ICACT20260214 Question.1
Questioner: abilaaniloshan@gmail.com 2026-02-10 오전 10:05:32 |
ICACT20260214 Answer.1
Answer by Auhor hominpark@etri.re.kr 2026-02-10 오전 10:05:32
Chrome  Click!! |
| Why did you choose an explainable, feature-based Health Index instead of an end-to-end deep learning classifier for fault detection? |
We deliberately chose interpretable features over black-box deep learning models to ensure field engineers can understand, validate, and trust the results. By using 14 engineered features (e.g., spectral centroid, kurtosis), engineers can correlate specific feature deviations with physical defect mechanisms (such as stiffness changes), enabling actionable root cause analysis |
ICACT20260214 Question.2
Questioner: abilaaniloshan@gmail.com 2026-02-10 오전 10:05:55 |
ICACT20260214 Answer.2
Answer by Auhor hominpark@etri.re.kr 2026-02-10 오전 10:05:55
Chrome Click!! |
| How does the dual-threshold strategy improve reliability compared to a simple fault/no-fault decision? |
The dual-threshold strategy (τ
1
,τ
2
) creates a Caution zone that acts as an operational buffer. This zone improves reliability by absorbing uncertainty and filtering out approximately 60-70% of potential false alarms. Instead of triggering an immediate stop, it initiates a secondary verification, thereby preventing unnecessary production line stoppages |
ICACT20260214 Question.3
Questioner: hhl9438@chungbuk.ac.kr 2026-02-10 오전 10:06:11 |
ICACT20260214 Answer.3
Answer by Auhor hominpark@etri.re.kr 2026-02-10 오전 10:06:11
Chrome Click!! |
| The Caution zone triggers an automated re-inspection. Does this frequent re-checking significantly increase the total inspection time, and is there a limit to how many times it retries? |
According to the sources, the Caution zone triggers an automated re-inspection protocol to confirm status and filter noise. While specific retry limits or exact time increases are not detailed, the system is designed to operate autonomously with mobile manipulators. This approach aims to reduce idle time and costs associated with manual inspections, thereby maintaining overall operational efficiency without human intervention |
ICACT20260214 Question.4
Questioner: md02anas02@gmail.com 2026-02-10 오전 10:06:32 |
ICACT20260214 Answer.4
Answer by Auhor hominpark@etri.re.kr 2026-02-10 오전 10:06:32
Chrome Click!! |
| Did you use a public industrial dataset or data from a specific manufacturing plant? |
We did not use a public dataset; instead, we utilized 1,080 real-world percussion test samples collected from generator stator wedge inspections at actual power plants. This dataset comprises Normal (320), Caution (320), and Fault (440) samples collected across multiple units and time periods |
