Telemetry in.
Explainable faults out.
STAFDS simulates spacecraft telemetry, runs it through a deterministic validation-and-filtering pipeline, and flags abnormal conditions using rule-based diagnostic logic you can actually read and trust — not a black box.
What actually happens to a packet
Every simulated telemetry packet passes through the same fixed sequence — nothing is skipped, and nothing on the frontend decides fault state independently.
What's inside the prototype
Each module below is independently testable and maps to a real ground-segment concept — scaled down to something one engineer can build, run, and explain.
Telemetry Simulator
Generates 18 spacecraft parameters across thermal, power, attitude, comms, and system groups — with configurable seed and sampling rate.
Packet Validation
Catches missing fields, NaNs, timestamp anomalies, and duplicate or out-of-sequence packets before they reach the pipeline.
Signal Filtering
Raw, moving-average, and EMA views of every channel, toggleable on the live chart with configurable window and alpha.
Fault Detection
Deterministic, explainable rules — never a statistical black box — with hysteresis to prevent alarm flapping near a threshold.
Health Scoring
0–100 composite score with penalty weighting per fault severity, classified NOMINAL through CRITICAL.
Scenario Control
Switch between Overheating, Low Battery, High Current, Comm Loss, and Recovery to watch the full fault lifecycle play out.
Event Timeline
Every fault detection, acknowledgement, recovery, and scenario change is recorded with a UTC timestamp.
Playback & Export
Replay recorded telemetry through the identical live pipeline, or export history to CSV for offline analysis.
Engineering Reports
Generates parameter statistics, fault summaries, and health trends for a selected monitoring period.
Four rule types, one explanation each
Rule-based prototype diagnostic logic — every fault the system raises comes with the specific rule and observed value that triggered it, not a confidence score.
Threshold crossing + hysteresis
Warning and critical boundaries per parameter, with a margin required before a fault is marked recovered — stops alarms from flapping right at the line.
Rate-of-change detection
Flags a value that jumps further in one sample than is physically realistic for that parameter.
Stuck-sensor detection
A value that hasn't moved for a configurable number of samples is treated as a possible frozen or disconnected sensor.
Correlation rules
Combines several parameters — e.g. rising current, falling voltage, and rising temperature together — to catch power-system anomalies no single threshold would.
Built to run local-first
No paid APIs, no cloud dependency, no remote fonts or CDN calls once installed.
See the fault lifecycle end to end
The live demo runs entirely in your browser — start the simulator, switch to Overheating, and watch a fault get detected, escalate, and recover in real time.
▶ Launch Live Demo