Sign In

Plant Library

How RadMah AI produces physically-valid SCADA telemetry across a broad range of industries: layered fidelity tiers, a deep library of pre-built plant templates for well-studied verticals, and a composition path for everything else.

Automatic fidelity selection

The runtime picks the highest-fidelity path it has evidence for. Customers never pick the tier — the platform chooses based on the natural-language description, matched template, and available coverage.

PathData gradeWhat it does
Pre-built plant templateTraining-gradeFull mechanistic physics for well-studied verticals. Purpose-built templates with calibrated dynamics, realistic disturbances, protocol bindings, and cyber-effect overlays.
Composed plantComposition-gradeThe platform composes a plant from a validated library of unit-operation building blocks (control, electrical, fluid, material) connected in a directed graph. Every run is validated against the same quality gate.
Statistical fallbackAugmentation-gradeUsed for protocol-centric signals where physics is a nuisance parameter, so the requested operating profile governs mean, variance, and temporal structure.

Pre-built plant templates

Each pre-built template ships a full domain model: named signals with physical bounds, realistic dynamics, an observation layer that emits true/measured/reported triples, domain invariants, and a parameter schema the platform fills in from the customer’s description.

VerticalDomain coverage
Water / wastewaterTreatment process dynamics, pH chemistry, biological kinetics, realistic diurnal demand patterns.
Power & substationTransformer thermal behaviour, generator inertia, battery state-of-charge, tap-changer action, protection coordination.
Oil & gas midstreamPipeline pressure behaviour, compressor stability, separator operation, multi-component composition tracking.
Chemical processReactor dynamics with temperature-dependent kinetics, mass and energy balance, jacket cooling, conversion and selectivity tracking.
Buildings / HVACZone thermal comfort, chiller and boiler part-load behaviour, damper dynamics, occupancy-driven setpoint shaping.
ManufacturingDiscrete-event behaviour, cycle-time tracking, conveyor dynamics, buffer management, operational KPIs.
MiningVentilation pressure, flotation cell dynamics, thickener level, ore throughput.
Food processingThermal pasteurisation, fermentation dynamics, batch cooking, cold-chain tracking.
LogisticsConveyor networks, buffer occupancy, discrete item tracking.
Transport (rail, port, maritime)Hoist kinematics, belt mechanics, rail block occupancy, signalling state, marine propulsion behaviour.
Hydro & damsReservoir dynamics, spillway behaviour, penstock valve control, flood-control timing.

Industries outside this list are supported through the composition path — the platform generates a plant structure from a library of validated building blocks rather than picking a ready-made template. Either path passes the same validation gate.

Composition library

A library of validated unit-operation building blocks across four families, each with its own internal dynamics and integration policy. Blocks are connected through typed ports; the platform builds the graph and dispatches the integration in the correct order.

  • Control: PID, setpoint trackers, alarm panels, interlocks, sequencers.
  • Electrical: breakers, relays, buses, transformers, loads, metering panels.
  • Fluid: pumps, valves, tanks, pipes, flow meters, pressure transmitters.
  • Material: conveyors, buffers, mixers, reactors, hoists.

Determinism & the seal

Every run is reproducible bit-for-bit given the tuple (description, seed, parameter set, engine version). The dispatch path, plant fabrication output, and simulation state all feed into the evidence seal — two runs with the same seal hash produce identical artifacts.