Executed engine · visible state · exportable memory

The GFIS simulator is working software, with every state change connected to evidence.

This is controlled simulator proof: a trained service bundle, nine 48-hour scenarios, methane and VFA/ALK outputs, physics-bound auditing, Control Room handoff and persistent run export.

Accepted executed replay

9 × 48 h
Controlled scenarios

Nominal, overload, pH shock, temperature disturbance, reduced HRT, high solids, recovery and boundary challenges.

432
Hour-indexed rows

Every row retains methane yield, VFA/ALK, stability state, upper bound and violation flag.

144
Warning-state rows

Overload and boundary conditions move the virtual soft sensor away from the stable regime.

Executed 48-hour methane, VFA ALK and physics-bound simulator traces

Executed from the trained Python service. The low-VS boundary challenge records a physics violation before correction; organic-overload conditions maintain a warning-state VFA/ALK response.

Open complete traces Open scenario summary Open run metadata

Control Room: model-backed decision surface

GFIS Control Room showing process controls, predictions and memory export

Historical localhost capture, 9 July 2026. Operators change temperature, pH, OLR, HRT, TS and VS; GFIS displays methane, physics bound, VFA/ALK, violation state, scenario sweep and exportable run memory. The historical “not trained” panel is superseded by accepted replay metadata confirming the trained bundle loaded.

Visible input-to-output path

Every operator input is shown beside methane prediction, physical ceiling, VFA/ALK estimate and stability interpretation. Scenario state can be sent directly to the industrial simulator.

Experiment memory

Prediction, optimization, sweep, 48-hour trace and simulator handoff records retain parameters, outputs and effects for CSV, JSON and report export.

Industrial simulator: state propagation and alarms

GFIS industrial simulator with process variables, alarms and memory panel

Historical localhost capture, 9 July 2026. The simulator receives the Control Room state, recalculates methane and VFA/ALK behaviour, records warning/critical interface states and exposes simulation-memory export.

Operator variableBiological interpretationGFIS visible response
TemperatureMicrobial-rate and mesophilic-efficiency changeMethane trajectory and drift response
pHAcidification or inhibition riskYield reduction and stability message
OLROrganic overload and VFA accumulation pressureVFA/ALK warning and scenario comparison
HRTContact time and washout riskTemporal methane response
TS / VSSubstrate and biodegradable-solids availabilityYield response and methane feasibility ceiling

Product evidence versus plant validation

Working product evidence

Supported now

  • Executable methane prediction
  • Stateful multi-hour simulation
  • VFA/ALK warning behaviour
  • Physics-bound detection
  • Scenario handoff and export memory
Next external layer

Requires plant/shadow-mode data

  • Industrial methane accuracy
  • Measured VFA/ALK calibration
  • Sensor uptime and latency
  • Operator warning precision
  • Closed-loop control performance