Recognizing and identifying defect patterns on magnetic media

Patent metadata

Field	Value
Publication number	US-9964607-B1
Application number	US-201715687894-A
Country	US
Kind code	B1
Filing date	Aug 28, 2017
Priority date	Aug 28, 2017
Publication date	May 8, 2018
Grant date	May 8, 2018

Abstract

A method includes generating synthetic data related to known defect patterns on surfaces of magnetic media using parameterized rules. A classifier model is trained with the synthetic data so that the classifier model learns how to detect and identify defect patterns on magnetic media. Performance of the classifier model is validated by using real defect pattern data. The classifier model is deployed for use in identifying defective data patterns on magnetic media test specimens. The classifier may be used before or after clustering defect data points on surfaces of magnetic media.

First claim

What is claimed is:

1. A method comprising:
generating synthetic data related to known defect patterns on surfaces of magnetic media using parameterized rules;
training a classifier model with the synthetic data so that the classifier model learns how to detect and identify defect patterns on magnetic media;
validating performance of the classifier model by using real detect pattern data; and
deploying the classifier model for use in identifying defective patterns from data collected from magnetic media test specimens.

2. The method of claim 1 , wherein the classifier model comprises a convolutional neural network (CNN).

3. The method of claim 1 , wherein the synthetic data comprises simulated images of surfaces of magnetic media having defect data patterns.

4. The method of claim 3 , wherein generating the synthetic data comprises randomly selecting simulated defect patterns for each simulated image of magnetic media.

5. The method of claim 4 , wherein the defects for each simulated image of magnetic media is written to a file along with labels identifying the simulated defect patterns.

6. The method of claim 5 , wherein each file is used in training the classifier model.

7. A computer-implemented method comprising:
classifying defect patterns on a test specimen of magnetic media using a classifier model that is trained by synthetic data generated from parameterized rules related to known defect patterns so as to identify what defect patterns exist on the test specimen of magnetic media;
resolving conflicting defect patterns that were identified during classification using a rule-based fusion model; and
clustering the identified defect patterns to determine which defect points belong to which defect patterns on the test specimen of magnetic media.

8. The method of claim 7 , wherein the classifier model comprises a convolutional neural network (CNN).

9. The method of claim 7 , wherein the test specimen comprises an image of a surface of magnetic media.

10. The method of claim 9 , wherein classifying defect patterns on the image of the surface of magnetic media comprises performing one or more consecutive convolutional computations of enhancing or de-enhancing pixels on the image.

11. The method of claim 10 , wherein classifying defect patterns on the image of the surface of magnetic media comprising performing an operation of pooling after one or more convolutional computations in order to reduce the resolution of the image.

12. The method of claim 11 , wherein after the operations of convolution and pooling are complete, the image is flattened and further classified or labeled with defect pattern types.

13. A system of recognizing and identifying defect patterns on surfaces of magnetic media, the system comprising:
a user interface configured to provide controls to a user to select defect data files of magnetic media to be submitted for recognition and identification of defect patterns;
a file processor configured to load and divide the selected defect data files of magnetic media into surfaces for processing;
a plurality of surface processors configured to process the selected defect data files of magnetic media in parallel to increase the speed at which defect data is processed during manufacturing of magnetic media.

14. The system of claim 13 , further comprising a main processor that generates and displays detailed results and summary data fields when surface processing is completed.

15. The system of claim 13 , further comprising a main processor that loads configuration data to configure the plurality of surface processors.

16. The system of claim 15 , wherein the user provides the configuration data loaded by the main processor, the configuration data comprising setting a number of parallel surface processors that are to be used.

17. The system of claim 13 , wherein at least one of the surface processors comprises an engine that classifies defective data on the surface of magnetic media into defect patterns before clustering defect points on the surface of the magnetic media.

18. The system of claim 13 , wherein at least one of the surface processors comprises an engine that clusters defect points on the surface of the magnetic media before classifying the defective data on the surface of magnetic media into defect patterns.

19. The system of claim 13 , wherein the plurality of surface processors comprise a hybrid of surface processors where at least one of the surface processors comprises an engine that classifies defective data on the surface of magnetic media into defect patterns before clustering defect points on the surface of the magnetic media and at least one of the surface processors comprises an engine that clusters defect points on the surface of the magnetic media before classifying the defective data on the surface of magnetic media into defect patterns.

20. The system of claim 13 , wherein when the user interface submits select defect data files of magnetic media for recognition and identification of defect patterns the user interface uploads the select defect data files to a web server.

Assignees

• Seagate Technology Llc

Inventors

• Propes Nicholas C
• Baldini Ryan
• Liwag Rommel
• Tjahjadi Arswendo

Classifications

• G06V10/82

  using neural networks · CPC title

• G06F2218/12

  Classification; Matching · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06V10/454

  Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN] · CPC title

• G06N3/09

  Supervised learning · CPC title