What it does

AlignIQ aligns multi year pipeline inspection data to a single physical reference. It anchors inspections using girth welds, corrects distance drift locally, and then matches defects with confidence scoring. The result is a clean, reliable view of which defects truly persist and which changes are real.

How we built it

We first align all inspection runs using girth welds as fixed anchors. Then we correct distance drift between welds using piecewise normalization. Once the pipe is aligned, we match defects only within the same joint using position and shape information, and assign a confidence score to each match. One inspection is used as a master reference to keep comparisons consistent across years.

Challenges we ran into

Inspection datasets from different years are inconsistent and incomplete. Distance errors are not uniform and cannot be fixed with a single offset. Some defects lack attributes, making matching uncertain. Balancing strict matching with flexibility required careful tuning.

Accomplishments that we’re proud of

We built an approach that mirrors how the pipe is physically constructed. We reduced false defect movement caused by distance drift. We created a clear, explainable workflow that works even with imperfect data. The system scales cleanly to more than two inspection runs. We have provided with industry standard classification of the output

What we learned

Physical reference points matter more than reported distances. Most false corrosion growth comes from misalignment, not actual damage. Confidence based matching is far safer than forced one to one comparisons. Good visuals make complex engineering ideas easy to understand.

What’s next for AlignIQ

Integrating raw signal matching for higher confidence cases. Adding uncertainty visualization for engineers and decision makers. Automating weld matching and confidence tuning. Extending AlignIQ to support strain, movement, and crack inspection data.

Built With

  • fastapi
  • pydantic
  • python
  • streamlit
  • uvicorn
Share this project:

Updates