基于BP神经网络技术的实验数据分析处理

许多热灾害实验的危险性和成本相对较高,很难进行多次实验,且数据较少。针对这些的问题,笔者根据BP神经网络基本原理,建立多层反向传播的神经网络,以现有实验数据为训练样本,对神经网络进行训练,利用训练后的网络对待测样本进行预测。以喷射火作用下液化气储罐热响应实验的数据处理为例,利用BP神经网络进行压力值预测。将预测的结果和实验数据进行比较,结果表明了神经网络对实验数据进行处理和预测的可行性。

Process and Analysis of Experiment Data Based on BP Neural Network Technique

The risk and cast of thermal hazard experiments are high, repeated experiments were difficult to be carried out, so the experiment data were rather rare. Based of the basic principle of BP neural network, the multi-layers back propagation neural network was established and trained by the real experiment data. The Wained network was used to predict the eases which should be tested. Taking the process and analysis of experiment data from the thermal response of liquefied gas tanks subjected to jet fire as an example, the pressure values were predicted by the BP neural network Comparison of the predicted results with the experiment data proved that the feasibility of the method.