Skip to content

sets1623/AI_Driven_Digital_Twin_Report.pdf

Repository files navigation

AI-Driven Digital Twin (Level-1)

Multi-Physics System Modeling, Estimation, and Health Monitoring


Project Overview

This project implements a Level-1 Digital Twin framework for an ISS-inspired space system, focusing on system-level modeling, estimation, control logic, and health monitoring.

The Digital Twin combines:

  • Electrical (battery) dynamics
  • Thermal dynamics
  • Life support (ECLSS-inspired) dynamics
  • State estimation and fault detection
  • Health degradation and Remaining Useful Life (RUL) estimation

The objective is to study how Digital Twin architectures can be built for complex engineering systems, with emphasis on clarity, correctness, and scalability, rather than high-fidelity mission replication.


What This Project Does (Clearly Stated)

This project creates a closed-loop simulation environment that:

  1. Models nonlinear multi-physics system dynamics
  2. Estimates internal system states using an Extended Kalman Filter (EKF)
  3. Applies optimization-based control logic (mini-MPC style)
  4. Detects abnormal behavior using statistical fault detection (NIS)
  5. Predicts battery health degradation and Remaining Useful Life (RUL)

All components are implemented in a modular and extensible architecture.


System Architecture (Level-1)

State Vector (8 States)

The system is represented using a unified nonlinear state vector:

  • SOC – Battery state of charge
  • V_bus – Electrical bus voltage (simplified)
  • T_core – Internal system temperature
  • T_radiator – Radiator temperature
  • O2_mass – Cabin oxygen mass
  • CO2_mass – Cabin CO₂ mass
  • P_cabin – Cabin pressure
  • H_battery – Battery health index

This structure enables future expansion while remaining computationally simple.


Subsystem Modeling

1. Electrical (Battery) System

  • Nonlinear SOC dynamics
  • Charging and discharging efficiency
  • Solar input disturbance (orbital sunlight)
  • Load variation
  • Health degradation linked to SOC stress

2. Thermal System

  • Heat generation from electrical load
  • Radiative heat rejection
  • Pump-controlled heat transfer
  • Core and radiator temperature dynamics

3. Life Support System (Simplified ECLSS)

  • Oxygen generation and consumption
  • CO₂ generation and scrubbing
  • Cabin pressure variation
  • Crew metabolic load as disturbance

Estimation and Monitoring

Extended Kalman Filter (EKF)

  • Estimates SOC and battery health
  • Handles nonlinear system behavior
  • Includes process and measurement noise modeling

Fault Detection

  • Uses Normalized Innovation Squared (NIS)
  • Detects degradation faults after fault injection
  • Evaluates filter consistency statistically

Health & Prognostics

  • Continuous battery health degradation modeling
  • Smoothed Remaining Useful Life (RUL) estimation
  • Suitable for predictive maintenance studies

Control Logic

  • Optimization-based control using grid search

  • Minimizes:

    • SOC deviation
    • Health degradation rate
    • Control effort
  • Enforces safety constraints on SOC and health

This represents a simplified Model Predictive Control (MPC) concept.


Numerical Integration

  • System dynamics are propagated using a Runge–Kutta 4th Order (RK4) solver

  • Ensures numerical stability and accuracy

  • Clean separation between:

    • State definition
    • System dynamics
    • Integration logic

Project Structure

AI_Digital_Twin/
│
├── main_simulation.py          # Full closed-loop simulation
│
├── state_vector.py             # 8-state system definition
├── system_dynamics.py          # Nonlinear multi-physics model
├── rk4_solver.py               # RK4 numerical integrator
│
├── System_Architecture_Level1.md
│
├── requirements.txt
└── README.md

Compiled files (*.cpython-313) are generated artifacts and not part of core logic.


Tools and Technologies

  • Python
  • NumPy, SciPy
  • Matplotlib (visualization)
  • Control systems & estimation theory

The focus is on engineering modeling and reasoning, not software complexity.


What This Project Is

✔ A research-oriented Digital Twin framework ✔ A system-level engineering study ✔ A learning platform for estimation, control, and health monitoring ✔ Suitable for:

  • Final-year projects
  • Research internships
  • Control systems & Digital Twin portfolios

What This Project Is NOT

✘ Not a real ISS model ✘ Not connected to live space data ✘ Not a mission-certified flight system ✘ Not a black-box AI project


Key Learning Outcomes

  • Multi-physics system modeling
  • State-space formulation
  • EKF implementation
  • Fault detection using statistical methods
  • Health degradation & RUL estimation
  • Digital Twin architecture design

Future Scope

  • Higher-fidelity physical models
  • Coupled estimation across all states
  • Advanced MPC formulations
  • Integration with real open telemetry datasets
  • Extension to lunar or Mars habitat systems

Author

Tamil Selvan Electrical & Electronics Engineering Interests: Digital Twins • Control Systems • Estimation • Space Systems


License

MIT License – Free for educational and research use.


Final Note for Reviewers

This project emphasizes clarity, correctness, and engineering discipline in Digital Twin development, serving as a solid foundation for advanced research and system expansion.


About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages