Application of BP Neural Network Model in the Recycled Concrete Performance Prediction

Abstract

When the construction wastes were used as raw materials of Recycled concrete, the type and replacement ratio of recycled aggregates should be considered in addition to mix proportion. It is very difficult to describe the complicated nonlinear relationship between different indexes. Through analyzing design process of BP neural network model, the appropriate network parameters were selected, the BP neural network model about performance of recycled concrete is established. After the BP neural network was trained, the 7-20-3 BP neural network model is established to realize nonlinear mapping about performance of recycled concrete. The results show that the established BP network model can accurately predict the recycled concrete slump, 28d compressive strength and elastic modulus, which has a good applicability to the concrete. InstructionEvery year to billion tons of construction wastes in our country were produce, it is very difficult to handle. Through screening and crushing the waste brick and concrete in the construction waste, the recycled aggregate was produced which can replaced natural aggregate to produce recycled concrete, and it can solved the problems of the construction waste. Domestic and foreign scholars have done a lot of experimental study on the basic properties of recycled concrete, but the source of the recycled aggregates, crushing technology, physical properties are different. In addition to mix proportion, the physical properties of recycled aggregate which include apparent density, bulk density, water absorption and crushing index and replacement ratio of recycled aggregates must be considered. So we need a kind of intelligent analysis tool to realize the recycled concrete strength and working performance and mixture ratio, and the nonlinear mapping between classification of recycled aggregate performance.Artificial neural network theory regards biological neural network as the prototype, analyzing the microstructure, imitating the running mechanism of the biological neural networks, then the numerical calculation model is got which can solve the complex nonlinear relationship between simple input signals. This paper uses the BP neural network model to realize the study on the properties of recycled concrete, the non-linear relation between recycled concrete working performance and various influencing factors was build. And the BP network model is adopted to the practical application to make a prediction of the recycled concrete slump, 28d compressive strength and elastic modulus and verify the applicability of the model. Setting of BP Neural Network Parameters Selection the network structureMany factors affecting the performance of recycled concrete, such as water cement ratio, sand ratio, types and replacement ratio of recycled aggregate, and cement varieties and dosage, additive, etc.

Use of Artificial Intelligence for Predicting Parameters of Sustainable Concrete and Raw Ingredient Effects and Interactions

“…Furthermore, the poor bonding between mortar and RA is also one of the primary reasons for the decreased compressive strength of recycled concrete [ 48 ]. The concrete properties are reduced by up to 10–25% when replacing natural aggregates with RAs, depicting the considerably reduced compressive strength of concrete in the case of 100% substitution [ 49 , 50 ]. Therefore, to avoid the consumption of testing time, cost, and raw materials, prediction models based on experimental data are usually developed to forecast the concrete compressive strength [ 4 , 34 , 51 , 52 , 53 ].…”