Improved forecasting model to estimate stability of rock slope

The stability of slope is significant in civil engineering, and the structure and physical and mechanical properties of slope rock hold characteristics of macroscopic and microscopic discontinuity, high nonlinearity. The stability of slope is greatly affected by geological structure and its construction. However, these factors are random, fuzzy and changeable, so the slope is an uncertain, nonlinear, dynamic and complicated system; and it is difficult to describe such nonlinear characteristics of this system with traditional methods. Therefore, the stability of large and complicated rock slopes could not be accurately forecasted. A novel forecasting method, which is an alternation and iterative algorithm based on simulated annealing applied in neural network (NN) is presented. Under the same initial conditions, the comparison of the new method with the traditional NN algorithm is conducted; and the result shows the superiority and efficiency of the former. Based on comprehensively analyzing the mechanism of stability loss of rock slopes and the main factors affecting the stability of rock slopes, the compound indices were proposed for forecasting model as influencing factors. Then, the cases of rock slopes in hydropower projects are taken as training samples and the unlearning cases are forcast; and the results is satisfying, showing that the forecasting accuracy is superior to traditional BP neural network. Therefore, the forecasting model put forward here can get safety factor of different rock slopes quickly and accurately; and it can provide a new approach for selecting slope design economically and rationally.