Automated User Resource Analyzer

A.U.R.A.

ECLSS Predictive Maintenance System

AI-powered anomaly detection and predictive maintenance for International Space Station life support systems — keeping crew safe before failures happen.

0 ISS Modules
0 Sensor Parameters
0 Fault Types Detected
0 ML Models
0 Second Refresh Rate

Mission-Critical Life Support Intelligence

As human spaceflight ventures farther from Earth, reactive maintenance is a liability. A.U.R.A. transforms ECLSS from reactive to proactive — predicting failures before they endanger crew.

Real-Time Web Application

A.U.R.A. is a full-stack web application built on FastAPI with live WebSocket data streaming. Sensor readings from 7 ISS modules update every second across 20 parameters — accessible from any browser, no installation required.

Four machine learning models work in concert: Isolation Forest detects anomalies, Random Forest classifies fault type, LSTM predicts remaining useful life, and a Deep Q-Network recommends corrective actions.

ISS Modules Monitored

JLP & JPM Node 2 Columbus US Lab Cupola Node 1 Joint Airlock
DIGITAL TWIN — LIVE 3D MODEL

Everything in One Dashboard

Eight integrated views give operators complete situational awareness

🛸
Digital Twin
Interactive SVG ISS schematic with live color-coded sensor overlays. Click any module to drill into sensor detail. Fault states light up in real time.
WebSocket Live
📊
Dashboard
20+ sensor readings across 9 ECLSS subsystems in a unified panel. Auto-refreshes every second. Historical queue holds 10,000+ rows for fast lookups.
1s Refresh
📈
Sensor Detail & Trends
Drill into any parameter for historical trend charts. Identify drift, degradation patterns, and approaching thresholds before they trigger alerts.
Chart.js
🚨
Alerts System
Two-tier alert system: Critical (active fault, immediate action) and Warning (parameter drifting toward threshold). Zero false positives in 1,000 nominal samples.
2-Tier Alerting
🤖
AI Analyst
LLM-powered analyst synthesizes all model outputs into natural-language maintenance recommendations. Explains the why, not just the what.
LLM Integration
🔧
Maintenance Planner
DQN-recommended corrective actions mapped to specific faults. 11 action types from seal injection to valve isolation, ranked by confidence and urgency.
DQN Recommended

Four Models, One Decision

Each model passes its output downstream — raw sensor noise to actionable recommendation in under a second

🌲
Isolation Forest
Unsupervised anomaly detection. Scores every incoming reading — flags deviations from learned normal operation.
🌳
Random Forest
Classifies the fault type from 8 categories when IF flags an anomaly. Confidence >92% triggers direct bypass to action.
🧠
LSTM Predictor
60-step sliding window with multi-head attention. Predicts failure probability and Remaining Useful Life in hours.
🎯
DQN Recommender
Deep Q-Network maps 32-dim system state to optimal corrective action from 11 remediation procedures.

8 Critical Fault Types Detected

Random Forest classifies each anomaly into one of eight mission-critical fault categories with confidence scoring

Cabin Leak
O₂ Generator Failure
O₂ Leak
CO₂ Scrubber Failure
CHX Failure
Water Processor Failure
Trace Contaminant Filter Saturation
NH₃ Coolant Leak

See It in Action

Every screen built for clarity under pressure

AURA
Digital Twin Dashboard Sensor Detail Trends Alerts
● NOMINAL
Module Status
US Lab
Node 1
Node 2
Columbus
Cupola
Airlock
JLP&JPM
Digital Twin

Live 3D ISS model in Three.js with color-coded module health indicators and interactive orbit controls.

AURA
Digital Twin Dashboard Sensor Detail Trends Alerts
⚠ FAULT
ECLSS System Overview Refresh
US LAB NOMINAL
O2 PP
20.87
CO2 PP
0.38
Humidity
52.1%
Temp
21.5°C
DQN: No Action Needed (96%) LSTM: NOMINAL
COLUMBUS FAULT
⚠ Detected: CO₂ Scrubber Failure
CO2 PP
1.31
Temp
24.8°C
DQN: Replace CO₂ Filter (89%) LSTM: CRITICAL
Dashboard

All locations with live sensor bars, fault banners, and DQN + LSTM status footers.

AURA
Dashboard Sensor Detail Trends Alerts AI Analyst
⚠ FAULT
Location
Columbus ▾
Parameter
CO₂ Partial Pressure ▾
Points
100 ▾
Load
1.31 mmHg ANOMALOUS limit 0.70
1.4 1.0 0.6 0.2 -100 -50 now
TimestampValue
2025-01-01 14:23:021.31 mmHg ⚠
2025-01-01 14:23:010.68 mmHg
Sensor Detail

Historical Chart.js line charts with anomaly-flagged points, nominal band overlays, and data table.

AURA
Dashboard Sensor Detail Trends Alerts AI Analyst
● NOMINAL
Location
US Lab ▾
Readings
100 ▾
Analyze
CO₂ Partial PressureCRITICAL
1.31 mmHg
τ: 0.78p: <0.001slope: +0.12/tick
↑ increasing
Immediate action recommended — CO₂ scrubber saturation.
TemperatureWARNING
24.8 °C
↑ increasing
Trends

Mann-Kendall trend analysis with severity-coded cards showing τ, p-value, slope, and recommendations.

AURA
Sensor Detail Trends Alerts 2 AI Analyst Maintenance
⚠ FAULT
Alert History Refresh Ack All
CRITICAL
CO₂ Scrubber Failure (97.2%) — Columbus
2025-01-01 14:23:01
Ack
CRITICAL
NH₃ Coolant Leak (91.5%) — Node 2
2025-01-01 14:18:44
Acked
WARNING
Humidity Drift — Columbus
2025-01-01 14:09:04
Ack
Alerts

Two-tier alert log with severity badges, confidence scores, timestamps, and per-alert acknowledgement.

AURA
Trends Alerts AI Analyst Maintenance ⚙ Settings
● NOMINAL
AI Analyst
Scope
All Locs ▾
Ollama (local)
Live Anomalies
14:23:02ANOMALOUSCO₂ (97%)
What's wrong in Columbus?
CO₂ Scrubber Failure confirmed. RF 97.2% confidence. CO₂ PP at 1.31 mmHg — limit 0.70. LSTM RUL: 4.2 hrs. DQN: Replace CO₂ filter cartridge.
Ask about anomalies, faults…
AI Analyst

LLM chat with sidebar showing live anomalies — synthesizes all four model outputs into plain-English analysis.

Ready to See Everything?

Explore the full feature breakdown or see rigorous testing validation for every system component.

Full Feature Breakdown Testing & Validation