Research | Jayant Patil

Areas of Interest

Computational Methods

Efficient approaches for reliability assessment.

Recursive techniques for functional reliability assessment of radial infrastructure systems
Closed-form techniques for reliability assessment for multi-component and multi-state systems
High-performance computing for creating large-scale synthetic datasets for training ML models

Machine Learning for Reliability Assessment

Novel ML approaches for reliability analysis of complex networked systems.

Augmented AI framework with Graph Neural Networks for estimating all-terminal reliability
ML informed stratified sampling for estimating source-terminal reliability
CNN and CAE based sensor fault detection, localization, and correction
Computer Vision techniques for utility asset digitization and condition assessment

Electric Grid Modeling and Analysis

Real-world data and synthetic models to analyze electric grid reliability.

Parametric synthetic power flow models for power transmission systems
Computer Vision for detecting utility assets in public streetview imagery
Radial infrastructure models for power distribution systems
Equity analysis of grid reliability leveraging web-scraped outage data to identify disparities in service quality

Featured Projects

Electric Grid Modeling using Network Science & Computer Vision

Multi-Scale Electric Power System Modeling

Developed innovative techniques to model integrated electric power systems from interconnection level grid to distribution level transformers, for building-level power availability analysis.

Created reduced-sized transmission system models using equivalencing techniques
Applied Voronoi polygons to identify substation service areas
Identified utility poles from street view imagery using Mask-RCNN model
Demonstrated approach using Lumberton, NC and Galveston Island, TX case studies

Network Science Computer Vision GIS Deep Learning Python TensorFlow NetworkX

Reliability Assessment of Radial Infrastructure Networks

Recursive Technique

Developed a novel recursive method for efficiently computing system-level functional reliability of radial infrastructure networks, significantly outperforming traditional Monte Carlo Simulation methods.

Created an application-agnostic formulation applicable to electric power distribution, water systems, solar farms, and other radial infrastructure
Validated with real and synthetic case studies including balanced trees, synthetic solar farms, and real power distribution networks
Achieved 1000x performance improvement in computational time over traditional Monte Carlo methods

Dynamic Programming Python NetworkX Pandas

Assessing Equity in Electric Power Outages

Socioeconomic Analysis of Power Disruptions

Conducted a comprehensive equity analysis of electric power outages in a major US county, examining relationships between outage patterns and socioeconomic factors across different block groups.

Web scraped outage data using python requests and deployed using GitHub Actions
Analyzed detailed power outage data using geospatial techniques
Discovered significant inequity in power disruptions (Gini coefficient of 0.63)
Discovered that higher-income areas experience fewer disruptions per capita and shorter outage durations than middle and low-income areas
Provided actionable insights for utilities and policymakers to improve equity in outage recovery

Web Scraping Equity Analysis Statistical Analysis Python GitHub Actions GeoPandas

Augmented AI Framework with GNN for Reliability

All-Terminal Reliability Approximation using Graph Neural Networks

Designed a novel Augmented Artificial Intelligence framework combining Graph Neural Networks with traditional estimation methods to improve the accuracy of ATR predictions for power networks.

Created one of the largest synthetic dataset of network reliability with 1.1M graphs using high-performance computing
Engineered PNAConv-based GNN models for ATR prediction achieving 99.8% coefficient of determination with actual ATR values
Reduced mean absolute error by 78% (0.681×10⁻³ to 0.147×10⁻³) through smart selection of GNN model vs. traditional methods
Established new benchmarks for network reliability assessment using ML techniques

Graph Neural Networks Parallel Computing Python NetworkX PyTorch Geometric

CNN and CAE based Sensor Fault Detection, Localization, and Correction

Deep Learning for automatically detecting and correcting faults in sensor data

Developed a novel deep learning framework for detecting, classifying, localizing, and reconstructing faults in sensor data of structural monitoring systems using Convolutional Neural Networks and Convolutional Autoencoders.

Achieved 98.7% accuracy in fault type detection across missing, spiky, random, and drift type faults
Demonstrated 99% accuracy in sensor data reconstruction
Published in Mechanical Systems and Signal Processing journal with 160+ citations till date

CNN Convolutional Autoencoders MATLAB Python Keras TensorFlow Talos

Reliability of Wind Turbine Systems

Closed-Form technique for system reliability assessment

Created an efficient method for reliability assessment of systems with multi-state components, providing flexible system failure definitions and component importance assessment.

Reduced computational time by 1000x compared to Monte Carlo Simulation methods
Evaluated component importance using reliability and cost-based metrics
Demonstrated efficacy with historical wind turbine failure data

Reliability Analysis MATLAB

Publications

Patil, J. & Dueñas-Osorio, L. (In Preparation). "Recursive technique for efficient computation of functional reliability of radial infrastructure systems."
Patil, J., Lin, C.Y., Cha, E.J., & Dueñas-Osorio, L. (Under Review). "Integrating Synthetic Power Flow and Computer Vision Models for Assessing Building-Level Outages."
Herkal, S., Patil, J., & Dueñas-Osorio, L. (Under Review). Closed-form reliability and risk assessment for the upkeep of multi-component and multi-state wind turbine systems.
Nofal, O., Rosenheim, N., Kameshwar, S., Patil, J., Zhou, X., van de Lindt, J. W., ... & Wang, C. (2025). "Methodology for Interdependent Population–Building–Infrastructure Posthazard Functionality Assessment for Communities." Journal of Structural Engineering, 151(5), 04025048.
Nofal, O., Rosenheim, N., Kameshwar, S., Patil, J., Zhou, X., van de Lindt, J. W., ... & Jeon, H. (2024). "Community‐level post‐hazard functionality methodology for buildings exposed to floods." Computer-Aided Civil and Infrastructure Engineering.
Nofal, O., Rosenheim, N., Patil, J., Zhou, X., Kameshwar, S., van de Lindt, J. W., & Duenas-Osorio, L. (2023). "Community-level approach for a socio-physical flood post-hazard functionality assessment." In ASCE Inspire 2023 (pp. 339-348).
Jana, D., Patil, J., Herkal, S., Nagarajaiah, S., & Duenas-Osorio, L. (2022). "CNN and Convolutional Autoencoder (CAE) based real-time sensor fault detection, localization, and correction." Mechanical Systems and Signal Processing, 169, 108723.
Nofal, O., Rosenheim, N., Patil, J., Zhou, X., van de Lindt, J. W., Duenas-Osorio, L., & Cha, E. J. (2022). "Interdependent Households-Buildings-Networks Community-Level Post-Hazard Functionality Assessment Methodology." Proceedings of the 8th International Symposium on Reliability Engineering and Risk Management (ISRERM 2022), Hanover, Germany.
Birchfield, A. B., Patil, J., Paredes, R., & Dueñas-Osorio, L. (2021, April). "Preliminary Analysis of Network Fragility and Resilience in Large Electric Grids." In 2021 IEEE Power and Energy Conference at Illinois (PECI) (pp. 1-6). IEEE.