Study Of Gray Neural Network Model Of Enterprise's Safety Devotion

Lack of safety devotion and unscientific devotion structure are important reasons of restricting the level of enterprise's production safety. For enterprises of Shandong province, the security level is relatively high, but also some problems exist to varying degrees. Production safety research, especially in safety devotion, can not only improve the overall level of safety in Shandong province, but also help to improve the utilization rate of limited resources, so as to achieve the purpose of improving steadily safety level.Based on inquisition data of Shandong province's enterprises by Shandong province safety produce surveillance management bureau, the safety production situation was analyzed, the trends of various safety devotion and accidents were researched. Absolute numerical of various safety devotion were maintained sustainable growth, but their's increment speed all below production value's speed commonly. Accidents Poisson distribution function and casualties Poisson distribution functions were got. The relationship between production safety devotion and accidents were studied and function relationships were simulated by using least squares curve fitting tool.According to the characteristics of enterprise safe devotion, the model of grey system theory combined with neural network was put forward to establish enterprise safety devotion model, so the model can solve the problems about data volatility and large data modeling. Result of the model were distribution data about update transformation devotion, security measures devotion, safety technology devotion, labor protection devotion, occupational disease devotion, property insurance premium, transportation insurance premium and personal insurance premium in premise of production value goal, millions of man-hour casualties and millions of man-hour accidents goal. When determined the structure of neural network, error results of different hidden nodes and three different training function of the learning process were compared to get the best neural network structure. Error inspection was done and compared. Application of this model was done to predict safety devotion finally.