论文部分内容阅读
常规D-S(Dempster-Shafter)决策融合方法由于其自身理论不足,不能很好直接处理决策结果偏差大、冲突大的传感器融合问题,因而对于信息高冲突情况下的转子微弱故障融合诊断存在着失效问题。针对该类问题与不足,借鉴复杂网络的舆论传播、社会学习理论及多智能体一致性决策的相关概念与思路,从避免决策结果冲突大的传感器直接进行融合的角度进行改进,提出相邻模块化加权D-S融合方法。该方法首先根据初步结果进行相邻节点与模块划分,只有决策距离在相邻界限值范围内的相邻模块节点才能进行决策融合;对于同一模块内相邻节点,根据各节点决策权重及初步决策结果采用加权D-S融合方法进行决策融合;针对融合结果再进行相邻节点模块划分与融合,依此步骤进行循环划分与融合,直到所有模块与节点均不相邻;最后采用专家权威决策方法确定权重和最大的模块融合结果作为最终的传感器网络一致性决策结果。通过多传感器网络的转子故障模拟实验对所提方法进行验证,应用结果表明:所提方法可以较好解决少数传感器诊断正确、而多数诊断错误的信息高冲突条件下的局部微弱故障融合诊断问题。
Because of its own theory, conventional DS (Dempster-Shafter) fusion method can not directly deal with sensor fusion problem with big deviation and big conflict in decision-making results, so there is a failure problem in faint fault fusion diagnosis . Aiming at the problems and shortcomings of this kind of problems, this paper uses the concepts and ideas of the public opinion dissemination of complex networks, social learning theory and the multi-agent consensus decision-making to improve the system from the point of view of avoiding the conflict of decision- Weighted DS fusion method. In the method, adjacent nodes and modules are divided according to the preliminary results. Only the adjacent nodes with the decision distance within the range of the adjacent thresholds can make the decision fusion. For the adjacent nodes in the same module, according to the decision weight of each node and the initial decision The results were weighted by the DS fusion method for decision fusion. According to the fusion result, the adjacent node modules were divided and merged again. According to this procedure, the circulation and integration were repeated until all the modules were not adjacent to the nodes. Finally, weights were determined by authoritative decision making method And the largest module fusion results as the final sensor network consistency decision-making results. The proposed method is validated by the rotor fault simulation experiment of multi-sensor networks. The application results show that the proposed method can solve the local faint fault fusion diagnosis problem with few sensors correctly diagnosed, while most of them have high conflict conditions.