The Diagnostics Data Pipeline
Raw diagnostic signals become usable knowledge only after a pipeline synchronizes, calibrates, labels, and routes them reliably.
From signal to knowledge
A fusion plant generates enormous streams of diagnostic data. On their own these are raw counts and voltages. A data pipeline turns them into calibrated, time-aligned, labeled quantities that control systems, the digital twin, and analysts can trust and use. The pipeline is infrastructure, but the science depends on it.
Pipeline stages
- Acquisition: capture signals at their native rate with accurate timestamps.
- Synchronization: align every channel to a common clock.
- Calibration: convert raw values to physical units with known error.
- Quality control: flag bad channels and out-of-range values.
- Routing and storage: deliver to consumers and archive for later.
Two paths, two budgets
A fast path serves real-time control within a strict latency budget, keeping only what control needs. A thorough path serves analysis and archiving, doing the full processing without the latency pressure. Separating them lets each be optimized for its purpose.
Feeding the models
The pipeline feeds state estimation, the digital twin, anomaly detection, and the long-term archive that trains future models. Every downstream capability inherits the pipeline's quality.
For both machines
The Hyperion breeder and the burner have different diagnostics but share the pipeline architecture, so tooling and expertise transfer between them.
Provenance
Every processed value carries its provenance, the raw source, calibration, and processing version, so results remain reproducible and auditable.