Mobile Learning Console

Probabilistic Digital Twin for Transmission Pipelines

A phone-first reconstruction of Ayello et al.'s IPC 2020 pipeline case: uncertain inputs, Bayesian threat inference, and inspection decisions.

IPC 2020 19 km dry-gas line
PoF in 2031 0.02% Target 5%
Water exposure 0.0 mo/y Dry gas baseline
Dominant threat Uniform Posterior node
Inspection year No trigger PoF threshold
Risk status Low threat Below threshold

Live Model Stage

From uncertain inputs to inspection decisions

The baseline keeps Table 2 inputs as probability distributions. The twin propagates uncertainty into flaw depth and probability of failure instead of collapsing it into one fixed value.

  1. Probabilistic inputs
  2. Bayesian network
  3. Threat mechanisms
  4. Decision loop

1. Spatial Pipe State

Corrosion mechanisms along the pipeline

3D-style view
Uniform corrosion The highlighted threat chain is synchronized with the Bayesian network and decision panel.
Uniform corrosion Localized corrosion Erosion MIC

2. Bayesian Network

Evidence updates threat probabilities

Draggable
P(threat | evidence) ∝ P(evidence | threat) × P(threat)
Water0% pH/O2/CO2verify Flow0.5/0.5 Uniform22% Localized20% Erosion15% MIC11% Flaw18% PoF0% Actionwatch

The Bayesian network passes evidence on water, chemistry, and flow to multiple threat mechanisms, then aggregates them into flaw depth and PoF.

3. Future Risk

Flaw depth and probability of failure

Fig. 4 logic

Teaching simulation calibrated to the paper's baseline and water-present narratives.

4. Decision Maker Panel

When should inspection happen?

Threshold
Inspection timing No trigger PoF below 5%

When PoF reaches the unacceptable-risk threshold you define, the next inspection date appears.

Decision log is empty.

Scenario Inputs

Operate as the decision maker

Learning Story

The four questions this page answers

1 Why was it proposed?

Conventional digital twins can look precise while hiding missing data, sensor limitations, and model uncertainty.

2 How does it solve the problem?

Inputs enter as distributions, models exchange probabilities, and Bayesian-network nodes update when new evidence arrives.

3 What should the decision maker do?

Set the PoF threshold, test evidence, compare mechanisms, choose inspection timing, choose locations, and prioritize new data.

4 What risk is controlled?

The core risk is false certainty: late inspection, missed mechanisms, and poor data spending caused by untracked uncertainty.

Paper Evidence

Source-backed anchors

Problem: uncertainty in the information layer can propagate into unsafe conclusions.

Mechanism: Bayesian networks connect prior knowledge, observations, and conditional probability tables.

Decision: unacceptable risk is set as a PoF threshold; inspection timing and data collection follow from the posterior risk.

Evidence image from paper page 10