| CPC G01M 13/00 (2013.01) [G01M 13/028 (2013.01); G01M 13/045 (2013.01); G06F 11/0736 (2013.01); G06F 11/079 (2013.01)] | 7 Claims |
|
1. An apparatus for diagnosing an equipment failure using a sound spectrogram image, the apparatus comprising:
a sensor unit having a fixing gear including a magnet or ring detachable from an equipment, and configured to collect sound data and collection position information of the sound data from one side of the equipment; and
a controller configured to convert the collected sound data into a spectrogram image, perform machine learning based on the converted spectrogram image, a type of the equipment, and the collected collection position information of the sound data, and determine whether a failure occurs in the equipment related to the collected sound data based on a machine learning result,
wherein the controller performs data classification for training by classifying each of a plurality of pieces of previously collected equipment-specific sound data into any one type of normal sound, shaft failure sound, bearing damage sound, cavitation sound, impeller failure sound, and motor bearing sound, converts the plurality of pieces of classified equipment-specific sound data into spectrogram images, divides the plurality of converted equipment-specific spectrogram images into a train set and a test set, and then performs a training and test function for a preset failure diagnosis and failure cause determination model using the divided train set and test set,
the controller performs machine learning using the converted spectrogram image, the type of the equipment, and the collected collection position information of the sound data as input values of the failure diagnosis and failure cause determination model and determines whether the failure occurs in the equipment related to the collected sound data based on a machine learning result, and
when it is determined that the failure occurs in the equipment related to the collected sound data, the controller classifies a cause of the failure as any one of a shaft failure, bearing damage, cavitation, an impeller failure, and a motor bearing failure based on the machine learning result.
|