基于支持向量机的室内轰燃预测模型研究

Study on Predicting Model for Indoor Flashover Based on Support Vector Machine

为对室内轰燃进行准确预测,针对室内轰燃样本的不足在一定程度上制约了其应用,为此运用SVM技术构建室内轰燃预测的数学模型。在小样本条件下,应用工具软件LIBSVM进行仿真,并将SVM模型预测结果和人工神经网络预测结果进行对比。结果显示,SVM技术能较好地解决小样本和模型预测精确度之间的矛盾,SVM模型其预测精度及可行性高于神经网络模型。实例表明,由于室内火灾受多种因素影响,传统的预测方法存在一定的局限性,而SVM模型预测法预测的结果与试验结果比较一致。

In order to accurately predict flashover,based on the fact that insufficient samples restrict the knowledge-based method to some extent on predicting the flashover,a mathematical model for predicting indoor flashover was built using SVM technology.Under the condition of only a small quantity of samples,the application tool software LIBSVM was used to simulate the flashover prediction model and artificial neural network prediction model.The result shows that the SVM model can solve the contradiction of precision between small samples and prediction models with a higher accuracy and feasibility than neural network prediction model.The application example demonstrates that the traditional prediction methods have certain limitations in the application of predicting flashover due to many factors affecting indoor fires,but the SVM model can get a good prediction result which is relatively in accord with the experimental results.