Research at DAIL

Data-driven intelligence for complex systems

We develop practical and reliable AI-enabled methods for nuclear engineering, biomedical applications, materials design and analysis, and neuroimaging.

Statistical signal processingMachine learningDeep learningPhysics-based modelingAI agentsMultimodal data

What we do

From multimodal signals to reliable decisions

Our lab combines statistical signal processing, machine learning, deep learning, and AI agents to analyze ECG, EEG, fMRI, advanced neural imaging, materials data, and nuclear power plant operational signals.

Our goal is to enable intelligent diagnosis, biometric security, brain functional analysis, materials discovery, and decision support in complex real-world systems.

01MeasureAcquire and reconstruct multimodal signals
02ModelCombine physical knowledge and data
03InterpretIdentify patterns, events, and mechanisms
04SupportDeliver reliable and actionable guidance

Current research

Our methods move across engineered and living systems while sharing a common focus on trustworthy analysis and interpretation.

01

Nuclear engineering

Digital twin development for NPPs and SMRs

Digital twins for nuclear power plants and small modular reactors enhance safety, resilience, and operational efficiency by connecting physical simulation with intelligent monitoring.

  • SMR simulator construction with physical reactor models
  • AI/ML event detection, classification, and localization
  • Fault-tolerant monitoring for reliable diagnosis
  • Optimization and LLM-enabled decision support
  • Digital twin integration, connectivity, and visualization
Digital
Twin
SimulatorAI / MLMonitoringDecision
02

Industrial applications

Smart systems for industrial challenges

We combine AI, algorithms, and physics-based modeling to improve the safety, efficiency, and innovation of complex industrial systems.

  • Battery state-of-charge estimation using residual convolutional neural networks
  • Spent fuel pool loading-pattern optimization with integrated CFD and genetic algorithms
  • Water-grid leakage detection and localization using LSTM and MUSIC-based subspace projection
  • LLM-guided ML and GA inverse design for high-entropy alloys
  • AI and image-based characterization of atomic-level material structures
DataMeasure
→
ModelAnalyze
→
AIOptimize
03

Biometric security

Biometric recognition using ECG

ECG signals contain distinctive physiological characteristics that make them suitable for identity recognition while introducing important privacy and security questions.

  • Deep-learning and machine-learning identification
  • Identity verification and systematic performance evaluation
  • Robustness analysis under real-world variation
  • Attacks that threaten privacy or shift decision boundaries
  • Security-aware biometric system design
ECG→Feature→Model→Identity
04

Domain-adaptive intelligence

AI agents for high-stakes decisions

We combine large language models, retrieval-augmented generation, and multi-agent systems to solve complex knowledge and decision-making challenges.

  • Executable decision logic for emergency procedure support
  • Reduced operator cognitive load during major incidents
  • RAG-driven nuclear knowledge query
  • Multi-agent retrieval and analysis of regulations and procedures
  • Dual-agent reflection for standardized materials data extraction
AI
Agent
LLMRAGToolsHuman
05

Neuroscience

Brain research and neuroimaging

We develop AI-driven imaging and brain-signal analysis methods to study neural dynamics and support biomedical applications.

  • AI-based denoising for clearer four-dimensional neural imaging
  • Analysis of locomotion velocity and cerebellar signals
  • Temporal delay and neural response dynamics
  • EEG source reconstruction and localization using MUSIC
  • High-frequency neural dynamics during movement
Measured signals
Reconstruct and localize
Neural dynamics
06

Biomedical applications

Cognitive science and clinical research

We integrate neuroimaging, statistical analysis, and AI to establish objective biomarkers for brain function and clinical conditions.

  • Entropy-based analysis of resting-state fMRI
  • Late-life depression, suicidality, and cognitive performance
  • Deep-learning prediction of clinical severity
  • RS-EEG functional connectivity and phase synchronization
  • PLI/WPLI noise reduction and SVM-based prediction
fMRIEEG
Statistics + AI
Objective
biomarkers

Collaboration

Have a complex data problem?

We welcome academic and industrial collaborations in nuclear engineering, smart systems, biomedical signals, materials, neuroimaging, and AI.

Start a conversation