Abstract:To grasp the effect of different dosages on the strength of gypsum cementitious material, the gypsum-based composite cementitious materials were prepared by orthogonal test method using building desulfurized gypsum as raw materials, and the BP Neural Network of the main factors affecting 7d compressive strength of desulfurized gypsum block was established. Based on the BP neural network model, the conditions of different influencing factors such as whisker content were optimized. The results show that the effects of various factors on the 7d compressive strength of gypsum block were in descending order cement, slag content, water reducer content, retarder content, neutralization slag content, and whisker content. However, the process parameters optimized by the BP neural network model are that the whisker content is 6.40%, the polycarboxylate superplasticizer content is 1.28%, the ratio of cement to slag content was 1 of 3, the calcination neutralization slag content is 2.00% and the dosage of citric acid retarder is 0.114%. Under this condition, the obtained gypsum-based composite cementitious material which 7d compressive strength is 14.62MPa increased by 2.024% compared with the results of the orthogonal test. And optimization using BP neural network model can reduce the dosage of admixtures to a certain extent, which the dosage of polycarboxylate superplasticizer, whiskers and citric acid retarder decrease by 0.12%, 0.60% and 0.006%, respectively. The research has important reference significance on the recycling of gypsum waste and reducing the cost of mine filling.