基于非参数估计的决策融合模型及其应用研究

针对传统bayes决策融合算法中,类条件概率密度的不确定性以及固定的类条件概率密度估计值所引起的融合系统误判,本文将非参数估计方法引入到bayes决策融合算法中,提出了一种改进的bayes决策融合模型。该模型在融合过程中每次进行bayes最大后验概率准则判决后,均判断当前的类条件概率密度是否达到预期的精度,并在未达到预期精度的情况下利用parzen窗算法完成条件概率密度的逐步构造逼近。通过在丰满水电仿真系统的温度故障诊断过程中的实际应用表明,该算法在性能上优于传统bayes方法,可有效地提高诊断判决正确率。

The Research of Decision Fusion Model Based on the Non-Parameter Estimation and Its Application

In the traditional Bayes decision fusion algorithm, fault decision of the fusion system often occurred because of the uncertainty or fixed value of conditional probability density. In this paper, a new improved Bayes fusion model, which introduced non-parameter estimation method into traditional Bayes fusion criterion, was proposed. In the improved model, judgment about the precision of conditional probability density was made after the decision based on Bayes maximum a posteriori criterion in the process of fusion, and conditional probability density was modified by the Parzen window method when the precision was lower than expected value. In the practical application, the method had been successfully applied in the temperature fault diagnosis system of Hydroelectric Simulation System of Jilin Fengman, the performance preceded that of the traditional Bayes criterion, and the correct ratio of diagnosis result had been efficiently improved.