Statistical modeling and multi-objective optimization of road geopolymer grouting material via RSM and MOPSO. (15th February 2021)

• Record Type: Journal Article
• Authors: Tian, Zhongnan; Zhang, Zhengqi; Zhang, Kaiwen; Tang, Xiuming; Huang, Shuolei
• Is Part Of: Construction & building materials. Volume 271(2021)
• Publication Date: 2021-02-15
• Subjects: Road engineering -- Geopolymer -- Grouting material -- Multi-objective optimization -- Statistical modeling -- Road reinforcement; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2020.121534
• Languages: English
• ISSNs: 0950-0618

Abstract

Highlights: As a green and carbon-friendly grouting material, RGGM was prepared with industrial by-products at ambient temperature. Models of the initial fluidity and mechanical strength of RGGM were developed using Response Surface Methodology (RSM). RSM coupled with MOPSO algorithm was implemented for predicting and optimizing the properties of RGGM. The predicted results obtained just brought a difference around 5% from the experiments results. Abstract: As a green and carbon-friendly material, road geopolymer grouting material (RGGM) has a great application potential as a replacement of cement grouting materials in subgrade reinforcement. However, the diversity of its components, including mineral precursors and activators, may mislead engineers in evaluating its performance, especially its fluidity and strength, which limits the popularization and application of RGGM. For this, a Response Surface Methodology (RSM) coupled with multi-objective particle swarm optimization (MOPSO) algorithm was developed for predicting and optimizing the variables that affect the initial fluidity and mechanical strength of RGGM. Initially, a Central-Composite Design (CCD) of RSM has been adopted to establish the statistical models of three responses (28d compressive strength, 28d flexural strength and initial Marsh time) of RGGM. Then, the significant effect of different variables on the three responses was investigated by ANOVA and 3D surface diagrams. Finally, MOPSO algorithm was applied to find global solutions using the three statistical models and optimize the multi-performances of RGGM. It is found that all statistical models are significant because the R-squared are more than 0.9, which can also be verified by the Normal probability plot and Predicted vs actual plot. The effect of water consumption on the three responses is the most significant among the single factors, and for the interaction factors, the interaction between water and NaOH is the most significant. Through the optimization of RSM coupled with MOPSO algorithm, the optimal solutions are predicted and verified only about 5% difference from the experimental results. This optimization method (RSM + MOPSO) is helpful for engineers and researchers to have a comprehensive understanding of the performances of RGGM, which will promote its application to cleaner road reinforcement in road engineering.