基于高斯过程回归的橡胶玻璃化温度的预测研究

Study on Prediction of Glass Transition Temperature of Rubber Based on Gaussian Process Regression

以实例数据为样本,利用高斯过程回归分析苯乙烯和丁二烯含量对溶聚丁苯橡胶(SSBR)玻璃化温度(T_(g))的影响,从而预测SSBR的T_(g)。结果表明:高斯过程回归建立的SSBR的T_(g)预测模型可靠和有效;与反向传播神经网络模型相比,高斯过程回归模型解决小样本问题具有优越性,可为更多复杂耗时的试验数据预测提供有效解决方案,对一定范围内的苯乙烯和丁二烯含量对SSBR的T_(g)的影响进行定性和定量分析。

In this study,the influence of styrene and butadiene contents on the glass transition temperature(T_(g))of solution-polymerized styrene-butadiene rubber(SSBR)was analyzed by using Gaussian process regression,and the method to predict the T_(g) of SSBR was established.The results showed that,the T_(g) prediction model of SSBR established by Gaussian process regression was feasible and effective.Compared with the back propagation(BP)neural network model,the Gaussian process regression model had advantages in solving the small sample problem,and could provide an effective solution for more complex and timeconsuming test data prediction.It was effective in the qualitative and quantitative analysis of the influence of styrene and butadiene content in a certain range on the T_(g) of SSBR.