Early fault detection of rotating machinery through chaotic vibration feature extraction of experimental data sets

Fault detection of rotating machinery by the complex and non-stationary vibration signals with noise is very difficult, especially at the early stages. Also, many failure mechanisms and various adverse operating conditions in rotating machinery involve significant nonlinear dynamical properties. As a novel method, phase space reconstruction is used to study the effect of faults on the chaotic behavior, for the first time. Strange attractors in reconstructed phase space proof the existence of chaotic behavior. To quantify the chaotic vibration for fault diagnosis, a set of new features are extracted. These features include the largest Lyapunov exponent; approximate entropy and correlation dimension which acquire more fault characteristic information. The variations of these features for different healthy/faulty conditions are very good for fault diagnosis and identification. For the first time, a new chaotic feature space is introduced for fault detection, which is made from chaotic behavior features. In this space, different conditions of rotating machinery are separated very well. To obtain more generalized results, the features are introduced into a neural network to identify different faults in rotating machinery. The effectiveness of the new features based on chaotic vibrations is demonstrated by the experimental data sets. The proposed approach can reliably recognize different fault types and have more accurate results. Also, the performance of the new procedure is robust to the variation of load values and shows good generalization capability for various load values.