Fault diagnosis approach for rotating machinery based on dynamic model and computational intelligence

• Proposes a novel diagnostic approach combing dynamic model and intelligent methods.
• The dynamic model of a rotating machinery is established to produce fault data.
• Improved empirical mode decomposition is developed to extract more fault features.
• Fuzzy support vector machine with new membership is presented to recognize faults.
• Fault diagnosis results validate the accuracy and superiority of the approach.

In order to achieve accurate diagnosis for rotating machinery automatically and considering that test data under actual fault conditions are rather difficult to obtain, a novel fault diagnosis strategy based on rotor dynamics and computational intelligence was proposed in this paper. Considering the nonlinear restoring force of ball bearing, the dynamic equation of a rotor–bearing system containing four typical faults was deduced with lumped mass method. Vibration responses of the system under various conditions of different rotational speeds, fault types and fault degrees were acquired. An alternative empirical mode decomposition (EMD) method improved by wavelet packet decomposition was developed to process the fault signals. Time–frequency characteristics calculated via the improved EMD as well as statistical parameters of the signal in time- and frequency-domains were extracted as fault features. Then, fuzzy support vector machine (FSVM) optimized by multi-population genetic algorithm was adopted to identify the state of the system automatically. Fault diagnosis results validate the effectiveness of the proposed approach as well as its superiority over commonly used support vector machines. The performances of different fault features and the anti-noise capability of the approach were also investigated. Results demonstrate that the proposed approach is very suitable for engineering application owing to its high accuracy and strong robustness.