Autogenous shrinkage prediction on high performance concrete of fly ash based on BP neural network

The article adopts test data of neural network for autogenous shrinkage to train and predict on the data which doesn't join training. The article's prediction is on the basis of common medium sand, 5-31.5mm limestone rubble, second class fly-ash, P.O42.5 silicate cement, considering factors include five ones such as ratio of water and cement, sand rate, content of cement, content of fly ash, etc. By adjusting various parameters of neural network structure, it obtains three optimized results of neural network simulation. The error between concrete autogtenous shrinkage value of neural network prediction and trial value is within 3%, which can meet requirement of the concrete engineering.