Experimental Assessment, Optimization, and Multi-scale Modeling of Alkali-Silica Reaction (ASR) Through Machine-Learning Techniques
Abstract
Multi-scale experimental investigation and machine-learning modeling of alkali-silica reaction (ASR) were undertaken in this research. The experimental assessment was implemented in three different levels including aggregate, mortar, and concrete. Several aggregates of different reactivity were tested to obtained their activation energy (AE) by modifying the classification method. Seven different types of fly ash (FA) were characterize and investigated to determine the influencing FA indicators affecting ASR expansion. Based on design of experiment (DOE), accelerated mortar bar test (AMBT) was carried out on 150 mortar bars with different reactive aggregates, FA types, FA percentages, and temperatures. The ASR expansion of the mortar samples was measured at an extended time up to 60 days to match the test duration of the concrete test. Accelerated concrete cylinder test (ACCT) was also implemented on different mixes and at different temperatures. Based on the experimental plan, the tests were designed and the data were collected on three levels of aggregate, mortar, and concrete. In order to develop multi-scale predictive models for ASR expansion of mortar and concrete at the phase of mix design, machine-learning techniques were utilized for the first time to predict the ASR expansion using Artificial Neural Network (ANN), Genetic Programming (GP), and Adaptive Inference Neuro-Funny System (ANFIS). The performance of different models were evaluated using different performance criteria and the prediction results were compared. Moreover, close-form formulations were also derived for ANN and GP model. The results obtained indicated that the multi-scale assessment along with the machine-learning models can constitute a promising and powerful approach to predict the ASR expansion in concrete structures.
Subject
Alkali-Silica Reaction (ASR)Experimental Assessment
Optimization
Multi-scale Modeling
Machine-learning Techniques
Citation
Jalal, Mostafa (2020). Experimental Assessment, Optimization, and Multi-scale Modeling of Alkali-Silica Reaction (ASR) Through Machine-Learning Techniques. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /191690.