航空发动机诊断知识的动态获取与柔性诊断技术

Knowledge Dynamic Acquisition and Flexible Diagnosis Technique of Aero-Engines

针对军用航空发动机的状态监测与故障诊断问题,研究了航空发动机的诊断知识动态获取模型及柔性诊断技术。建立了可扩展诊断样本库,实现样本库中故障征兆和故障模式的动态增减,以增加系统的柔性和可扩展性;运用粗糙集理论对样本集进行处理,实现冗余属性的约简、冗余样本的去除及样本冲突的消除;用神经网络通过对处理后的样本集进行学习以动态获取知识,将实际诊断样本输入到训练好的神经网络模型即可得到诊断结果。整个诊断过程具有充分的可扩展性和柔性,当有新样本加入时,按上述步骤进行处理即可实现诊断知识的动态获取和诊断。算例表明了方法的正确性和有效性。

Aimed at the military aircraft aero--engine condition monitoring and fault diagnosis, the aero engine diagnosis knowledge dynamic acquisition model and flexible diagnosis technique were studied. Firstly, a sample database which can be extended was established to accommodate the dynamic change of fault symptoms and fault modes, and flexibility of diagnosis system was increased greatly. Secondly, the sample set was pre processed by Rough Sets （RS） theory, and the redundancies of samples＇ attributes and redundant samples were removed. Thirdly, Artificial Neural Network （ANN） was trained by the processed samples, and diagnosis knowledge was obtained. Finally, a practical diagnosis sample was input into trained ANN, and the results can be obtained. Whole diagnosis process has full flexibility and adaptability. When a new sample is added, above mentioned steps are repeated, and a new knowledge can be dynamically obtained. In the end, an example is used to verify the validity and correctness of the diagnosis model.