Ilias Bilionis
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Ilias Bilionis has a total of 107 co-authors. Here are the top co-authors by number of shared publications:
- Jitesh H. Panchal
Purdue University
10 shared publications - Shirley J. Dyke
Lyles School of Civil and Construction Engineering, Purdue University, 585 Purdue Mall, West Lafayette, IN 47907-2088, USA
6 shared publications Alejandro Strachan
4 shared publications- Atharva Hans
School of Mechanical Engineering Purdue University 47906 West Lafayette IN USA
4 shared publications - Dionysios Aliprantis
Elmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA
4 shared publications Jitesh Panchal
School of Mechanical Engineering Purdue University 47907-2088 West Lafayette Indiana
4 shared publications
Latest Publications
- Open Access
- 6 August 2026
- Published by Elsevier in Materials & Design
- Vol. 269
Active-learning-guided discovery of alumina-forming refractory complex concentrated alloys
- 6 August 2026
- Published by Elsevier in Materials & Design
- Vol. 269
Open AccessActive-learning-guided discovery of alumina-forming refractory complex concentrated alloys
- 5 August 2026
- Published by Elsevier in Acta Materialia
- Vol. 318
Bayesian identification of fibrous insulation thermal conductivity towards design of spacecraft thermal protection systems
- 5 August 2026
- Published by Elsevier in Acta Materialia
- Vol. 318
Bayesian identification of fibrous insulation thermal conductivity towards design of spacecraft thermal protection systems
- 31 July 2026
- Published by SAGE Publications in International Journal of Engine Research
Modeling diesel output particulate matter as the Ornstein-Uhlenbeck process
Diesel engine particulate matter (PM) is one of the most challenging emission constituents to predict. As engines become cleaner and emissions levels drop, manufacturers need reliable methods to quantify the PM generated by production engines. Production engines typically do not have calibrated, time-resolved sensors for continuous engine-out PM mass estimation, so manufacturers rely on predictive models using available engine state measurements. In practice, this requires a computationally inexpensive model that provides PM estimates with calibrated uncertainty. Complex, multiscale physics make mechanistic models intractable and traditional data-driven methods struggle in transient drive cycles due to the stochastic nature of PM generation. Using high-frequency experimental PM measurements from transient engine tests, we introduce a novel PM model based on the Ornstein-Uhlenbeck (OU) process. The OU process is a mean-reverting stochastic process commonly used in financial modeling and is defined as the solution to a stochastic differential equation (SDE). We modify the OU process by parameterizing the terms of the SDE as functions of the engine state, which are then fit with a maximum likelihood estimate. In a synthetic example, we verify the ability of our model to learn a time-varying, parameterized OU process. We then train the model using real experimental data designed to dynamically cover the engine operating space and test the trained model on EPA-regulated drive cycles. For most drive cycles, we find the method accurately predicts cumulative PM mass output across time.
- 31 July 2026
- Published by SAGE Publications in International Journal of Engine Research
Modeling diesel output particulate matter as the Ornstein-Uhlenbeck process
Diesel engine particulate matter (PM) is one of the most challenging emission constituents to predict. As engines become cleaner and emissions levels drop, manufacturers need reliable methods to quantify the PM generated by production engines. Production engines typically do not have calibrated, time-resolved sensors for continuous engine-out PM mass estimation, so manufacturers rely on predictive models using available engine state measurements. In practice, this requires a computationally inexpensive model that provides PM estimates with calibrated uncertainty. Complex, multiscale physics make mechanistic models intractable and traditional data-driven methods struggle in transient drive cycles due to the stochastic nature of PM generation. Using high-frequency experimental PM measurements from transient engine tests, we introduce a novel PM model based on the Ornstein-Uhlenbeck (OU) process. The OU process is a mean-reverting stochastic process commonly used in financial modeling and is defined as the solution to a stochastic differential equation (SDE). We modify the OU process by parameterizing the terms of the SDE as functions of the engine state, which are then fit with a maximum likelihood estimate. In a synthetic example, we verify the ability of our model to learn a time-varying, parameterized OU process. We then train the model using real experimental data designed to dynamically cover the engine operating space and test the trained model on EPA-regulated drive cycles. For most drive cycles, we find the method accurately predicts cumulative PM mass output across time.
- 26 July 2026
- Published by Elsevier in International Journal of Thermal Sciences
- Vol. 230
Entropy switch for Bayesian thermal network discovery
- 26 July 2026
- Published by Elsevier in International Journal of Thermal Sciences
- Vol. 230
Entropy switch for Bayesian thermal network discovery
- Open Access
- 25 July 2026
- Published by Elsevier in Medical Image Analysis
- Vol. 114, 104230
Patient-specific unsupervised neural reconstruction of cardiac blood flow from 4D flow MRI
- 25 July 2026
- Published by Elsevier in Medical Image Analysis
- Vol. 114, 104230
Open AccessPatient-specific unsupervised neural reconstruction of cardiac blood flow from 4D flow MRI
- 14 July 2026
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM/ASA Journal on Uncertainty Quantification
- Vol. 14 (3) , 1008-1044
An Interpretation of the Brownian Bridge as a Physics-Informed Prior for the Poisson Equation
Many inverse problems require reconstructing physical fields from limited and noisy data while incorporating known governing equations. A growing body of work within probabilistic numerics formalizes such tasks via Bayesian inference in function spaces by assigning a physically meaningful prior to the latent field. In this work, we demonstrate that Brownian bridge Gaussian processes can be viewed as softly enforced physics-constrained priors for the Poisson equation. We first show equivalence between the variational problem associated with the Poisson equation and a kernel ridge regression objective. Then, through the connection between Gaussian process regression and kernel methods, we identify a Gaussian process for which the posterior mean function and the minimizer to the variational problem agree, thereby placing this PDE-based regularization within a fully Bayesian framework. This connection allows us to probe different theoretical questions, such as convergence and behavior of inverse problems. We then develop a finite-dimensional representation in function space and prove convergence of the projected prior and resulting posterior in Wasserstein distance. Finally, we connect the method to the important problem of identifying model-form error in applications, providing a diagnostic for model misspecification.
- 14 July 2026
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM/ASA Journal on Uncertainty Quantification
- Vol. 14 (3) , 1008-1044
An Interpretation of the Brownian Bridge as a Physics-Informed Prior for the Poisson Equation
Many inverse problems require reconstructing physical fields from limited and noisy data while incorporating known governing equations. A growing body of work within probabilistic numerics formalizes such tasks via Bayesian inference in function spaces by assigning a physically meaningful prior to the latent field. In this work, we demonstrate that Brownian bridge Gaussian processes can be viewed as softly enforced physics-constrained priors for the Poisson equation. We first show equivalence between the variational problem associated with the Poisson equation and a kernel ridge regression objective. Then, through the connection between Gaussian process regression and kernel methods, we identify a Gaussian process for which the posterior mean function and the minimizer to the variational problem agree, thereby placing this PDE-based regularization within a fully Bayesian framework. This connection allows us to probe different theoretical questions, such as convergence and behavior of inverse problems. We then develop a finite-dimensional representation in function space and prove convergence of the projected prior and resulting posterior in Wasserstein distance. Finally, we connect the method to the important problem of identifying model-form error in applications, providing a diagnostic for model misspecification.
- Open Access
- 4 June 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 331
Experimental study on the impact of long communication delays on autonomous decision-making in deep space habitats
- 4 June 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 331
Open AccessExperimental study on the impact of long communication delays on autonomous decision-making in deep space habitats
- Open Access
- 26 May 2026
- Published by Elsevier in SSRN Electronic Journal
Active Learning Discovery of High Temperature Oxidation Resistant Refractory Complex Concentrated Alloys
Refractory complex concentrated alloys (RCCAs) are of significant interest for advanced high-temperature applications due to their wide compositional range and potential for attractive mechanical properties and oxidation resistance. However, their compositional complexity challenges conventional alloy discovery approaches. Here, an active learning framework integrating Gaussian process regression with Bayesian global optimization is introduced to accelerate the identification of oxidation-resistant RCCAs. Focusing on aluminum-containing quaternary systems, alloy and oxide descriptors were used to predict oxidation performance at 1000°C. Beginning with a dataset of 81 experimentally validated alloys, the framework iteratively selected batches of five alloys, balancing exploration and exploitation to minimize experimental costs. After six iterations, two alloys—nominal Al₃₀Mo₅Ti₁₅Cr₅₀ and Al₄₀Mo₅Ti₃₀Cr₂₅—exhibited specific mass gains below 1 mg/cm² after 24 h at 1000°C in air, forming adherent external α-Al₂O₃ scales with parabolic oxidation kinetics consistent with diffusion-limited scale growth. Multiobjective analysis further demonstrates that these alloys achieve high specific hardness (>0.12 HV·m³/kg) and thermal expansion compatibility with thermal barrier coating systems, suggesting their potential as candidate bond coat materials, warranting further investigation. This work underscores the efficacy of active learning in traversing complex compositional landscapes and offers a scalable strategy for developing advanced materials for extreme environments.
- 26 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessActive Learning Discovery of High Temperature Oxidation Resistant Refractory Complex Concentrated Alloys
Refractory complex concentrated alloys (RCCAs) are of significant interest for advanced high-temperature applications due to their wide compositional range and potential for attractive mechanical properties and oxidation resistance. However, their compositional complexity challenges conventional alloy discovery approaches. Here, an active learning framework integrating Gaussian process regression with Bayesian global optimization is introduced to accelerate the identification of oxidation-resistant RCCAs. Focusing on aluminum-containing quaternary systems, alloy and oxide descriptors were used to predict oxidation performance at 1000°C. Beginning with a dataset of 81 experimentally validated alloys, the framework iteratively selected batches of five alloys, balancing exploration and exploitation to minimize experimental costs. After six iterations, two alloys—nominal Al₃₀Mo₅Ti₁₅Cr₅₀ and Al₄₀Mo₅Ti₃₀Cr₂₅—exhibited specific mass gains below 1 mg/cm² after 24 h at 1000°C in air, forming adherent external α-Al₂O₃ scales with parabolic oxidation kinetics consistent with diffusion-limited scale growth. Multiobjective analysis further demonstrates that these alloys achieve high specific hardness (>0.12 HV·m³/kg) and thermal expansion compatibility with thermal barrier coating systems, suggesting their potential as candidate bond coat materials, warranting further investigation. This work underscores the efficacy of active learning in traversing complex compositional landscapes and offers a scalable strategy for developing advanced materials for extreme environments.
- Open Access
- 17 May 2026
- Published by Elsevier in SSRN Electronic Journal
End-to-end Framework for Autonomous Building Control Deployment via Domain-Randomized Reinforcement Learning and Arbitration Policy Selection
The paper presents an end-to-end framework for autonomous building control that integrates domain-randomized reinforcement learning (DR-RL) with Arbitration Policy Selection (APS) to handle variability across policy development, adaptation, and long-term operation under real deployment conditions. During development, domain randomization generates a diverse set of plausible building environments from limited prior information, enabling pretraining that exposes policies to variability in building models without requiring source–target similarity assumptions. During initial field deployment, APS automatically selects the most suitable pretrained policy based on short-term operational performance, mitigating negative transfer and eliminating the need for retraining or accurate target models. During long-term operation, the learning-based control policies continue to adapt to previously unseen conditions, enabling sustained performance under evolving real-world dynamics without requiring manual intervention.The proposed approach is first validated using simulations for a single-zone and multi-zone DOE reference building, where APS-selected policies achieve 11.5%–50.6% improvements in initial performance and 37.4%–50.0% gains in long-term performance, compared to training from scratch, while reducing adaptation time from multiple weeks to less than two days. The framework is further evaluated in an actual office building, achieving 7.8%–10% reward improvement and 13.6%–15.9% energy reduction over baseline control while maintaining thermal comfort. Following a short adaptation phase, the approach rapidly stabilizes and consistently outperforms baseline rule-based control and RL policy trained on nominal building model) under real-world deployment conditions. Beyond performance improvement, the developed framework reduces setup to a minimal configuration process requiring less than one day of low-dimensional input specification, after which policy generation, evaluation, and selection are performed automatically. In both simulation and field deployment, the results show that combining DR-RL with APS provides a practical pathway toward scalable deployment of advanced control systems and offers a foundation for AI-driven building operation with minimal human intervention.
- 17 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessEnd-to-end Framework for Autonomous Building Control Deployment via Domain-Randomized Reinforcement Learning and Arbitration Policy Selection
The paper presents an end-to-end framework for autonomous building control that integrates domain-randomized reinforcement learning (DR-RL) with Arbitration Policy Selection (APS) to handle variability across policy development, adaptation, and long-term operation under real deployment conditions. During development, domain randomization generates a diverse set of plausible building environments from limited prior information, enabling pretraining that exposes policies to variability in building models without requiring source–target similarity assumptions. During initial field deployment, APS automatically selects the most suitable pretrained policy based on short-term operational performance, mitigating negative transfer and eliminating the need for retraining or accurate target models. During long-term operation, the learning-based control policies continue to adapt to previously unseen conditions, enabling sustained performance under evolving real-world dynamics without requiring manual intervention.The proposed approach is first validated using simulations for a single-zone and multi-zone DOE reference building, where APS-selected policies achieve 11.5%–50.6% improvements in initial performance and 37.4%–50.0% gains in long-term performance, compared to training from scratch, while reducing adaptation time from multiple weeks to less than two days. The framework is further evaluated in an actual office building, achieving 7.8%–10% reward improvement and 13.6%–15.9% energy reduction over baseline control while maintaining thermal comfort. Following a short adaptation phase, the approach rapidly stabilizes and consistently outperforms baseline rule-based control and RL policy trained on nominal building model) under real-world deployment conditions. Beyond performance improvement, the developed framework reduces setup to a minimal configuration process requiring less than one day of low-dimensional input specification, after which policy generation, evaluation, and selection are performed automatically. In both simulation and field deployment, the results show that combining DR-RL with APS provides a practical pathway toward scalable deployment of advanced control systems and offers a foundation for AI-driven building operation with minimal human intervention.
- 4 May 2026
- Published by SAGE Publications in International Journal of Engine Research
Bayesian calibration of engine-out NOx models for engine-to-engine transferability
Accurate prediction of engine-out NOx is essential for meeting stringent emissions regulations and optimizing engine performance. Traditional approaches rely on models trained on data from a small number of engines, which can be insufficient in generalizing across an entire population of engines due to sensor biases and variations in input conditions. In real world applications, these models require tuning or calibration to maintain acceptable error tolerance when applied to other engines. This highlights the need for models that can adapt with minimal adjustments to accommodate engine-to-engine variability and sensor discrepancies. While previous studies have explored machine learning methods for predicting engine-out NOx, these approaches often fail to generalize reliably across different engines and operating environments. To address these issues, we propose a Bayesian calibration framework that combines Gaussian processes (GP) with approximate Bayesian computation to infer and correct sensor biases. Starting with a pretrained model developed using nominal engine data, our method identifies engine specific sensor biases and recalibrates predictions accordingly. By incorporating these inferred biases, our approach generates posterior predictive distributions for engine-out NOx on unseen test data, achieving high accuracy without retraining the model. Our results demonstrate that this transferable modeling approach significantly improves the accuracy of predictions compared to conventional non-adaptive GP models, effectively addressing engine-to-engine variability and improving model generalizability.
- 4 May 2026
- Published by SAGE Publications in International Journal of Engine Research
Bayesian calibration of engine-out NOx models for engine-to-engine transferability
Accurate prediction of engine-out NOx is essential for meeting stringent emissions regulations and optimizing engine performance. Traditional approaches rely on models trained on data from a small number of engines, which can be insufficient in generalizing across an entire population of engines due to sensor biases and variations in input conditions. In real world applications, these models require tuning or calibration to maintain acceptable error tolerance when applied to other engines. This highlights the need for models that can adapt with minimal adjustments to accommodate engine-to-engine variability and sensor discrepancies. While previous studies have explored machine learning methods for predicting engine-out NOx, these approaches often fail to generalize reliably across different engines and operating environments. To address these issues, we propose a Bayesian calibration framework that combines Gaussian processes (GP) with approximate Bayesian computation to infer and correct sensor biases. Starting with a pretrained model developed using nominal engine data, our method identifies engine specific sensor biases and recalibrates predictions accordingly. By incorporating these inferred biases, our approach generates posterior predictive distributions for engine-out NOx on unseen test data, achieving high accuracy without retraining the model. Our results demonstrate that this transferable modeling approach significantly improves the accuracy of predictions compared to conventional non-adaptive GP models, effectively addressing engine-to-engine variability and improving model generalizability.