神经网络和SVM多传感器融合的隧道CO体积分数研究

利用BP,RBF神经网络、支持向量机(SVM)不同信息融合模型对高速公路隧道中失效CO传感器数据融合研究,比较了单一融合模型融合效果和不同最优加权信息融合模型融合效果。仿真实验表明:3种单一融合模型对隧道CO体积分数融合的有效性,其中单一SVM融合模型效果最好。最优加权融合性能均优于单一融合模型,其中,BP与SVM最优加权融合精度最高。还分析单一融合模型输出之间的冗余度对最优加权融合精度的影响规律,该规律为参与最优加权融合的单一模型筛选提供了一种新方法。

Research on tunnel CO volume fraction of neural network and SVM multi-sensor fusion

To compare effects of the single integration model fusion,and the different optimal weighting information fusion model effects through the study by information fusion model of BP,RBF neural networks, support vector machine(SVM) integration on the failure CO sensor data fusion in the highway tunnel.The simulation test results show that the effectiveness of integration of three kinds of single-fusion model on tunnel CO volume fraction,among which the single SVM fusion model is the best.Optimal weighted fusion performance is superior to a single fusion model,in which BP and SVM,the optimal weighted fusion has the highest precision. The effect rule of the redundancy of precision of single fusion precision between the model outputs on optimal weighted fusion precision is also analyzed.The rule provides a new way to select single models of the optimal weighted fusion.