支持向量机回归在电子器件易损性评估中的应用

针对现有的以概率统计理论为基础的方法和模糊神经网络法必须建立在大量统计数据基础之上，以及模糊信息扩散估计法可能对器件失效阈值估计过高的问题，提出将模糊信息处理技术用于对原始实验数据的处理，得到训练样本，在此基础上利用支持向量机回归预测一定功率的高功率微波辐照条件下电子器件的损伤概率。仿真结果表明:该方法与模糊神经网络法都较好地给出了预测结果，但该方法具有更高的精度（均方根误差为7.40610-5），并且克服了在样本数据减半的小样本情况下模糊神经网络法可能出现野值的缺陷。

Application of support vector regression to vulnerability assessment of electronic devices illuminated or injected by high power microwave

Aiming at the problems that the existing methods based on the probability statistical theory and the fuzzy neural network method must be built on the foundation of large quantities of statistical data, and the failure thresholds of electronic devices estimated by the fuzzy information diffusion could be higher than the actual ones, a new method is presented that the raw experimental data are processed by the fuzzy information processing technology to obtain the training samples, on the basis of which the damage probabilities of electronic devices illuminated or injected by the high power microwave are predicted by support vector regression. The simulation results show that the fuzzy neural network and the new method both achieve good prediction results. But the results of the latter are more accurate and it overcomes the defect that errors could occur in the results predicted by the fuzzy neural network under the condition of small samples.