simulation/fault_simulator.py, plus a small number of payloads derived from the public ESA Anomaly Detection Benchmark. Synthetic and real-derived data are labelled separately. No dataset has been published.When a spacecraft enters safe mode, every day offline costs millions in lost science data, mission delays, and operational overhead.
Select a preset incident or paste raw telemetry JSON. SENTINEL's AI agent will ingest and diagnose in real time.
An illustrative causal chain showing how a single sensor anomaly could cascade into safe mode entry. This graph is fixed demo content, not inferred from telemetry, and its node confidences are NOT EVALUATED.
Chronological fault progression for the scripted scenario. Entries are pre-written and revealed on a fixed timer — they are not derived from telemetry.
Subsystem risk heatmap and health radar for the selected scripted scenario. The bar and radar shapes are illustrative demo values; anomaly confidence and evidence counts are NOT EVALUATED because no model computes them yet.
Click any subsystem to inspect its telemetry and highlight related causal nodes in the failure graph. All values shown are simulated demo data from the scripted scenario — not measurements from a spacecraft.
An illustrative recovery sequence written by hand to show the intended output shape. These steps were not generated by SENTINEL, have not passed safety validation, and carry no measured success probability.
A fixed, hand-written reference list of publicly reported spacecraft anomalies, included for context. SENTINEL does not currently match telemetry against an incident database, and no similarity score is computed.
An autonomous AI pipeline that sits between the spacecraft and mission control — diagnosing, reasoning, and planning without waiting for a ground pass.