/view/255