Postfire Residual Capacity Of Steel Fiber Reinforced Volcanic Scoria Concrete Using PSO-BPNN Machine Learning

With the rapid development of the global economy,the large carbon emissions of various industries have become a threat to the survival of human beings.In order to cope with the problem of excessive carbon emissions,it is imperative to develop new green building materials.As a kind of natural green light aggregate,volcanic scoria has abundant deposits all over the world,and replacing ordinary crushed stone aggregate with volcanic slag aggregate can improve the economic and ecological benefits of the construction industry.The application of volcanic scoria can alleviate the problem of carbon emission to a certain extent,which has good social benefits and makes a great contribution to sustainable development.In this study,an artificial neural network(ANN)was used to predict the mechanical properties of steel fiber-reinforced volcanic slag concrete,with the following main findings.(1)In this paper,720 steel fiber volcanic scoria concrete cubic specimens were designed and prepared,and volcanic scoria aggregate replacement rate,steel fiber dosage,water-cement ratio and high temperature were selected as the research factors for the mechanical properties of steel fiber volcanic scoria concrete.High temperature tests,compressive strength tests and split tensile strength tests were performed on the concrete,and the obtained experimental data were established as a neural network database.(2)Based on the data,this paper uses volcanic scoria aggregate replacement rate,steel fiber admixture,water-cement ratio and high temperature as neural network inputs.Six neural network models,including the conventional BP neural network(BPNN),BP neural network optimized using genetic algorithm(GA-BPNN)and BP neural network optimized using particle swarm algorithm(PSO-BPNN)were developed to model the compressive strength and split tensile strength of volcanic scoria concrete separately.After testing the appropriate neural network structure and selecting the appropriate training parameters and algorithms,the neural networks were trained.After the training,the different models established in this paper were evaluated for their prediction accuracy using mean absolute percentage error(MAPE),root mean square error(RMSE)and correlation coefficient(R~2).(3)Finally,a parametric analysis was carried out using the PSO-BPNN model to investigate the effects of different volcanic slag aggregate substitution rates,steel fiber blending,water-cement ratio and high temperature on the residual strength of volcanic scoria concrete after high temperature.