AI for Engineering Systems

Our research establishes statistically principled and decision-oriented AI methods for complex engineering systems whose performance depends on evolving states, uncertain failure mechanisms, and operational decisions. We integrate domain knowledge with probabilistic modeling, Bayesian inference, Gaussian processes, survival analysis, reinforcement learning, and uncertainty quantification to monitor system behavior, diagnose abnormal operations, predict failure time and failure mode, and prescribe maintenance or operational actions. Applications include multi-component production systems, IoT-enabled smart and connected systems, multistage manufacturing systems, structural damage diagnosis, and service systems where reliable prediction must ultimately support better engineering decisions.

Selected publications

[J11] J. Lee, T. Srikitrungruang, and S. Jahani, "Knowledge-Guided Reinforcement Learning for Preventive Maintenance Planning in Economically Dependent Multi-Component Systems", IISE Transactions, accepted, 2025.
[J12] S. Aghaee Dabanghan Fard, A. Deep, M. Kim, J. Lee, "Bayesian Joint Model of Multi-Sensor and Failure Event Data for Multi-Mode Failure Prediction'', Technometrics, 2026.
[J2] J. Lee, J. Son, S. Zhou, and Y. Chen, "Variation source identification in manufacturing processes using Bayesian approach with sparse variance components prior," IEEE Transactions on Automation Science and Engineering, vol. 17, no. 3, pp. 1469–1485, 2020.
[J5] C. Huang, J. Lee, Y. Zhang, S. Zhou, and J. Tang, "Mixed-input Bayesian Optimization Method for Structural Damage Diagnosis," IEEE Transactions on Reliability, vol 72, no. 2, pp. 678-691, 2023. doi:10.1109/TR.2022.3179602

Preventive Maintenance on multi-component system (IoT-enabled Smart and Connected Production System)
- Goal: Establishing a reinforcement learning method combined with a stochastic model to find the optimal preventive maintenance policy in a multi-components production system.
- Methodologies: Deep Reinforcement learning, deep neural network, stochastic process modeling

This work is done with a Ph.D. student, T. Srikitrungruang and Collaborator, Dr. S. Jahani

[J11] J. Lee, T. Srikitrungruang, and S. Jahani, "Knowledge-Guided Reinforcement Learning for Preventive Maintenance Planning in Economically Dependent Multi-Component Systems", IISE Transactions, accepted, 2025.
- This work was selected as a finalist in the QSR Best Referred Paper Competition in 2025 INFORMS Conference.

Prognostics with diagnostics capability : Real-time failure time and types prediction and uncertainty quantification
- Methodologies: Bayesian Hierarchical Modeling, Convolutional Gaussian Process, Variational Inference, Proportional Cox Hazard model

This work is done with a Ph.D. student, S. Aghaee
[J12] S. Aghaee Dabanghan Fard, A. Deep, M. Kim, J. Lee, "Bayesian Joint Model of Multi-Sensor and Failure Event Data for Multi-Mode Failure Prediction'', Technometrics, 2026.
- This work was selected as a finalist in the QCRE Best Student Paper Competition in 2025 IISE Conference .

IoT-enabled Smart and Connected System

Diagnosis on Faulty Operation leveraging domain knowledge IoT-enabled Smart and Connected Manufacturing System
- Goal: Identifying faulty operations with excessive variation in a multistage manufacturing system; establishing a sparse-variance statistical model.
- Methodologies: Random-effects model, hierarchical Bayesian prior modeling

[J2] J. Lee, J. Son, S. Zhou, and Y. Chen, "Variation source identification in manufacturing processes using Bayesian approach with sparse variance components prior," IEEE Transactions on Automation Science and Engineering, vol. 17, no. 3, pp. 1469–1485, 2020.

Google Sites

Report abuse