Background: Crystal oscillator plays a major role in the RF system as the frequency source
while often encountered in vibration condition. In order to maintain the normal operation of the system,
implementing estimation and prediction is necessary as a strategy based on performance degradation.
However, there is little research addressing the issues associated with health status evaluation and
Residual Useful Life (RUL) prediction of the RF crystal oscillator.
Methods: This paper proposes a method to assess the health status and the RUL tested in a case study
under vibration condition.
Conclusion: Our approach utilizes acceleration sensitivity(AS) as indicator and the M-estimation based
optimally-pruned extreme learning machine (M-OPELM) as technology, obtaining ideal predicting
results according to the actual RUL.