Neural Network To Predict Self-compacting Concrete With Its Performance

Self-compacting concrete (SCC) is one of high performance concrete (HPC) that can consolidate under its own weight without any vibration. The SCC has the following advantages: remarkably reducing the noisy pollution and the worker s labor intensity in construction, deducing the rough surface or segregation because of missing-vibration or excessive-vibration in conventional construction, resolving the quality-defect problems resulting from lacking-vibration in the section of complex shapes and densely-packed reinforcement area. At the same time, large of industrial solid waste such as fly ash and blast furnace slag is utilized in the proportioning of SCC. It is helpful in comprehensive utilization of resource and environment protection, so the SCC belongs to "green concrete", it is a branch of HPC which will be developed in the future.In the foregoing chapters of this paper, experiments are carried out complying with orthogonal test-schema. The slump and slump flow of fresh concrete are measured and the compressive strength of hardened concrete are tested. Then, the effect of the factors (the w/c ratio, sand content by mass of aggregate, the percentage of silica fume and slag) on the self-compacting capability of the mass is analysed. And then, the optimum value of every effect factors is acquired according to the comprehensive-equilibration theory. Based on this, by using P.O. 52.5 cement, SCC with two grades (C40, C50) are mixed successfully. Simultaneously, regressive analysis is carried out between the 28d compressive strength of SCC and the c/w ratio with no variation of other factors. The optimum mix proportions of two grades SCC are recommended.Due to the diversity and complexity of the SCC ingredients which influence each other, it is difficult to describe the relationship between the workability of fresh concrete and a single mixing effect factor, so as the mechanical properties of hardened concrete. In the subsequent chapter, the properties of fresh or hardened concrete are predicted by using neural network that can be used to carry out nonlinear regression with multi-factors. Firstly, a predicting neural network model is founded with the mass of composing-materials as input and the properties of fresh or hardened concrete as output. The nonlinear relationship is mastered by the model after the algorithm was selected and the model was trained with abundant samples. Then the model is used to predict the outcome of some test that have been done formerly. It shows that the artificial neural network can be used as a new approach to predict the performance of SCC and it will be helpful in proportioning the SCC mixtures in the near future.