nuluh/thesis
1
0
Fork 0
Code Issues 47 Pull Requests Actions Packages Projects Releases 1 Wiki Activity

[EXP] Alternative undamage case data #99

New Issue
Open
opened 2025-06-26 03:24:54 +00:00 by nuluh · 0 comments
nuluh commented 2025-06-26 03:24:54 +00:00 (Migrated from github.com)
Copy Link

Hypothesis

Constructing label 0 (“healthy”) using complement columns (all non-damaged indices) from each BD{i}.TXT—rather than only using zzzBU.TXT—will yield more realistic and generalizable models for vibration-based structural damage detection, due to the global propagation of local damage effects in fixed-fixed grid structures.

Background & Motivation

Current practice uses only zzzBU.TXT (all healthy) for label 0, but literature and experimental evidence (Shahri var & Bouwkamp 1982, Yu & Huo, Wang et al., Tadros) show that even remote healthy nodes exhibit altered vibration signatures when a single bolt is loosened elsewhere in the structure. This effect is especially pronounced in fixed-fixed grids due to global changes in the stiffness matrix. A more realistic label 0 should therefore include healthy columns from damaged scenarios (complement indices in BD{i}.TXT).

Dataset

  • Data files: zzzBU.TXT (baseline), zzzBD{i}.TXT (i=1..30)
  • For label 0: use all complement columns from each zzzBD{i}.TXT (non-damaged indices) and compute STFT vectors. To balance across classes, sample ~107 STFT vectors per sensor end per file (4 sensor ends × 6 files × 107 ≈ 2565 vectors).
  • For damaged labels 1–6: use all 513 STFT vectors per sensor end per file (5 sensor ends × 513 = 2565 vectors).
  • Result: each label (0–6) has exactly 2565 STFT vectors.

Methodology

  1. Extract non-damaged columns from each zzzBD{i}.TXT and compute STFT for each sensor end.
  2. Balance each class to 2565 STFT vectors as described in the Dataset section.
  3. Train and evaluate models using the balanced dataset, comparing performance against the original unbalanced approach (label 0 from zzzBU.TXT only).
  4. Analyze classification accuracy and robustness for healthy vs. damaged classes and the impact of dataset balancing.

Parameters & Hyperparameters

Evaluation Metrics

  • Accuracy
  • F1-score (macro-averaged)
  • Confusion matrix analysis for healthy vs. damaged
  • Effect of balancing on class-wise performance

Notebook Location

notebooks/stft.ipynb

Dependencies

  • May depend on preprocessing updates from prior data pipeline issues

References

  • Shahri var & Bouwkamp (1982): Model study of effects of damage on vibration properties (PDF)
  • Yu & Huo: Grid truss vibration changes (paper)
  • Wang et al., Tadros, review articles (Wang SSRN, Tadros thesis)

Additional Notes

No response

### Hypothesis Constructing label 0 (“healthy”) using complement columns (all non-damaged indices) from each BD{i}.TXT—rather than only using zzzBU.TXT—will yield more realistic and generalizable models for vibration-based structural damage detection, due to the global propagation of local damage effects in fixed-fixed grid structures. ### Background & Motivation Current practice uses only zzzBU.TXT (all healthy) for label 0, but literature and experimental evidence (Shahri var & Bouwkamp 1982, Yu & Huo, Wang et al., Tadros) show that even remote healthy nodes exhibit altered vibration signatures when a single bolt is loosened elsewhere in the structure. This effect is especially pronounced in fixed-fixed grids due to global changes in the stiffness matrix. A more realistic label 0 should therefore include healthy columns from damaged scenarios (complement indices in BD{i}.TXT). ### Dataset - Data files: zzzBU.TXT (baseline), zzzBD{i}.TXT (i=1..30) - For label 0: use all complement columns from each zzzBD{i}.TXT (non-damaged indices) and compute STFT vectors. To balance across classes, sample ~107 STFT vectors per sensor end per file (4 sensor ends × 6 files × 107 ≈ 2565 vectors). - For damaged labels 1–6: use all 513 STFT vectors per sensor end per file (5 sensor ends × 513 = 2565 vectors). - Result: each label (0–6) has exactly 2565 STFT vectors. ### Methodology 1. Extract non-damaged columns from each zzzBD{i}.TXT and compute STFT for each sensor end. 2. Balance each class to 2565 STFT vectors as described in the Dataset section. 3. Train and evaluate models using the balanced dataset, comparing performance against the original unbalanced approach (label 0 from zzzBU.TXT only). 4. Analyze classification accuracy and robustness for healthy vs. damaged classes and the impact of dataset balancing. ### Parameters & Hyperparameters - ### Evaluation Metrics - Accuracy - F1-score (macro-averaged) - Confusion matrix analysis for healthy vs. damaged - Effect of balancing on class-wise performance ### Notebook Location notebooks/stft.ipynb ### Dependencies - May depend on preprocessing updates from prior data pipeline issues ### References - Shahri var & Bouwkamp (1982): Model study of effects of damage on vibration properties ([PDF](https://nehrpsearch.nist.gov/static/files/NSF/PB83148742.pdf)) - Yu & Huo: Grid truss vibration changes ([paper](https://onlinelibrary.wiley.com/doi/10.1155/2015/246480)) - Wang et al., Tadros, review articles ([Wang SSRN](https://papers.ssrn.com/sol3/Delivery.cfm/261d87f0-9623-4ad2-9bfa-6d134236d572-MECA.pdf?abstractid=4286065&mirid=1), [Tadros thesis](https://krex.k-state.edu/server/api/core/bitstreams/118ce825-0e69-474c-9ffa-cf7084f9a0d9/content)) ### Additional Notes _No response_
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
bug
Something isn't working
 critical
Block Progress
 documentation
Improvements or additions to documentation
 documentclass
Improvements or additions to LaTeX documentclass
 duplicate
This issue or pull request already exists
 enhancement
New feature or request
 good first issue
Good for newcomers
 help wanted
Extra attention is needed
 invalid
This doesn't seem right
 latex
Related to the writings
 question
Further information is requested
 wontfix
This will not be worked on
No Label critical
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
nuluh
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: nuluh/thesis#99
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?