基于人工神经网络的基坑钢支撑轴力预测研究

Research on Prediction of Steel Support Axial Force of Excavation Based on Artificial Neural Network

以温度、湿度、风速、升降温速率、阳光照射时间和阳光直射钢支撑面积作为神经网络的输入因子,钢支撑轴力作为输出因子,基于BP神经网络,建立深基坑非施工期钢支撑轴力预测模型,利用Matlab软件对神经网络进行训练,并结合实测值对钢支撑轴力值进行检验。结果表明:外界环境温度对支撑轴力影响呈现"M"式变化前进的规律;其神经网络预测模型回归分析线性相关系数都达到0.75以上;通过对预测轴力值与实测值的综合对比分析,测试样本预测相对误差最大值为0.183,相对误差最小值为0,预测的结果基本符合现场钢支撑轴力的变化趋势;该模型程序具有适应能力强、预测准确度高、操作简便等优点。该钢支撑轴力预测研究为非施工期基坑安全问题提供新的参考依据。

Non-construction period prediction model based on BP neural network set up.In the BP neural network mode,temperature,humidity,wind speed,rising and cooling rate,sun exposure time and direct sunlight steel support area taken as input data,the steel support axial force taken as output datasoftware used to train the BP neural network and then steel support axial force tested by combining with the measured values.The results show that:External ambient temperature influences the axial force of the support and presents the law of repeated M"progression.The linear correlation coefficient of the results of the regression analysis of the steel supporting its BP neural network prediction model all reach 0.75 or more.Through comprehensive analysis of predicted axial force values and monitored values,the maximum relative error of the test sample prediction is 0.183,and the minimum relative error is 0.The predicted results basically reflect the trend of the steel support axial force.The model program has the characteristics of strong adaptability,high prediction accuracy and easy operation.The prediction of the axial force of the steel support provides a new reference for the safety of foundation pits during non-construction period.