Abstract. The evaluation model of university laboratory safety based on BP neural network is presented. By reference of the domestic research for laboratory safety evaluation and combining with of laboratory characteristics of Jinlin University, the safety evaluation index system of university laboratory is selected． This model quantifies the evaluation index of laboratory safety into definite data as the input for BP network, and taking it output as the evaluation results. The BP network is trained by Trainlm function with training samples which have been obtained by experts, and then it is simulated with testing data. The experiments show that the error between the training value of network simulation and the actual evaluation results is very small, then the applicability of the evaluation model is illustrated. The existing information of laboratory safety evaluation with relatively successful cases can be used in the model, hence the experience of experts can be accumulated for providing more scientific and quantitative criteria for the laboratory safety evaluation．

To solve the problem that the accure data can't be pushed by the failure of local sensor in intelligence greenhouse system, it was presented that the Apriori algorithm which was based on association rule applied in the prediction of sensor fault data. Forcasting the greenhouse environment temperature is provided as an example in this paper, firstly, the classic Apriori algorithm is modified. Then it was used in the prediction of fault sensor data. The experimental results show that the improved Apriori algorithem could quickly find the association rule between the parameters in Greenhouse, thus estimated the range of the parameters of the fault sensor and the method could be proved to be feasible.