标准BP神经网络算法和附加动量法在沉降监测中的应用研究

在沉降监测工程实践中,由于采用的预测方法和项目沉降趋势的不同,预测结果的精度会有差异。而随着应用的不断深入,对现有算法进行改进以发挥算法的优势,成为目前预测算法研究的主流。在本文的研究中,编程实现了标准BP神经网络算法和附加动量法。通过研究学习率对标准BP神经网络算法的影响,确定学习率的大小。然后,研究了动量项对附加动量法收敛速度的影响,确定了动量项的取值。最后,对两种算法的稳定性和算法的效率以及预测精度等方面进行比较,探究两种算法的特点。

Research on Application of Standard BP Neural Network Algorithm and Back-propagation With Momentum in the Practice of Engineering Settlement Monitoring

In the practice of engineering settlement monitoring,according to the difference of forecast methods and characteristics of the settlement trend,the accuracy of the prediction results will be different. With the deepening applica-tion,improving available algorithms to make the advantages of each algorithm a difference,has become the mainstream of research. In this paper,author implemented the standard BP neural network algorithm and back-propagation with mo-mentum. By researching the effect of different learning rates for the standard BP neural network algorithm to determine the size of the learning rate. Then,author studied the effect of momentum on convergence speed of back-propagation with momentum,to determine the value of momentum. Finally,comparing the stability and efficiency of algorithms and forecast accuracy to explore the characteristics of the two algorithms.