基于混沌吸引子的飞轮故障检测

针对飞轮早期故障难以检测、精确数学模型难以建立的问题，提出一种基于混沌吸引子特征的故障检测方法。该方法利用辅助曲面函数与系统参量构造离散动力系统，通过迭代产生近似混沌吸引子，正常数据与故障数据所产生的混沌吸引子形态不同，以此为特征进行故障检测。仿真结果表明，该方法构造的离散动力系统能够稳定地产生混沌吸引子；产生的混沌吸引子与初始迭代点无关；同种故障在不同工况下的特征相同；混沌吸引子特征对微小幅度的故障敏感。

Fault detection for flywheels based on chaotic attractor

Aimed at the problem that the early fault of the flywheel is difficult to detect and the precision mathematical model is difficult to be established, a fault detection method based on the characteristics of chaotic attractor is proposed. This method uses the auxiliary curved surface function and the system parameters to construct the discrete dynamical system. The approximate chaotic attractors obtained from normal data through iteration are different with the ones obtained from fault data. The difference could be used as feature for fault detection. The simulation results show that the discrete dynamical system constructed by this method can generate the chaotic attractor stably. The chaotic attractor is independent of the initial iteration point. The same faults under different working conditions have the same characteristics. The chaotic attractor feature is sensitive to small fault.