Intelligence decision mechanism for prediction of compressive strength of self-compaction green concrete via neural network

Abstract

Civil engineering has a specific position for different forms of concrete as a century-old material. One of the most popular materials for human consumption is made from this substance. In recent years, solutions have been presented for the manufacturing of concrete, making concrete environmentally friendly and also making current waste useable as additives for concrete. In this research, the mechanical features of self-compacting green concrete (SCGC) comprising of compressive, tensile and flexural strength have been studied while diverse amounts of calcium carbide waste (CCW) and rice husk ash (RHA) were tested as partial cement replacements at 3, 7 and 28 days of the curing age. Then, according to available experimental results, artificial intelligence algorithms including Emotional Neural Network-chaotic particle swarm optimization (EANN-CPSO), First-principles molecular-dynamics (FPMD), and conventional Linear Regression (LR) model were applied for predicting the mechanical features of self-compacting concrete (SCC) while adding up to 10% RHA and up to 20% CCW in the SCGC blends. It was reported that the addition of CCW decreases the workability of SCGC mixes and raises the compressive strength (CS) at 28 days for SCGC mixes including 10% of RHA and without CCW compared to control mixes. Also, in both the testing and training phases, the minimum R2 value FPMD, EANN-CPSO and LR models is about 0.904. It was found that the hybrid models of EANN-CPSO-S3, FPMD -S3 and LR-S3 show the most accurate performance with R2 = 0.997 and 0.970 for EANN-CPSO-S3, R2 = 0.967 and 0.954 for FPMD -S3 and R2 = 0.934 and 0.929 for LR-S3 in both phases. The result shows that additional models might increase the performance of the model, such as new algorithms, hybrid models and optimization approaches.