基于BP网络的隔震初步设计系统应用研究

Application Study on Isolation Structure Initial Design Based on BP Neural Network

利用已有隔震工程设计资料,以BP网络建立了隔震初步设计系统对隔震初步设计结果进行预测。其中BP网络是在前人所做工作基础上进行了改进,主要是输入部分增加了总高度和是否有地下室,输出部分变成直接求出水平向减震系数和隔震支座最大位移。经过改进的BP网络预测结果和实际设计结果的对比分析表明,预测结果的误差控制在一定的范围内可用于实际工程应用。所以该神经网络模型可为隔震设计人员提供一种快速、有效的隔震初步设计方法。

Based on the existing isolation engineering design data, the BP network is used to isolation preliminary design to predict the preliminary design of the isolation results. BP neural network is improved on the basis of previous work , including adding the overall height and whether the basement in the input part, the horizontal damping coefficient and the maximum displacement of the isolation bearing can be obtained directly in the output part. The prediction resuhs of the improved BP network are compared with the actual design results. The results show that if the error controlled to a certain range, the improved BP network can be used in practical engineering applications. So the neural network model for the isolation design provides a fast, effective isolation preliminary design method.