基于故障敏感度的证据权重计算方法及其应用

发布时间：2018-07-17 02:13

【摘要】：典型证据权重计算方法存在只有少数传感器判断正确而多数判断错误的高冲突证据的加权D-S决策融合问题,针对此问题,提出一种基于故障敏感度的证据权重计算方法。首先,通过核函数主元分析(kernel principal component analysis,简称KPCA)提取非线性的敏感特征;其次,基于故障检测原理计算该特征的故障敏感度,并将其作为传感器的故障敏感度;最后,计算得到基于故障敏感度的传感器决策权重,并将该权重及等权重法和基于决策距离方法的权重共同应用于转子故障模拟实验台的融合检测与诊断中。结果表明,该方法能对故障敏感、包含故障信息多的传感器赋予更高的权重值,提高其决策地位和作用,反之则赋予较小的权重,"弱化"其决策地位和作用。通过证据权重的"调节"作用,使得该方法无论是在只有少数传感器发现故障的证据高冲突情况还是在冲突不大或无冲突时,均取得了更好的决策融合结果。
[Abstract]:There exists a weighted D-S decision fusion problem in a typical evidence weight calculation method, in which only a few sensors are correct and the majority of the evidence is wrong. In view of this problem, a method for calculating the weight of evidence based on fault sensitivity is proposed. Firstly, the nonlinear sensitive feature is extracted by kernel function principal component analysis (kernel principal component analysis,). Secondly, the fault sensitivity of the feature is calculated based on the principle of fault detection, and it is regarded as the fault sensitivity of sensor. The weight of sensor decision weight based on fault sensitivity is calculated, and the weights based on equal weight method and decision distance method are applied to the fusion detection and diagnosis of rotor fault simulation test bench. The results show that the method can give higher weight value to sensor with more fault information and improve its decision-making position and function. Otherwise, it can give smaller weight and "weaken" its decision-making position and function. By "adjusting" the weight of the evidence, the method can obtain better results of decision fusion, whether in the case of high conflict of evidence found by only a few sensors or when the conflict is small or no conflict.
【作者单位】：空军工程大学航空航天工程学院;先进航空发动机协同创新中心;
【基金】：国家自然科学基金资助项目(51105374) 航空科学基金资助项目(20142196019) 陕西省自然科学基础研究计划资助项目(2015JM5207)
【分类号】：TH17

【参考文献】