Granted Patent B1

US 11,348,599 · App. 17/155,279 · Granted May 31, 2022

Inventors: Deepak Gowda (Cary, NC); Sathish Kumar Bikumala (Round Rock, TX)

Assignee: Dell Products L.P.

G10L25/51G06N3/08G10L25/30G11B27/36H04R1/08

Abstract

Claims (84)

1. A computer-implemented method comprising:

recording digital audio corresponding to audio detected by an audio transducer in proximity to moving mechanical components of a hard disk drive; and

determining a health status of the hard disk drive based on a classification of the recorded digital audio, wherein the recorded digital audio is classified using a trained machine learning model, the determining comprising

providing the digital audio to an input of an audio feature extractor to generate one or more audio feature vectors, wherein the audio feature extractor includes a neural network; and

providing the one or more audio feature vectors to a hard disk drive health status classifier, wherein the hard disk drive health status classifier provides an output classification corresponding to a health status of the hard disk drive based on the one or more audio feature vectors, wherein the hard disk drive health status classifier includes a neural network.

2. The computer-implemented method of claim 1 , further comprising:

detecting a health status check event indicating that a check of the health of the hard disk drive is to be undertaken; and

in response to the health status check event,

analyzing the digital audio corresponding to the audio detected by the audio transducer for classification, wherein the digital audio that is classified occurs within an analysis window.

3. The computer-implemented method of claim 2 , wherein the health status check event includes one or more of:

a read operation of the hard disk drive;

a write operation of the hard disk drive;

a park operation of the hard disk drive;

actuation of a spindle motor of the hard disk drive;

actuation of a head armature of the hard disk drive;

a test operation initiated in response to a testing schedule; and

a test operation initiated in response to detection of a hard disk drive error.

4. The computer-implemented method of claim 1 , further comprising:

executing audio pre-processing operations on the digital audio; and

providing the pre-processed digital audio to the input of the audio feature extractor to generate the one or more audio feature vectors.

5. The computer-implemented method of claim 1 , wherein

the audio feature extractor includes a pre-trained, modified VGGish neural network.

6. The computer-implemented method of claim 1 , wherein

the trained machine learning model is trained using digital audio samples corresponding to one or more of:

audio of a properly operating hard disk drive;

audio of a hard disk drive having scratches;

audio of a hard disk drive having bad sectors;

audio of a hard disk drive having spindle issues;

audio of a hard disk drive having bearing issues; and

audio of a hard disk drive having head issues.

7. A system comprising:

a hard disk drive controller;

a non-transitory, computer-readable storage medium embodying computer program code including instructions executable by the hard disk drive controller and configured for:

recording digital audio corresponding to audio detected by an audio transducer in proximity to moving mechanical components of a hard disk drive; and

determining a health status of the hard disk drive based on a classification of the recorded digital audio, wherein the recorded digital audio is classified using a trained machine learning model, the determining comprising

providing the digital audio to an input of an audio feature extractor to generate one or more audio feature vectors, wherein the audio feature extractor includes a neural network; and

providing the one or more audio feature vectors to a hard disk drive health status classifier, wherein the hard disk drive health status classifier provides an output classification corresponding to a health status of the hard disk drive based on the one or more audio feature vectors, wherein the hard disk drive health status classifier includes a neural network.

8. The system of claim 7 , wherein the instructions are further configured for:

detecting a health status check event indicating that a check of the health of the hard disk drive is to be undertaken; and

in response to the health status check event,

analyzing the digital audio corresponding to the audio detected by the audio transducer for classification, wherein the digital audio that is classified occurs within an analysis window.

9. The system of claim 8 , wherein the health status check event includes one or more of:

a read operation of the hard disk drive;

a write operation of the hard disk drive;

a park operation of the hard disk drive;

actuation of a spindle motor of the hard disk drive;

actuation of a head armature of the hard disk drive;

a test operation initiated in response to a testing schedule; and

a test operation initiated in response to detection of a hard disk drive error.

10. The system of claim 7 , wherein the instructions are further configured for:

executing audio pre-processing operations on the digital audio; and

providing the pre-processed digital audio to the input of the audio feature extractor to generate the one or more audio feature vectors.

11. The system of claim 7 , wherein

the audio feature extractor includes a pre-trained, modified VGGish neural network.

12. The system of claim 7 , wherein

the trained machine learning model is trained using digital audio samples corresponding to one or more of:

audio of a properly operating hard disk drive;

audio of a hard disk drive having scratches;

audio of a hard disk drive having bad sectors;

audio of a hard disk drive having spindle issues;

audio of a hard disk drive having bearing issues; and

audio of a hard disk drive having head issues.

13. A non-transitory, computer-readable storage medium embodying computer program code, the computer program code comprising computer executable instructions configured for:

recording digital audio corresponding to audio detected by an audio transducer in proximity to moving mechanical components of a hard disk drive; and

determining a health status of the hard disk drive based on a classification of the recorded digital audio, wherein the recorded digital audio is classified using a trained machine learning model, the determining comprising

providing the digital audio to an input of an audio feature extractor to generate one or more audio feature vectors, wherein the audio feature extractor includes a neural network; and

providing the one or more audio feature vectors to a hard disk drive health status classifier, wherein the hard disk drive health status classifier provides an output classification corresponding to a health status of the hard disk drive based on the one or more audio feature vectors, wherein the hard disk drive health status classifier includes a neural network.

14. The non-transitory, computer-readable storage medium of claim 13 , wherein the instructions are further configured for:

detecting a health status check event indicating that a check of the health of the hard disk drive is to be undertaken; and

in response to the health status check event,

analyzing the digital audio corresponding to the audio detected by the audio transducer for classification, wherein the digital audio that is classified occurs within an analysis window.

15. The non-transitory, computer-readable storage medium of claim 14 , wherein the health status check event includes one or more of:

a read operation of the hard disk drive;

a write operation of the hard disk drive;

a park operation of the hard disk drive;

actuation of a spindle motor of the hard disk drive;

actuation of a head armature of the hard disk drive;

a test operation initiated in response to a testing schedule; and

a test operation initiated in response to detection of a hard disk drive error.

16. The non-transitory, computer-readable storage medium of claim 13 , wherein the instructions are further configured for:

executing audio pre-processing operations on the digital audio; and

providing the pre-processed digital audio to the input of the audio feature extractor to generate the one or more audio feature vectors.

17. The non-transitory, computer-readable storage medium of claim 13 , wherein

the audio feature extractor includes a pre-trained, modified VGGish neural network.

Assignments (9)

Cited By (2)

US 12,242,328 US 12,243,559