基于D-S证据理论的航空发动机振动故障分析

摘要

Aeroengine has complex structure harsh work conditions, the effective analysis on its vibration is an important means for fault diagnosis. Since different feature value gives different capability in vibration analysis, in order to integrate analyses results obtained in different feature items, the Dempster-Shafer evidential theory-based information fusing method is adopted to fuse the diagnosis results of BP neural networks with different features. Furthermore, aiming at the practical working status of aeroengine, a belief function construction approach using output statistics of neural networks is put forward. Through the experimental analysis results of the vibration signals observed during the trial running of the aeroengine, it shows that this algorithm can effectively improve the precision rate of aeroengine vibration fault recognition.%航空发动机结构复杂且工作条件恶劣,对其振动的有效分析是进行故障诊断的重要手段.由于不同特征量对振动具有不同的分析能力,为了综合利用不同特征项下的分析结果,采用基于D-S证据理论的信息融合方法对不同特征下的BP神经网络的诊断结果进行融合,并针对航空发动机实际工作状况提出一种利用神经网络的输出统计值构造信度函数的方法.通过对实测航空发动机试车时振动信号的实验分析结果表明,该算法可以有效地提高航空发动机振动故障识别的准确率.