Abstract: Aiming at the problem that Dempster-Shafer (D-S) theory was not effectively solve the problem existing in the combination of highly conflicting evidence,by setting the weight of evidence and constructing the distance function,the optimal weighted value of evidence was obtained from the evidence source with improved Particle Swarm Optimization (PSO) algorithm,thereby the fusion performance was improved.Due to the defects of traditional PSO which were early maturity and low efficiency in later iterations,a new self-adaptive function was constructed and a self-adaptive algorithm integrated by improved PSO and D-S theory was proposed,which realized a double-optimized,dual combination fusion method.Simulation experiment results indicated that this method could self-adaptively acquire the optimized weighted value of evidence source,and effectively overcome the conflict existing in the multi-evidence synthesis.Compared with other PSO algorithm,this method applied a better performance.As the method was applied to multi-Agent intelligent diagnosis of the command and control system for the first time,better fault diagnosis results was acquired.

Key words: dempster-shafer theory, particle swarm optimization, self-adaptive weight, multi-Agent, fault diagnosis

摘要： 针对D-S证据理论在融合决策中不能有效解决高度冲突证据的合成问题,从证据源出发设置证据权值,并构造证据权值距离函数,利用改进的粒子群算法对其进行优化获取最优权值,从而改进融合性能。构建一种新型自适应函数,提出改进的自适应权重粒子群优化算法与D-S综合算法(AWIPSO-DS),实现一种“双优化,双结合”的融合方法。仿真实验表明,该方法能自适应获取最优证据源权值,有效克服多证据冲突合成问题,与其他改进的粒子群算法相比具有良好的适用性能。首次将该方法运用于指挥控制装备多Agent智能诊断系统上,具有较好的故障诊断效果。

关键词: D-S证据理论, 改进的粒子群算法, 自适应权重, 多Agent, 故障诊断