Time series, sensors and structural-health-monitoring triage — AI for Civil Engineering

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

Check sampling, synchronisation and missing intervals.
Build a seasonal or persistence baseline.
Calculate and interpret a standardised anomaly score.
Separate drift, environment and possible damage hypotheses.
Design monitoring escalation and fallback.

Engineering context and methodSource §Lesson 06 · Engineering context and method · GOV-02 · GOV-03 · RES-01 · RES-07 · DATA-03

Sensor observations combine structural response with temperature, traffic, moisture, calibration, synchronisation and data-loss effects. Freeze acquisition metadata and quality flags before modelling. Split by time, retain later periods for testing, and compare with persistence or seasonal baselines. An alert is a triage signal; diagnosis needs engineering context, corroborating sensors and inspection evidence.

Illustrative monitoring triage. A threshold crossing routes evidence to competent review; it does not diagnose damage.Original STRUCTURA review diagram · technical sign-off pending

Verified worked exampleSource §Lesson 06 · Verified worked example · GOV-02 · GOV-03 · RES-01 · RES-07 · DATA-03

WORKED EXAMPLE

Standardised monitoring excursion

A reference window has mean 120 microstrain and standard deviation 4 microstrain; a new value is 134 microstrain.

Difference
134 − 120
14 microstrain
Standardise
14 / 4
z = 3.5
Interpret
compare quality, temperature and neighbouring sensors
review trigger

Result. z = 3.5 relative to the declared window. This is not a probability of damage or a diagnosis.

Practical lab · 5 h lesson effortSource §Lesson 06 · Practical lab · 5 h lesson effort · GOV-02 · GOV-03 · RES-01 · RES-07 · DATA-03

Use a synthetic, checksum-frozen multichannel signal with temperature and missing intervals.
Implement quality gates, a persistence baseline and a rolling robust anomaly score.
Test on a later time block and inspect at least five alerts.
Write an alert packet with context, confidence, abstention, escalation and sensor-failure fallback.

Failure modes to investigateSource §Lesson 06 · Failure modes to investigate · GOV-02 · GOV-03 · RES-01 · RES-07 · DATA-03

Random time split.
Clock drift or missing intervals ignored.
Temperature effects described as damage.
Threshold selected after viewing the final events.
No fallback when sensors or communications fail.

Knowledge checksSource §Lesson 06 · Knowledge checks · GOV-02 · GOV-03 · RES-01 · RES-07 · DATA-03

Five review questions and answer rationales
Question	Answer rationale
What does z = 3.5 mean?	The value is 3.5 reference-window standard deviations above that window mean.
Does it diagnose damage?	No; alternative environmental, operational and data-quality causes remain.
Why use a later test period?	It better represents forecasting or future monitoring use.
Why use neighbouring sensors?	Corroboration helps distinguish local sensor issues from system response.
Who determines an asset action?	The authorised engineering process, not the anomaly score.

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

ASCE Policy Statement 573 and Code of Ethics.
Peer-reviewed structural-health-monitoring and computer-vision reviews listed as RES-07.
NIST AI RMF 1.0.