神经网络与灰色理论联合模型在地铁沉降预测中的应用

Application of Conjunctive Model of BP Neural Network and Gray Theory Application in Subway Settlement Prediction

地铁路基的过量沉降或不均匀沉降将导致线路运营条件的恶化,乘客舒适度降低,甚至危及行车安全。因此对路基工程后期的沉降控制和预测随着运营速度的提高而愈加急迫。根据某段地铁线路路基的实际沉降观测数据,将神经网络与灰色系统进行串联型结合:即先利用BP神经网络插值方法将不等时距的实测沉降数据序列转化为等时距数据序列,进而利用转化的等时距沉降序列依据灰色GM(1,1)模型对荷载稳定时间内的路基沉降进行预测。实验结果表明,该方法具有较高预测精度。

Excessive settlement and unequal settlement of subway subgrade will cause deterioration of line operating conditions, passengers comfort level reduction, and even endanger the traffic safety. So with the operating speed increased, it needs the control and prediction for the later period of subgrade engineering immediately. According to the accrual settlement observation data of subway line subgrade in some area, series-type combination the BP neural network and gray theory, that is, first, the BP neural network interpolation method is used to transform the measured data sequence of unequal interval to equal time interval data series, second, according to gray GM （1, 1） model, the transformation of equal time interval settlement sequence is used to predict subgrade settlement during stable time of load. The result indicates that this method has high prediction precision.