基于改进的BP网络的D-S证据理论及其应用

Dempster-Shafer Theory of Evidence Based on Improved BP and Its Application

目前,D-S证据理论凭借自身独特的处理不确定性问题能力,已在众多领域得到了广泛的应用。命题基本概率分配(BPA)的确定是D-S证据理论得以广泛应用的关键之一,将改进的BP网络运用到基本概率分配的确定过程中,使得BP网络和D-S证据理论两者有机地联合应用,这样既可利用D-S证据理论来表达和处理不确定信息,又可以充分发挥BP网络的自学习、自适应和容错能力。建立了基于BP网络的D-S证据理论的故障诊断模型,并给出了证据的融合算法。最后,仿真实验表明该模型可行。

Demmpster-Shafer theory of evidence had good performance in dealing with uncertain information and had been used in many fields.However,basic probability assignment（BPA） was one of the key problems in its application.In this paper,an improved BP was applied to solve such a problem and it was integrated with D-S theory organically to deal with uncertain information fusion.The self-learning and adaptive capability as well as fault-tolerance may come into effect completely.A model for fault diagnosis that using D-S theory of evidence based on BP was discussed in detailed and the fusion algorithm for combing all pieces of evidence was presented.Finally,an illustrative example of circuit fault diagnosis was given to show that the proposed approach was feasible.