论文部分内容阅读
使用支持向量机(SVM)研究涡轮气封减压试验系统中高压卸荷膜片式减压器的稳定性问题,主要集中于以往方法不易涉及的多结构参数变化.针对稀疏易有残缺的小样本空间,与BP(back propagation)神经网络模型进行对比,得出SVM方法在所研究数据集上的一些结论:SVM模型预测性能在多结构参数变化情形下优于BP神经网络模型,预测误差平均降低了25.5%;SVM的泛化性好于BP;在双参数、三参数情形下,SVM模型为气体减压器的设计提供了更好的决策支持,给出了优化结构参数的设计建议.
The support vector machine (SVM) is used to study the stability of the high pressure unloading diaphragm pressure reducer in the gas turbine pressure reduction test system, which mainly focuses on the changes of the multi-structure parameters which are not easily involved in the past methods. Sample space and BP (back propagation) neural network model are compared to draw some conclusions SVM method in the data set studied: SVM model prediction performance is better than the BP neural network model in the case of changes in multiple structural parameters, the prediction error average Which is 25.5% lower than that of the traditional SVM. The generalization of SVM is better than that of BP. Under the condition of two-parameter and three-parameter, the SVM model provides better decision support for the design of gas pressure reducer and gives the design suggestion for optimizing the structural parameters.