BP Network Based Mix Proportion Design of Self-Compacting Concrete

It was known that many parameters of raw materials, such as, strength of cement, mud content and modulus of fineness of river sand, maximum size of aggregate, content of´ needle-like/sheet-like crushed stone, loss of ignition and fineness of fly ash, may exert significant influence on the theology and mechanical properties of self compacting concrete(SCC). It is a dream of researchers to identify the influencing degree of various factors on performance of SCC so as to obtain optimal properties. By virtue of BP neural network approach, this paper employed strength of cement, mud content and fineness modulus of fineness of river sand, maximum size of aggregate, content of needle-like/sheet-like crushed stone, loss of ignition and fineness of fly ash as the input parameters, and the corresponding optimized mix proportion as the output to describe the nonlinear relationship between them. And the orthogonal experiment was designed for the purpose of training and verification of network. The results demonstrated that the pre-trained BP neural network trained by orthogonal test data may employ to predict the optimal concrete mix proportion. This approach may replace some waste-time and heavy laboratory tests. In addition, such method may real-time optimize mixture proportion. of self-compacting concrete, which has great effect on the quality control of manufacturing self-compacting concrete.