An on-line failure identification method for rotor systems

Unexpected failures of rotor system could cause serious damage or loss to a machine and its application. An effective and reliable failure identification system that can on-line detect or even predict the occurrence of failures, can minimize this potential risks. In this paper, a method composing of signal processing and identification algorithm to diagnosis the three major failures of a rotator system caused by impact forces, abnormal friction, and unbalanced rotation of the system etc. is proposed. Both Wavelet transform and Fast Fourier Transform were used to convert the collected signals. Subsequently, a diagnosis map containing distribution of characteristic values of converted signals was built for failure diagnosis. When dynamic signals of a rotor system is collected and converted, the characteristic value of the converted signal will then be computed and mapping to the diagnosis map. Finally, the possibility of occurrence of a failure mode can be determined by the degree of approach. To verify the proposed methodology, both of simulation and experiment were conducted, and the results have shown high feasibility and reliability of the detection system.