The Hu Yang Research Group centers its research on PHM, digital-twin modeling, and system resilience of complex engineering systems. Focusing on high-end equipment in aerospace, energy, transportation, and intelligent manufacturing, the laboratory builds an integrated framework of multimodal intelligence, cyber-physical modeling, and reliability assurance. Its major research areas include:

(1) Multimodal Large Models and Intelligent PHM Algorithms, integrating AI with physical principles for accurate system health prediction and diagnosis;

(2) Digital Twin–Driven Modeling and Intelligent Maintenance Optimization, achieving intelligent operation and maintenance through cyber-physical co-simulation;

(3) System Resilience and Belief Reliability Analysis, quantifying robustness and recovery capabilities through networked system modeling and uncertainty analysis.

Together, these three directions form a unified innovation architecture of Intelligence–Simulation–Assurance, providing theoretical foundations and technological solutions for next-generation engineering systems with higher reliability, adaptability, and autonomy.

1. Multimodal Large Models for Intelligence PHM Algorithms
This research direction focuses on multimodal large models for PHM of complex engineering systems. It establishes a complete technical chain—from data processing to model training and cyber-physical validation—by integrating sensor signals, operational parameters, maintenance logs, and environmental data to achieve cognitive modeling and intelligent optimization for high-reliability systems in aerospace, energy, and transportation domains. The main research topics include:

(1) Multimodal Data Processing for PHM
Developing cleaning, alignment, and representation-learning methods for heterogeneous data; constructing standardized multi-source data processing pipelines and efficient embedding frameworks to achieve unified representation and dynamic fusion of structural, temporal, visual, and textual modalities.

(2) Integrated Modeling of Multimodal PHM Large Models
Building a comprehensive multimodal cognitive model architecture that incorporates a Digital-Twin Module (physics-based constraints), a Knowledge-Enhanced Module (domain knowledge and expert rules), and a Dynamic Fusion Mechanism between them, enabling deep synergy between data-driven and physics-informed modeling.

(3) Multi-Agent Collaborative Training for PHM Large Models
Exploring multi-agent reinforcement learning (MARL) and collaborative optimization strategies for multi-task, multi-scenario, and multi-equipment PHM applications, enabling distributed perception, cooperative reasoning, and coordinated decision-making among intelligent agents.

(4) Cyber-Physical Validation of PHM Multimodal Models
Integrating real and simulated data for hybrid validation under various operational conditions, fault profiles, and mission scenarios to evaluate the model’s robustness, generalization, and engineering applicability—forming a sustainable PHM model verification and iteration framework.

The goal of this research direction is to establish a unified framework that combines data-driven learning, physical constraints, knowledge enhancement, intelligent collaboration, and cyber-physical validation, enabling predictive maintenance and autonomous health management across the entire lifecycle of complex engineered systems.

2. Digital Twin–Driven Modeling, Simulation, and Intelligent Maintenance Optimization
This research direction focuses on digital-twin-driven modeling, simulation, and intelligent maintenance optimization for complex engineering systems. By integrating Model-Based Systems Engineering (MBSE), multi-physics simulation, and data-driven optimization, the laboratory develops real-time virtual representations and predictive decision frameworks to enable high-reliability and efficiency in intelligent maintenance. Main research topics include:

(1) Digital Twin Architecture and Multi-Level Modeling Methods
Developing hierarchical digital-twin frameworks for mission, system, equipment, and component levels, establishing consistent mappings from requirements through functional, logical, physical, and behavioral domains for seamless cyber–physical synchronization.

(2) Multiphysics Modeling and Behavioral Simulation
Creating coupled simulations across mechanical, thermal, electrical, and fluid domains to analyze system responses and degradation mechanisms under realistic or extreme conditions, supporting physics-informed health monitoring and life prediction.

(3) State Assessment and Predictive Analytics Based on Digital Twins
Fusing simulation outputs with real-time sensor data to perform dynamic health evaluation, mission performance prediction, and risk assessment using data assimilation and hybrid modeling techniques.

(4) Intelligent Maintenance and Decision Optimization
Employing reinforcement learning, multi-objective optimization, and evolutionary computation to develop digital-twin-driven maintenance decision models for planning, scheduling, resource management, and contingency operations, achieving an optimal balance between availability and cost.

(5) Cyber–Physical Closed-Loop Validation and Simulation Platform Development
Establishing integrated verification platforms to test algorithms, models, and strategies in closed-loop settings, and developing scalable simulation systems using AnyLogic, Modelica, and Python for full-chain validation from virtual testing to real deployment.

This direction aims to create a “cyber–physical interconnection, intelligent decision-making, and closed-loop optimization” ecosystem for digital twins, enabling visualization, prediction, and optimization of complex systems throughout their lifecycle.

3. Complex System Resilience and Belief Reliability Analysis

This research direction focuses on the System Resilience and Belief Reliability Analysis of complex engineering systems in aerospace, energy, transportation, and intelligent manufacturing.
It integrates complex network theory, AI-based analytics, and digital-twin simulation to uncover the structural vulnerabilities and dynamic evolution mechanisms of large-scale interconnected systems under multi-disturbance and high-uncertainty conditions, establishing a quantitative and verifiable framework for system reliability and safety evaluation.

Key research topics include:

(1) Complex Network Modeling and Coupled System Analysis:
     Investigating the structural characteristics, coupling relationships, and dynamic evolution of complex systems through the construction of multi-layer, multi-domain, and multi-scale network models. The research emphasizes the design and dynamic construction of a OmniLink HyperNetwork, which enables high-dimensional interaction modeling across multiple entities, relations, and temporal scales. On this basis, the Structural Order Entropy is defined and computed to quantify the degree of order, complexity, and structural stability of systems, providing a mathematical foundation for analyzing robustness and evolutionary trends.

(2) Belief Reliability Theory and Uncertainty Quantification:
     Developing reliability modeling methods grounded in belief theory, integrating Bayesian inference, Dempster–Shafer evidence theory, and fuzzy logic to represent, update, and evaluate multi-source uncertainty information dynamically.

(3) System Resilience Evaluation and Metric Development:
     Studying the recovery and robustness characteristics of systems under perturbations, failures, and external shocks; constructing resilience metrics, dynamic models, and evaluation frameworks to support      recoverability and mission continuity analysis.

(4) Resilience Enhancement and Structural Optimization Strategies:
     Designing adaptive control, redundancy configuration, and structural reconfiguration methods for high-risk environments to enhance system-level resilience and implement proactive risk mitigation.

This direction establishes a new interdisciplinary paradigm that combines network science, entropy theory, and AI-driven reliability modeling, providing computationally grounded, predictive, and verifiable methodologies for resilience assurance in critical infrastructure and advanced engineering systems.