High-Performance Concrete compressive property prediction via deep hybrid learning

Abstract

The vast usage of concrete made it the second most used material after water. This volume of concrete consumes an enormous number of natural sources and chronically enhances environmental pollution by CO2 emission. Cementitious supplementary materials such as fly ash and micro silica help decrease the usage of cohesive materials in the concrete and improve concrete’s properties, specifically compressive strength. In addition, due to being the by-product materials of other industries, applying these materials contribute to the decline of environmental pollution. On the other hand, fly ash and micro silica decrease the ratio of water to cement and increase the compressive strength (CS) of concrete. High-Performance Concrete (HPC) is one of the types of concrete used in dams, bridges, etc. In order to achieve the compressive strength of HPC, it is necessary to conduct laboratory tests, which are not economical in terms of time and cost. For this reason, in the present study, the prediction of the CS of the mentioned concrete can be done based on soft-based and artificial intelligence. Furthermore, various mixed designs of HPC, such as fly ash and silica fume coupled with different percentages of plasticizers, are considered the base dataset for developing the prediction models. Neural network-based model hybridized with antlion optimization algorithm and biography-based optimization algorithm developed for compressive strength estimation. The result showed that the AMLP-I model with R2 and RMSE values of 0.9879 and 1.9003 accurately predicted compressive strength and can be referred to as the most qualitative prediction model compared to the BMLP model.