INDUSTRY FIELD GUIDE
Simulated natural gas pipeline event data for AI testing
A dashboard can continue showing the last reading after a signal stops updating.
NAICS 486210 · Pipeline Transportation of Natural Gas
A dashboard can continue showing the last reading after a signal stops updating. An assistant summarizing event evidence needs to distinguish a stale display from a verified observation. PHMSA publishes pipeline incident and reporting data. Use those resources to understand evidence types. Build any first sample in a detached simulation with invented values. PHMSA pipeline safety reporting index · PHMSA public incident data
Give the model two fictional sources reporting on the same demo interval. One stops updating and retains its prior value. Another contains a timestamped maintenance note. Include a second case where the note belongs to another scenario. Ask which readings are current and what evidence is missing. Require a human-review flag without suggesting an operational action.
Keep real telemetry, thresholds, precise locations, facility configuration, service details, and controls out of the packet. The benchmark is for offline evidence review only.
Describe a synthetic event review
Original Fiund exercise
A display's retained value is not a current sensor observation
A detached dashboard simulation uses arbitrary units and abstract sensor names. It contains no actual telemetry, threshold, location, facility configuration or control instruction.
| Record | Source | What it records |
|---|---|---|
| PG1 | Source A | Sensor DEMO-A last observed 5 U at 10:00. At 10:10 display still shows 5 but last-update field remains 10:00; no new sample. |
| PG2 | Maintenance source | Original note explicitly names DEMO-A reporting interruption at 10:02. It describes reporting only, with no measured process condition. |
| PG3 | Source B | DEMO-B shows 7 U, last observed 10:05. Attached interruption note names separate scenario DEMO-C; no B-specific explanation or 10:10 observation. |
Question for the model
As of 10:10, which values are last-known rather than current, and what explanations can be attributed to each source?
Show the reference answer and checks
PG1's five is A's last-known 10:00 observation, not a new 10:10 reading. PG2 supports an A-specific reporting interruption without establishing the underlying process state. PG3's seven is B's 10:05 observation; the C note cannot explain B's later missing update. Report timestamps, staleness and source-specific evidence, leaving current conditions unverified. The packet supports human review only, without recommending any pipeline action.
- Treating a retained display value as a fresh measurement.
- Applying another scenario's maintenance note or interpreting a reporting gap as a process condition.
Fields to preserve
- Sensor/scenario alias
- Defines source scope.
- Observation time versus display time
- Exposes stale data.
- Maintenance target and missing update
- Limits explanation.
Test another case
Add a fresh B reading at 10:10 while A remains stale. Require the two source states to update independently.
Use original simulated values. Exclude real settings and systems; the task remains an offline evidence review.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.