Modeling the Static Puncture (CBR) Properties of Non-woven Geotextiles Based on Neural Network and Multi-optimization

Abstract

The optimization of geotextile mechanical properties is crucial for enhancing their performance in civil engineering applications such as soil reinforcement and stabilization. This study focuses on the influence of manufacturing parameters on the static puncture (CBR) properties of polyester geotextiles. Polyester geotextile samples were manufactured using various parameters, including needle-punching density, penetration depth, calendering temperature, and speed. The mechanical properties of the samples, specifically strength and elongation, were evaluated using the CBR test according to EN ISO 12236. The data were analyzed using multivariate analysis of variance, followed by statistical analysis to determine the influence of the manufacturing parameters on the mechanical properties. Furthermore, the relationship between these parameters and the mechanical properties was modeled using artificial neural networks (ANN) and regression analysis. The results indicated that all manufacturing parameters significantly impacted the strength and elongation of the geotextiles. The ANN models, employing two hidden layers, predicted the strength and elongation with errors of 1.43% and 1.26%, respectively.