Skip to Content
You are currently on the new version of our website. Access the old version .
PolymersPolymers
  • Article
  • Open Access

12 April 2025

19 Pages

Prediction of Rheological Properties of PVA Fiber and Nano-SiO2-Reinforced Geopolymer Mortar Based on Back Propagation Neural Network Model Optimized by Genetic Algorithm

,
,
and
1
School of Water Conservancy and Transportation, Zhengzhou University, Zhengzhou 450001, China
2
State Key Laboratory of Tunnel Boring Machine and Intelligent Operations, Zhengzhou 450001, China
*
Author to whom correspondence should be addressed.

Abstract

The rheological properties of mortar are of vital importance to ensure the quality and durability of engineering structures, improving construction efficiency and adapting to different construction environments. This research focused on examining the rheological properties of geopolymer mortar (GM) with the incorporation of metakaolin (MK), nano-SiO2 (NS) and polyvinyl alcohol (PVA) fibers. The research focused on varying concentrations of PVA fiber ranging from 0 to 1.2% (interval of 0.2%) and NS ranging from 0 to 2.5% (interval of 0.5%). As the mix proportion optimization of GM is normally carried out experimentally, a significant amount of labor and material resources was consumed. Based on large amounts of authentic operation data, a prediction model of rheological properties for NS- and PVA-fiber-reinforced GM was developed using a back propagation (BP) neural network. Subsequently, the parameters were refined using a genetic algorithm (GA) to predict the rheological properties of GM reinforced with different dosages of NS and PVA fiber. Three rheological parameters, including static yield stress, plastic viscosity and dynamic yield stress, were used to evaluate the rheological properties of GM. Moreover, parameters of Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and Mean Absolute Error (MAE) were applied to assess the capability of the algorithms. When the GA–BP neural network was used, compared with the BP neural network, the coefficient of determination (R2) of the static yield stress, plastic viscosity, and dynamic yield stress increased by 4.40%, 2.11% and 15.28%, respectively, and the GA–BP neural network provided a superior fitting effect, higher prediction accuracy and faster convergence. Based on the outputs of the developed model, the GA–BP neural network can be adopted as a precise method to forecast the rheological properties of GM reinforced with NS and PVA fibers.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.