Abstract:
Based on the facts that BP neural networks have some flaws, such as characteristics of low efficiency and low accuracy. A new BP neural networks was put forward to forecast the surface settlement of pits, which depend on Simulated Annealing. The kernel functions was determined by the theory of optimization of overall situation, which could improve the efficiency and accuracy of BP neural networks. According to the typical pits excavation cases in Shanghai, the SA-BP neural networks that based on Simulated Annealing was built to forecast the surface settlement of pits. And the result was juxtaposed with normal distribution and skew distribution and the results of statistical analysis. The result shown that the results obtained by SA-BP neural networks were closer to reality than others, the error was low. And the method can be used to analysis and forecast the surface settlement of pits.