A Novel Prediction Method For Complex Geological Settlement Based On Improved Bp Neual Network-Taking Groundwater Resources Environmental Protection As An Example

The settlement of engineering geology directly affects the quality of groundwater resources, and then affects the water environment. However, at present, the influencing factors of geological subsidence are very complex, and the geological data is highly dispersed, which leads to a slow convergence rate of geological subsidence prediction and a large prediction error. In order to overcome this deficiency, in this paper, we proposed a complex geological settlement prediction method based on improved BP neural network. Momentum terms are introduced to improve BP neural network weights and learning efficiency. The factors that affect the settlement of the foundation are selected and input into the improved BP neural network and normalized. This method accelerates the convergence speed of the neural network during operation. By constructing a neural network model of geological settlement time series, complex geological settlement prediction is finally achieved. The experimental analysis results of engineering examples show that the proposed method has higher prediction accuracy of complex geological settlement and can reflect the nonlinear characteristics of complex geology. Under the effect of large loads, the prediction error gradually decreases. The test verifies that the prediction method is ideally applied, and it can provide a reliable reference for the protection of groundwater resources.