Reconstructing Spectra from IVHMS Condition Indicators

The US Army collects high resolution vibration data from accelerometers on over 3000 rotorcraft. This data is used to diagnose the health of the drive system for the purposes of Condition Based Maintenance. This high resolution data is not always captured and saved, however features computed from it are always available. This paper uses a machine learning technique to estimate the high resolution data from the features with negligible error up to 20 kHz. The results indicate that the Condition Indicators represent the spectrums very well, which was not previously demonstrated or assumed.