基于BP神经网络的膨润土导热性能预测

Prediction of the Thermal Performance of XJ-Bontonite Based on BP Neural Network

将新疆塔城地区天然钠基膨润土的导热性能及其影响因素作为一个系统来研究,在深入分析膨润土在不同干密度、不同含水量、不同孔隙度和不同饱和度条件下的导热性能实验的基础上,基于人工神经网络方法对新疆塔城地区天然钠基膨润土的热传导系数(作为系统输出)及其影响因素(作为系统输入)进行了系统模拟,建立了相应的模拟模型,设计了模拟算法,并用Microsoft Visual Studio为开发工具,研制了"膨润土导热性能模拟软件系统",用该软件系统成功实现了新疆塔城地区天然钠基膨润土导热性能的模拟及预测,结合导热性能实验数据进行软件实证分析,从软件运行的实证数据看,预测效果好、精度高;从控制(输入)的角度研究膨润土热传导热系数(输出)的预测问题,不仅为膨润土导热性能的预测和模拟提供了平台和工具,更重要的是为高放核废物处置库工程屏障开发高导热性能缓冲回填材料奠定了基础。

This paper regards the experimental heat - conducting property of XJ - bentonite and its influence factors as a system to study. After an in - depth analysis on the heat - conducting property of XJ - Bentonite was made under different dry density, water content, porosity and saturation. On the basis of the principle of neural network method, a systematic simulation to the heat - conducting property of XJ - bentonite （as the system output） and its influence factors （as the system input） was conducted and corre-sponding simulation model and designed algorithm were established. Furthermore, “ the simulation software system of the heat - conducting property of XJ - bentonite” was developed by using Microsoft Visual Studio as a developing instrument. With this system, the heat - conducting property of XJ - bentonite sys-tem simulation and prediction in future has been realized successfully. From the results of real experimen-tal data that the simulation software system operates, it can be concluded that this method has good effect and high precision. The research on the heat - conducting property of XJ - bentonite （ as output） in term of control （as input） has provided a platform and tool for prediction and simulation of the heat - conducting property of XJ - bentonite, more importantly it has laid a foundation for the design of highthermal conductivity buffer backfill material and the development of artificial barrier of high - level radioactive waste repository.