Application Of Fractal Theory In Accident Forecasting

The accident forecasting is an important content of the safety management. The purpose is to grasp the tendency of accidents under existing conditions, control the factors affecting accidents, and reduce the accidents. It is difficult to forecast accidents because of the characteristics of accidents such as nonlinearity, randomness and uncertainty. Fractal theory is a non-linear system theory which has been developed in recent years. The paper applies fractal theory and methods to forecast accidents and obtains certain accomplishment. The main conclusions are as follows:(1) Firstly, based on fractal theory , the paper concretely analyzes the fractal characteristic of the mine accident time series from July 2000 to April 2008 and calculates the correlation dimension and Kolmogorov entropy which are 5.292 and 0.0532 according to the technique of phase space reconstruction and G-P arithmetic. The results indicate that the accident time series has fractal characteristic. A relatively low extent of chaos explains that the mine security of our country is in a relatively stably state in recent years.(2) Secondly, the paper estimates the H index which is 0.95 through the Rescaled Range Analysis(R/S analysis). It indicates that the accident time series has positive long-rang correlation, shows stronger continuity and follows biased random walk.(3) At last, the paper builds two forecasting models. One is based on fractal theory, the other is based on grey theory. The model based on fractal theory was used to check the accuracy. The mean relative error is 4.6%. The model has highly forecasting precision. In this foundation, the model was used to forecast future data which are 66, 55, and 84. But the model based on grey theory does not meet the accuracy requirement. That is impossible to forecast. The results show that fractal theory is applied to the accident forecasting practicality and feasibility.