← Digital Twin Engine

Digital Twin Engine

Scenario Simulation

Fork the live process state, inject a perturbation, and propagate the modified graph forward through time. Compare outcomes across parallel scenarios — without touching real equipment. The what-if engine for first-principles digital twins.

LIVE PROCESS STATELIVEFlow42.0 L/sPressure3.3 barEfficiency86%Power18.6 kWFORK & PERTURBSfork_pointBaselineNo perturbation — live trajectory continuesPump TripPump efficiency set to 0 — cascade downstreamOptimizedVFD frequency increased to 55 HzPROPAGATIONt = 0h → 24h0h24hBASTRPOPTpropagate> dt=0.1s horizon=24h scenarios=3 step=1/20COMPARISON MATRIX — 24h SIMULATION HORIZONScenarioEfficiencyEnergy CostRisk ScoreRecommendationBaseline82%$1240LowContinue monitoringPump Trip42%$2201CriticalStart standby pump within 12sOptimized92%$991LowApply new VFD setpointscenario_engine> compare --baseline --trip --optimized --horizon=24h --metrics=efficiency,cost,riskRESULT: "Optimized" yields +10% efficiency, -$249/day vs baselineWARNING: "Pump Trip" cascade reaches critical state in 4.2s — standby pump response required

How Scenarios Work

Four stages from live state to decision-ready comparison.

01

Fork

Snapshot the current live graph state — every model state variable and every connection's latest value. The fork captures the complete state vector in ~2KB per model.

02

Perturb

Inject a change into the forked state — trip an equipment model, modify a setpoint, add a load, change an input condition. The perturbation is a targeted mutation of one or more state variables.

03

Propagate

Execute the modified graph forward through time. Each model re-evaluates with the perturbation cascading downstream through connected models in topological order. Performance is scoped by graph size and deployment.

04

Compare

Side-by-side comparison of outcomes across all scenarios against the live baseline. Efficiency, energy cost, risk scores, and actionable recommendations — computed for every forked branch.

Scenario Types

Five perturbation categories supported by the scenario execution engine.

TypePerturbationExampleTypical Horizon
Equipment TripSet a model output to zero/failed state"What if Pump 2 trips?"1min - 4h
Setpoint ChangeModify an input parameter"What if we increase VFD to 55 Hz?"1h - 24h
Load ForecastRamp an input over time"What if influent doubles from a storm?"4h - 72h
Failure CascadeTrip one model, observe downstream propagation"What if the blower fails during peak load?"1min - 8h
Maintenance WindowTake a model offline for a period"Can we service Clarifier 2 for 4 hours?"4h - 48h

Technical Specifications

Performance characteristics are scoped by graph complexity, source cadence, scenario horizon, and edge host capacity.

Scenario throughput

Parallel what-if evaluations sized during activation

Scoped

Max parallel scenarios

Concurrent forked graph instances sized by deployment

Scoped

Propagation latency

Per time-step, full DAG traversal depends on graph size

Scoped

State snapshot size

Per model, including all state variables

~2KB

Simulation horizon

Configurable forward projection window

1min - 72h

Time acceleration

Simulate 24h in 86 seconds

Up to 1000x

Test Your First Scenario

Fork the live process state, inject a perturbation, and see the system-wide impact propagate through your digital twin in under 100 milliseconds.