基于遗传算法的旋转机械故障诊断方法融合

ROTATING MACHINERY FAULT DIAGNOSIS COMBINATION OF METHOD BASED ON GENETIC ALGORITHM

针对任何单一性质故障特征、单一诊断方法难以实现在整个故障状态空间上准确诊断的局限性,提出基于遗传算法的旋转机械融合诊断方法。该方法能有效利用各种不同性质故障特征和不同诊断方法,使其发挥各自的优点,从而提高诊断的准确率。针对不同特征利用遗传算法将神经网络诊断和人工免疫诊断方法融合起来,使每一个诊断方法都在其优势空间区域发挥作用,使用小波包能量特征和双谱特征对两种诊断方法训练后,用遗传算法优化诊断融合权值矩阵对旋转机械进行实例诊断结果表明,该融合诊断方法能有效地提高故障诊断的准确率,并能提高诊断系统的鲁棒性。

The combination of fault diagnosis methods based on genetic algorithm for rotating machinery is presented, as there exists limitness for any single fault feature, any single diagnosis method to achieve the accurate diagnosis needs the whole diagnosis state area. This method can effectively use diversified different fault character and diagnosis methods that can present their advantage respectively, so that the diagnosis accuracy is improved. Neural network diagnosis method and artificial immune system diagnosis method are combined by using genetic algorithm. Two different characters, Wavelet Packet＂ energy＂ character and Bispectrum character, are used. After training the two fault diagnosis methods, the genetic algorithm is used to optimize diagnosis combination weight matrix. It is demonstrated from the diagnosis example of rotating machinery that the combination diagnosis method can improve the accuracy rate and diagnosis system robust quality effectively.