Prediction of the Mechanical Behaviour of HDPE Pipes Using the Artificial Neural Network Technique

Abstract

Actual statistics show that in recent years, more than 90% of the water distribution pipes installed in the world are made of plastic, exclusively polyethylene (PE). Due to the extensive use of these materials, it is necessary to have a good understanding of the mechanical properties of HDPE used for distribution system piping. For our study, we selected HDPE pipes as the material of choice. We then took a new approach to the analysis and prediction of mechanical properties, using new models based on Artificial Intelligence. In this paper, experimental tensile tests were conducted to obtain the mechanical properties of pipes. The first part of this work focuses on the mechanical tests, specifically tensile tests, while the second part centers on the numerical procedure for predicting the mechanical characteristics, a deep learning model was developed for prediction. The model was trained using a large dataset, including information on pipes. Specially designed deep learning architectures capture complex relationships and patterns in the data, enabling accurate predictions, Several ANN models were created to predict mechanical behaviour based on experimental data. We analyzed Bayesian regularization using MATLAB, an advantage of BR artificial neural networks is their ability to reveal potentially complex relationships. The results showed that the constructed prediction model is satisfactory since the M.S.E. value is nearly 0 (0.00023) and the  value is close to 1 (0.99934).  This study evaluates the advantages of our methodology by demonstrating the predictive power of an AI-based method and how well it predicts HDPE pipe behavior. The paper study will have significant effects on the water distribution and plastics industries.