A systematic review and assessment of concrete strength prediction models
Mylvaganam Nithurshan, Yogarajah Elakneswaran
Abstract
In Civil Engineering field, the compressive strength of concrete is a key parameter to design concrete structures and evaluate existing structures. Conventionally the measurement of the strength requires a considerable amount of time (~28 days) and cost. Thus, this paper comprehensively reviews the literature on strength prediction models focusing on the prediction mechanism and its prediction precision. The literature demonstrates that various techniques including mathematics and statistics, analytical, numerical, computational, homogenization and multi-scale model have been employed to develop a model. An analytical model which simply correlates compressive strength and strength influencing factors was first introduced based on a statistical analysis of experimental data and the ideology of concrete technology. With the advancement of concrete technology, those analytical models fail to account for the complexity of strength affecting factors and compute less accurate results. Although Machine Learning (ML) techniques have proven superior accuracy for estimating the mixture proportions and mechanical properties of novel concrete with complex cementitious material, the prediction mechanism is being ‘black box’ which limits its wide application. Since the microstructure of hydrated cement paste was identified as a strength controlling factor in concrete, several models have been developed simulating the hydration of cementitious materials and thus bridging between nano and macro properties such as compressive strength and young’s modulus of concrete. Even though those microstructural models give comparable results with experimental data, various assumptions and hypotheses were considered during the model development. In addition, some of the models could not directly correlate the macro properties of concrete. The review presented in this paper is strong evidence of the need for a powerful model which realistically predicts the mechanical properties of concrete incorporating fundamental physics and chemistry, thermodynamic consideration, bonding and packing arrangement between particles.