基于故障概率的联邦滤波鲁棒信息分配方法

考虑到故障隔离后故障信息的影响一般需要一段时间才能够消除,研究了当某个子系统发生传感器故障时,其他具有不同信息分配系数的无故障子系统的鲁棒性问题.基于理论分析提出了一种鲁棒信息分配思想,并指出了该思想应当满足的信息分配系数关系.然后给出了一个信息分配系数的专家自适应调整方法,即基于子系统在不同环境下可能发生的故障概率,建立相应的调整规则进行信息分配.从概率的角度,根据这种方法可使其他无故障子滤波器具有较强的抗干扰能力,从而使重构的联邦滤渡器快速输出不含故障信息的融合结果.理论分析和仿真说明该方法是可行的.

Robust information sharing scheme for federated filter based on fault probability

Considering that the fault information's influence on the NFF(new federated filter) regenerated by the GLFS(good local filters) cannot be removed after fault isolation until a period of time later,the robustness problem of information sharing of the federated filter using different information sharing schemes is investigated.By surveying how the information coefficients influence the robustness of the local filters,a robust information sharing idea is presented,and a necessary condition that this idea needs to satisfy is given.Then an adaptive expert information sharing method is presented based on the probabilities of sensor failures which can occur in each local filter under different operating conditions.From the viewpoint of probability,this method can make the other good local filters achieve a high resistance to the contamination from the faulty local filter before the failures are detected.Thus the NFF regenerated by the GLFS may promptly form a globally optimal solution not containing the fault information.Theoretical analysis and simulation show this idea is effective.