基于多源信息融合的发动机转子早期故障识别

Applying Method of Multi-Source Information Fusion to Achieving Early Diagnosis of Aero-Engine Rotor Fault

提出了将神经网络与D-S(Dempster-Shafer)证据理论相结合的发动机转子早期故障分级融合识别方法。文中对多源信息融合系统的基本结构、多源信息融合方法、早期故障融合识别过程等进行了分析和研究,并以发动机转子早期碰摩为对象进行了实验验证。结果表明,将神经网络与D-S证据理论相结合的早期故障分级融合识别方法,能够有效地提高发动机转子早期故障识别的快速性和有效性,利用神经网络的输出构造D-S融合推理中各焦元的基本概率赋值函数,避免了构造基本概率赋值函数时人为因素的影响,提高了故障识别精度。

Fig. 1 in the full paper shows the block diagram of our idea on how to achieve the early diagnosis of engine rotor fault. Section 2 proposes our information fusion method that combines neural network with the D-S （Dempster-Shafer） evidence theory to quickly diagnose early faults. In Fig. 2 of section 2, we give the block diagram of the process of the resolution of signals and their reconfiguration, such resolution and reconfiguration being the core of our method. In section 3, we did experiments on the experimental platform of an aero-engine rotor and measured the experimental data of six types of early fault, the measurement data being given in Table 1 and the diagnosis results being given in Table 2. On the basis of Tables 1 and 2, section 3 gives preliminarily three conclusions.