A novel Bayesian inference method for predicting optimum strength gain in sustainable geomaterials for greener construction

Abstract

There has been growing research interests in the study of sustainable geomaterials to reduce or replace the use of cement to promote greener construction. A machine learning technique based on Bayesian inference was proposed in this study to predict the optimum strength gain in sustainable geomaterials as an alternative to preliminary investigation of new materials and to supplement existing experimental design process. The proposed novel methodology was implemented using two established case studies on sustainable geomaterials previously studied by the second author: (i) fly ash-based geopolymer concrete and (ii) sustainable cementitious blends for soft soil stabilization in order to validate the proposed Bayesian methodology for wider application considering efficiency and sustainability as opposed to performing excessive conventional laboratory-based destructive tests. The eventual results show that the proposed Bayesian approach, which implements the 3-stage data training, validating, and updating process could reliably and accurately predict the strength of geomaterials, despite them having very different mix design requirements.