Construction of hardware failure prediction tool set using big data
| Project/Area Number |
15K00066
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | Japan Aerospace EXploration Agency |
|---|
Principal Investigator |
FUJITA Naoyuki 国立研究開発法人宇宙航空研究開発機構, 航空技術部門, 主幹研究開発員 (70358480)
|
|---|
| Research Collaborator |
ITO Rika
|
|---|
| Project Period (FY) |
2015-04-01 – 2018-03-31
|
| Project Status |
Completed (Fiscal Year 2017)
|
| Budget Amount *help |
¥3,640,000 (Direct Cost: ¥2,800,000、Indirect Cost: ¥840,000)
Fiscal Year 2017: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2016: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2015: ¥2,340,000 (Direct Cost: ¥1,800,000、Indirect Cost: ¥540,000)
|
|---|
| Keywords | 障害検出 / 原因特定 / 障害予測 / ビッグデータ |
|---|
| Outline of Final Research Achievements |
It is required to predict the failure of the hardware of the computer system, while the system has become large scale. We analyzed the prediction of hard disk failure by machine learning of S.M.A.R.T. information on hard disk. A so-called positive predictive value (PRE) was used as an evaluation index. As a result, it was observed that the closer the day of failure is, the more unique signs affecting PRE appear. In addition, by improving the detection accuracy by machine learning of S.M.A.R.T. information, it was shown that prediction of disability can be expected with considerably high precision. It was found that the detection power of the judging machine during the period close to the number of days before failure was high.
|
Report
(4 results)
Research Products
(8 results)
