Home Page
Analysis Page
Data Output Page
ML Future Prediction Images
ML Future Prediction Data

PipeAlign: Automated ILI Alignment & Integrity Analysis

Pipeline integrity analysis is high-stakes, but ILI data is messy: inconsistent vendor schemas, unit mismatches, and odometer drift make multi-run comparison slow and error-prone. Engineers often rely on manual spreadsheet alignment just to determine what actually changed.

PipeAlign turns raw, multi-year ILI files into a deterministic, traceable integrity analysis in minutes.

What It Does

PipeAlign automates multi-run ILI analysis end-to-end:

Ingests Excel/CSV ILI exports from different vendors
Normalizes schemas, units, and feature fields
Aligns inspection runs using girth welds as fixed physical anchors
Matches anomalies using deterministic physical tolerances
Explicitly flags new, missing, and ambiguous features
Computes corrosion growth rates
Generates tables, charts, and downloadable outputs
Projects integrity trends through 2030 (deterministic + optional ML)

How It Works

PipeAlign is deterministic by design. All core logic is explainable, auditable, and conservative.

1. Normalization

All runs are converted into a canonical physical representation:

Distance → meters
Clock position → degrees (0–360, circular)
Depth → percent wall thickness
Length / width → millimeters

This guarantees apples-to-apples comparison across vendors and years.

2. Weld-Anchored Alignment (Odometer Drift Correction)

Runs are aligned segment-by-segment using matching girth welds as fixed landmarks.

For each segment bounded by two adjacent welds:

scale  = (d_prev2 - d_prev1) / (d_later2 - d_later1)
offset = d_prev1 - scale * d_later1

For any feature with raw distance d_raw in the later run:

d_corrected = scale * d_raw + offset

This piecewise linear mapping corrects odometer drift without smoothing or overfitting, preserving physical traceability.

3. Deterministic Anomaly Matching

After alignment, anomalies are matched using hard physical constraints:

Axial distance tolerance
Circular clock difference
Feature type compatibility

Candidates failing any constraint are discarded.

Each remaining candidate pair is scored:

score =
  w_d   * |Δd| +
  w_c   * Δθ +
  w_dep * |Δdepth| +
  w_len * |Δlength| +
  w_wid * |Δwidth|

Lower score = higher physical similarity.

Ambiguity rule:
If the two best scores are within a small threshold ε, the anomaly is marked Ambiguous instead of forcing a match. This prevents false continuity in dense regions.

4. Growth & Projections

For matched anomalies:

growth_rate = (depth_2 - depth_1) / Δt

Deterministic projections: bounded extrapolation through 2030
Optional ML projections: advisory only; never override deterministic matching

Why This Matters

Problem	PipeAlign Approach
Vendor inconsistency	Canonical schema + unit normalization
Odometer drift	Weld-anchored segment alignment
Dense anomalies	Explicit ambiguity, no guessing
Trust gap	Deterministic math over black-box models

Key insight: In infrastructure analytics, defensibility beats sophistication. Engineers trust results they can audit.

Accomplishments

Replaced hours of manual spreadsheet alignment
Deterministic matching with explicit uncertainty
Outputs aligned with professional integrity workflows

What’s Next

Tolerance calibration using labeled review data
Multi-run continuity (e.g., 2007 → 2015 → 2022)
Hardened deployment with audit logs and persistence

Bottom Line

PipeAlign delivers accurate, auditable multi-run ILI alignment, with transparent math and explicit uncertainty — optimized for real integrity decisions.