STRUCTURA ACADEMIC · LESSON AREA

Transportation and pavement deterioration modelling

08 · Transportation and pavement deterioration modelling · AI for Civil Engineering

Course review
StandardInternational professional-learning synthesis · ABET 2026–27 · IEA GAPC v4 · ASCE ethics and AI responsibility · NIST AI RMF 1.0
Source3 source files
Review stateTechnical and publication gates pending
LEARNING OUTCOMES

After this chapter, you should be able to

  • Model section-level deterioration without repeated-record leakage.
  • Use future and grouped holdouts.
  • Compare with persistence and trend baselines.
  • Recognise treatment and policy feedback.
  • Separate prioritisation evidence from treatment authority.

Engineering context and methodSource §Lesson 08 · Engineering context and method · GOV-01 · GOV-02 · GOV-03 · RES-06 · DATA-03

Pavement records repeat measurements on sections and carry construction, climate, traffic and treatment history. Split by section and hold out later time. Record the exact LTPP Standard Data Release and table definitions because releases evolve. Historical treatment is not pure ground truth: decisions alter subsequent measurements and labels, creating feedback and potential inequity across the network.

Verified worked exampleSource §Lesson 08 · Verified worked example · GOV-01 · GOV-02 · GOV-03 · RES-06 · DATA-03

WORKED EXAMPLE

Two-section IRI comparison

Observed future IRI values are 1.80 and 2.40 m/km; persistence predicts 1.72 and 2.32; the model predicts 1.68 and 2.52.

  1. Persistence MAE

    (|1.72−1.80| + |2.32−2.40|) / 2

    0.08 m/km
  2. Model MAE

    (|1.68−1.80| + |2.52−2.40|) / 2

    0.12 m/km
  3. Decision

    0.12 > 0.08

    baseline wins this sample

Result. Prefer the simpler persistence result for this tiny illustration; a model name never overrides comparative evidence.

Practical lab · 6 h lesson effortSource §Lesson 08 · Practical lab · 6 h lesson effort · GOV-01 · GOV-02 · GOV-03 · RES-06 · DATA-03

  • Freeze the LTPP SDR version, tables, section IDs, units and exclusion logic.
  • Create persistence and simple trend baselines.
  • Hold out sections and later time; report climate and traffic slices.
  • Draft a prioritisation memo that explicitly excludes treatment authorisation.

Failure modes to investigateSource §Lesson 08 · Failure modes to investigate · GOV-01 · GOV-02 · GOV-03 · RES-06 · DATA-03

  • Same section in train and test.
  • Release or table definition not recorded.
  • Treatment information leaking from after the prediction time.
  • Agency decision treated as objective truth.
  • Network, equity and budget impacts omitted.

Knowledge checksSource §Lesson 08 · Knowledge checks · GOV-01 · GOV-02 · GOV-03 · RES-06 · DATA-03

Five review questions and answer rationales
QuestionAnswer rationale
Why group by section?Repeated records share construction and site properties.
Why hold out later time?Planning concerns future deterioration.
Can a prediction choose treatment?Not alone; structural, economic and network evidence is required.
Why record SDR version?LTPP releases are corrected and expanded.
What is feedback?Decisions based on predictions alter later observations and labels.
Use these as formative checks. Technical and editorial review remain pending.

Key points

  • Start from the accountable engineering decision and its consequence.
  • Compare against a transparent non-AI baseline.
  • Validate on a split that represents intended use and retain human authority.

Source references recorded by the supplied chapter

  • FHWA Long-Term Pavement Performance Standard Data Release 39 documentation.
  • Peer-reviewed pavement-ML review listed as RES-06.
  • NIST AI RMF 1.0.