Skip to content

Instructor station

github-actions[bot] edited this page Oct 20, 2026 · 1 revision

Instructor station

The instructor station at http://localhost:8121 is the simulator side of the glass. It is deliberately dark and violet, so nobody mistakes it for the dam SCADA. It starts and stops the hidden fault scenarios and shows the simulators' ground truth.

Instructor station

Hidden faults. The instructor station talks to the dam simulator and the CH PLC simulator directly over HTTP, never through the broker. Nothing in MQTT says that a scenario is active: learners and vendors diagnose from the data alone. Don't show this page to the people who are supposed to find the fault.

Scenarios A–H

Scenario (id) Failure mode Target Speed What changes Who notices first
A Turbine guide bearing degradation (tgb_bearing_degradation) #1 a unit weeks → ~20 min One TGB pad (option pad, default 3) creeps, then accelerates (~64 → ~95 °C); neighbours follow a little, the oil ~1–2 K; late on, the TGB 1X grows CH: PLC alarms at 75 / 85 °C. The customer's pad-1 RTD rises ~3 K
B Cooling-water strainer blockage (cooling_strainer) #2 a unit ~8 min All three cooling circuits fall to ~30 %; every bearing and the stator warm together Both: customer COOLING_WATER_FLOW_LOW within ~4 min; left alone, a bearing trip after ~25–30 min
C AGC parks a unit in the rough zone (agc_rough_zone) #7 a unit (default Unit2) live The plant controller's rough-zone table is bypassed for that unit: it gets ~22 MW at ~55 % gate, the others take the rest. Same plant MW CH: draft-tube pulsation at ~1.4 Hz, sub-synchronous shaft vibration, headcover vibration. The customer sees a unit at part load with swinging MW
D HPU pump wear / internal leakage (hpu_pump_wear) #11 a unit weeks → 30 min Pump output −55 %, leakage ×4: faster accumulator decay, more and longer pump runs CH: PumpStartsPerHour trend (rolling hour, so it lags), dP/dt of AccumulatorPressure
E Wicket gate shear pin break (shear_pin) #10 a unit instant ShearPinStatus = BROKEN; TGB vibration and headcover vibration step up at constant MW; WicketGateBreakawayAlarm while the gates move CH: PLC SHEAR_PIN_BROKEN (critical) and alarm-triggered waveforms
F Mechanical unbalance after maintenance (unbalance) #4 a unit step 1X ×4.2 at UGB, ×3.4 at LGB, ×1.8 at TGB with a 60° phase shift; UGB above the vibration alert CH only: the customer has no vibration probes
G RTD sensor fault (rtd_fault) #15 a unit ~25 min One CH RTD element (option tag, default Bearings/TGB/PadTemp02) drifts up to +20 °C; option mode: drift, flatline, spike. Quality stays GOOD CH, if it cross-checks neighbours. The customer's element A is fine
H Low reservoir, low head (low_reservoir) #8 + efficiency the plant ~17 min Inflow drops, the level falls 1.5 m/min to ~386.5 m a.s.l., head to ~82 m: lower efficiency, more flow per MW, gates near 100 %, more cavitation; 110 MW is no longer reachable Both: SCADA reservoir alarms and MWSetpointReachable false; CH headcover vibration and accelerometer bursts

Stopping H refills the reservoir at the same compressed rate.

Further scenarios

Scenario (id) Failure mode Effect
Oil cooler fouling (cooler_fouling) #2 Heat transfer halves over 30 min; flows stay normal, oil hotter, cooler outlet colder
TGB oil contamination (oil_water_ingress) #3 TGB oil level +8 %, oil temperature +3 K over 30 min
Misalignment / bearing clearance (misalignment) #5 2X grows, orbits turn elliptical (LGB, TGB), over 10 min
Hydraulic unbalance (hydraulic_unbalance) #6 A 1X component that grows with gate opening, not speed, over 10 min
Runner blade crack (runner_crack) #9 Blade-pass harmonics (2 × 65 Hz, sidebands ±1X) grow in the headcover accelerometer and TGB probes, over 30 min
Servomotor seal leak (servo_seal_leak) #12 HPU sump level drops, servo B lags the setpoint, over 30 min
Shaft seal wear (shaft_seal_wear) #13 Seal water flow, headcover sump pump starts and headcover pressure rise, over 30 min
Generator air-gap anomaly (air_gap) #14 Minimum air gap shrinks from ~11.6 to ~9.5 mm over 30 min (PLC alarm below 10.5)
Flood inflow (flood) — Level brought near 420 m (compressed), inflow 1,200 m³/s: the spillway opens in AUTO. Stopping restores the level
Generator protection trip (generator_trip) — Instant load rejection and emergency stop of one unit (default Unit1); reset at standstill
Debris on the trash rack (trash_debris) — Trash rack DP rises 40× faster for 15 min; the raker can't keep up

Ramp scenarios accept an option duration_s to change the compression. The failure-mode numbers refer to Failure modes.

Repair

Stop on a scenario "repairs" it at once: the fault term is gone, but the physics still has to recover (a hot bearing cools with its time constant). Reset all stops every scenario on every unit. A unit that tripped must still be reset and restarted from the dam SCADA.

API

curl -s localhost:8121/api/scenarios | python3 -m json.tool | head -40
curl -s -X POST -H 'Content-Type: application/json' -d '{"unit":"Unit1"}' \
  localhost:8121/api/scenarios/tgb_bearing_degradation/start
curl -s -X POST -H 'Content-Type: application/json' \
  -d '{"unit":"Unit1","options":{"mode":"flatline","tag":"Bearings/TGB/PadTemp03"}}' \
  localhost:8121/api/scenarios/rtd_fault/start
curl -s -X POST -H 'Content-Type: application/json' -d '{"unit":"Unit1"}' \
  localhost:8121/api/scenarios/tgb_bearing_degradation/stop
curl -s localhost:8121/api/ground_truth | python3 -m json.tool
curl -s -X POST localhost:8121/api/reset
Route
GET /api/scenarios Every scenario, dam and CH, with letter, group, target and options
GET /api/ground_truth Active scenarios with their severity, the plant controller's internals, the PLC alarms
GET /api/overview A few live values per unit, for orientation
POST /api/scenarios/<id>/start {"unit": "Unit1", "options": {...}} (409 if already running)
POST /api/scenarios/<id>/stop {"unit": "Unit1"}
POST /api/reset Stop everything

Clone this wiki locally