论文部分内容阅读
实际灌浆压力控制过程中,由于灌浆液的密度、粘度和地层等因素的影响,使得灌浆压力的变化具有不确定性、时变性和非线性特征。为了辨识、预测灌浆系统压力,提出了一种基于神经网络的多传感器数据融合技术。通过对灌浆工艺与机理分析得到该BP神经网络输入变量。该方法首先利用灌浆过程中采集的数据离线训练BP神经网络,获得一收敛的神经网络模型,然后用此神经网络模型实时预测所灌地层的灌浆压力。最后实验仿真结果表明,BP神经网络预测模型能够运用到灌浆系统中,模型的最大预测误差不超过15%,平均均方根误差仅为0.186。
In the process of actual grouting pressure control, the variation of grouting pressure is uncertain, time-varying and non-linear due to the influence of the density, viscosity and formation of grouting fluid. In order to identify and predict the pressure of grouting system, a multi-sensor data fusion technology based on neural network is proposed. The BP neural network input variables were obtained by analyzing the grouting technology and mechanism. In this method, the BP neural network is trained offline using the data collected during the grouting process to obtain a convergent neural network model. The neural network model is then used to predict the grouting pressure in the injected formation in real time. The experimental results show that the BP neural network prediction model can be applied to the grouting system. The maximum prediction error of the model does not exceed 15% and the average root mean square error is only 0.186.