Venkataramana Runkana - Academia.edu (original) (raw)

Uploads

Papers by Venkataramana Runkana

Research paper thumbnail of HyperLoRA for PDEs

arXiv (Cornell University), Aug 17, 2023

Physics-informed neural networks (PINNs) have been widely used to develop neural surrogates for s... more Physics-informed neural networks (PINNs) have been widely used to develop neural surrogates for solutions of Partial Differential Equations. A drawback of PINNs is that they have to be retrained with every change in initial-boundary conditions and PDE coefficients. The Hypernetwork, a model-based meta learning technique, takes in a parameterized task embedding as input and predicts the weights of PINN as output. Predicting weights of a neural network however, is a highdimensional regression problem, and hypernetworks perform sub-optimally while predicting parameters for large base networks. To circumvent this issue, we use a low ranked adaptation (LoRA) formulation to decompose every layer of the base network into low-ranked tensors and use hypernetworks to predict the low-ranked tensors. Despite the reduced dimensionality of the resulting weight-regression problem, LoRAbased Hypernetworks violate the underlying physics of the given task. We demonstrate that the generalization capabilities of LoRA-based hypernetworks drastically improve when trained with an additional physics-informed loss component (HyperPINN) to satisfy the governing differential equations. We observe that LoRA-based HyperPINN training allows us to learn fast solutions for parameterized PDEs like Burger's equation and Navier Stokes: Kovasznay flow, while having an 8x reduction in prediction parameters on average without compromising on accuracy when compared to all other baselines.

Research paper thumbnail of Reduced-order modeling of conjugate heat transfer in lithium-ion batteries

International Journal of Heat and Mass Transfer/International journal of heat and mass transfer, Aug 1, 2024

Research paper thumbnail of Virtual metrology for chemical mechanical planarization of semiconductor wafers

Journal of intelligent manufacturing, Mar 13, 2024

Research paper thumbnail of metafur: Digital Twin System of a Blast Furnace

Transactions of the Indian Institute of Metals, Jun 28, 2024

Research paper thumbnail of Machine Learning Based Modeling of Silicon Content of Molten Iron from a Blast Furnace

Research paper thumbnail of Machine learning based knowledge discovery and modeling of silicon content of molten iron from a blast furnace

Research paper thumbnail of Multi-Modal Instruction-Tuning Small-Scale Language-and-Vision Assistant for Semiconductor Electron Micrograph Analysis

Proceedings of the AAAI symposium series, May 20, 2024

We present a novel framework for analyzing and interpreting electron microscopy images in semicon... more We present a novel framework for analyzing and interpreting electron microscopy images in semiconductor manufacturing using vision-language instruction tuning. The framework employs a unique teacher-student approach, leveraging pretrained multimodal large language models such as GPT-4 to generate instruction-following data for zero-shot visual question answering (VQA) and classification tasks, customizing smaller multimodal models (SMMs) for microscopy image analysis, resulting in an instruction-tuned language-and-vision assistant. Our framework merges knowledge engineering with machine learning to integrate domain-specific expertise from larger to smaller multimodal models within this specialized field, greatly reducing the need for extensive human labeling. Our study presents a secure, cost-effective, and customizable approach for analyzing microscopy images, addressing the challenges of adopting proprietary models in semiconductor manufacturing.

Research paper thumbnail of Droplet impingement on a solid surface: Parametrization and asymmetry of dynamic contact angle model

Physics of Fluids, Jun 1, 2023

The study of the spreading behavior of droplets impinging on solid surfaces is of importance to a... more The study of the spreading behavior of droplets impinging on solid surfaces is of importance to applications such as inkjet printing and spray coating. The contact angle is an important parameter that influences the spreading behavior of droplets upon impingement on a solid surface. Computational fluid dynamics simulations studying droplet dynamics require a dynamic contact angle (DCA) model with an appropriate set of parameters to simulate the experimental system of interest. We propose a scheme to parameterize a DCA model and tune its parameters for systems of different levels of wettability. The developed DCA models show the varied response for advancing and receding phases of the droplet motion to emphasize the asymmetric nature of the relation between the contact angle and contact line velocity. These models enable accurate simulation of droplet impingement for a wide range of Weber number (We) and Reynolds number (Re) values. The proposed scheme helps tune the parameters of the DCA model in a systematic and quick manner, thereby enabling one to explore the design space better and also reduce the time to design and develop novel fluids and devices for applications dealing with impinging droplets.

Research paper thumbnail of Mathematical Modeling of Coagulation and Flocculation of Colloidal Suspensions Incorporating the Influence of Surface Forces

Research paper thumbnail of Advances in bioreactor control for production of biotherapeutic products

Biotechnology and Bioengineering, Mar 1, 2023

Advanced control strategies are well established in chemical, pharmaceutical, and food processing... more Advanced control strategies are well established in chemical, pharmaceutical, and food processing industries. Over the past decade, the application of these strategies is being explored for control of bioreactors for manufacturing of biotherapeutics. Most of the industrial bioreactor control strategies apply classical control techniques, with the control system designed for the facility at hand. However, with the recent progress in sensors, machinery, and industrial internet of things, and advancements in deeper understanding of the biological processes, coupled with the requirement of flexible production, the need to develop a robust and advanced process control system that can ease process intensification has emerged. This has further fuelled the development of advanced monitoring approaches, modeling techniques, process analytical technologies, and soft sensors. It is seen that proper application of these concepts can significantly improve bioreactor process performance, productivity, and reproducibility. This review is on the recent advancements in bioreactor control and its related aspects along with the associated challenges. This study also offers an insight into the future prospects for development of control strategies that can be designed for industrial‐scale production of biotherapeutic products.

Research paper thumbnail of Comparison of multi‐column chromatography configurations through model‐based optimization

Biotechnology Progress, Jul 16, 2023

Research paper thumbnail of Symbolic Regression for PDEs using Pruned Differentiable Programs

arXiv (Cornell University), Mar 13, 2023

Physics-informed Neural Networks (PINNs) have been widely used to obtain accurate neural surrogat... more Physics-informed Neural Networks (PINNs) have been widely used to obtain accurate neural surrogates for a system of Partial Differential Equations (PDE). One of the major limitations of PINNs is that the neural solutions are challenging to interpret, and are often treated as black-box solvers. While Symbolic Regression (SR) has been studied extensively, very few works exist which generate analytical expressions to directly perform SR for a system of PDEs. In this work, we introduce an end-to-end framework for obtaining mathematical expressions for solutions of PDEs. We use a trained PINN to generate a dataset, upon which we perform SR. We use a Differentiable Program Architecture (DPA) defined using context-free grammar to describe the space of symbolic expressions. We improve the interpretability by pruning the DPA in a depth-first manner using the magnitude of weights as our heuristic. On average, we observe a 95.3% reduction in parameters of DPA while maintaining accuracy at par with PINNs. Furthermore, on an average, pruning improves the accuracy of DPA by 7.81%. We demonstrate our framework outperforms the existing state-of-the-art SR solvers on systems of complex PDEs like Navier-Stokes: Kovasznay flow and Taylor-Green Vortex flow. Furthermore, we produce analytical expressions for a complex industrial use-case of an Air-Preheater, without suffering from performance loss viz-a-viz PINNs.

Research paper thumbnail of Real-time Health Monitoring of Heat Exchangers using Hypernetworks and PINNs

arXiv (Cornell University), Dec 20, 2022

We demonstrate a Physics-informed Neural Network (PINN) based model for realtime health monitorin... more We demonstrate a Physics-informed Neural Network (PINN) based model for realtime health monitoring of a heat exchanger, that plays a critical role in improving energy efficiency of thermal power plants. A hypernetwork based approach is used to enable the domain-decomposed PINN learn the thermal behavior of the heat exchanger in response to dynamic boundary conditions, eliminating the need to retrain. As a result, we achieve orders of magnitude reduction in inference time in comparison to existing PINNs, while maintaining the accuracy on par with the physics-based simulations. This makes the approach very attractive for predictive maintenance of the heat exchanger in digital twin environments.

Research paper thumbnail of Application of Reinforcement Learning for Real-Time Optimal Control of the Pellet Induration Process

Transactions of the Indian Institute of Metals

Research paper thumbnail of Transfer Entropy-Based Automated Fault Traversal and Root Cause Identification in Complex Nonlinear Industrial Processes

Industrial & Engineering Chemistry Research

Research paper thumbnail of Modeling and Simulation of Ultrafine Grinding of Alumina in a Planetary Ball Mill

Transactions of the Indian Institute of Metals

Research paper thumbnail of Artificial Intelligence for Monitoring and Optimization of an Integrated Mineral Processing Plant

Transactions of the Indian Institute of Metals

Research paper thumbnail of System for Optimizing and Controlling Particle Size Distribution and Production of Nanoparticles in Furnace Reactor

Research paper thumbnail of On-line optimization of induration of wet iron ore pellets on a moving grate

Research paper thumbnail of Selective flocculation of fines

A number of factors that affect selective fkJCcu1ation of fines have been identified and the effe... more A number of factors that affect selective fkJCcu1ation of fines have been identified and the effect of ~e of pa-rameters on p~ behavior has been explained with the help of a few case studies. Experiments with francolite-montmo-rillonite and francolite-palygorskite mixtures indicated that francolite recovery depends on pH and the type of dispersant. The results showed that removal of multivalent ionic species on clay mineral surfaces seems to enhance flocculation. and separation efficiency increases as Caz+ ions are removed from the surface. When chalcopyrite and quartz are present to-gether. it is however necessary to clean the fIocs obtained to remove entrapped quartz.

Research paper thumbnail of HyperLoRA for PDEs

arXiv (Cornell University), Aug 17, 2023

Physics-informed neural networks (PINNs) have been widely used to develop neural surrogates for s... more Physics-informed neural networks (PINNs) have been widely used to develop neural surrogates for solutions of Partial Differential Equations. A drawback of PINNs is that they have to be retrained with every change in initial-boundary conditions and PDE coefficients. The Hypernetwork, a model-based meta learning technique, takes in a parameterized task embedding as input and predicts the weights of PINN as output. Predicting weights of a neural network however, is a highdimensional regression problem, and hypernetworks perform sub-optimally while predicting parameters for large base networks. To circumvent this issue, we use a low ranked adaptation (LoRA) formulation to decompose every layer of the base network into low-ranked tensors and use hypernetworks to predict the low-ranked tensors. Despite the reduced dimensionality of the resulting weight-regression problem, LoRAbased Hypernetworks violate the underlying physics of the given task. We demonstrate that the generalization capabilities of LoRA-based hypernetworks drastically improve when trained with an additional physics-informed loss component (HyperPINN) to satisfy the governing differential equations. We observe that LoRA-based HyperPINN training allows us to learn fast solutions for parameterized PDEs like Burger's equation and Navier Stokes: Kovasznay flow, while having an 8x reduction in prediction parameters on average without compromising on accuracy when compared to all other baselines.

Research paper thumbnail of Reduced-order modeling of conjugate heat transfer in lithium-ion batteries

International Journal of Heat and Mass Transfer/International journal of heat and mass transfer, Aug 1, 2024

Research paper thumbnail of Virtual metrology for chemical mechanical planarization of semiconductor wafers

Journal of intelligent manufacturing, Mar 13, 2024

Research paper thumbnail of metafur: Digital Twin System of a Blast Furnace

Transactions of the Indian Institute of Metals, Jun 28, 2024

Research paper thumbnail of Machine Learning Based Modeling of Silicon Content of Molten Iron from a Blast Furnace

Research paper thumbnail of Machine learning based knowledge discovery and modeling of silicon content of molten iron from a blast furnace

Research paper thumbnail of Multi-Modal Instruction-Tuning Small-Scale Language-and-Vision Assistant for Semiconductor Electron Micrograph Analysis

Proceedings of the AAAI symposium series, May 20, 2024

We present a novel framework for analyzing and interpreting electron microscopy images in semicon... more We present a novel framework for analyzing and interpreting electron microscopy images in semiconductor manufacturing using vision-language instruction tuning. The framework employs a unique teacher-student approach, leveraging pretrained multimodal large language models such as GPT-4 to generate instruction-following data for zero-shot visual question answering (VQA) and classification tasks, customizing smaller multimodal models (SMMs) for microscopy image analysis, resulting in an instruction-tuned language-and-vision assistant. Our framework merges knowledge engineering with machine learning to integrate domain-specific expertise from larger to smaller multimodal models within this specialized field, greatly reducing the need for extensive human labeling. Our study presents a secure, cost-effective, and customizable approach for analyzing microscopy images, addressing the challenges of adopting proprietary models in semiconductor manufacturing.

Research paper thumbnail of Droplet impingement on a solid surface: Parametrization and asymmetry of dynamic contact angle model

Physics of Fluids, Jun 1, 2023

The study of the spreading behavior of droplets impinging on solid surfaces is of importance to a... more The study of the spreading behavior of droplets impinging on solid surfaces is of importance to applications such as inkjet printing and spray coating. The contact angle is an important parameter that influences the spreading behavior of droplets upon impingement on a solid surface. Computational fluid dynamics simulations studying droplet dynamics require a dynamic contact angle (DCA) model with an appropriate set of parameters to simulate the experimental system of interest. We propose a scheme to parameterize a DCA model and tune its parameters for systems of different levels of wettability. The developed DCA models show the varied response for advancing and receding phases of the droplet motion to emphasize the asymmetric nature of the relation between the contact angle and contact line velocity. These models enable accurate simulation of droplet impingement for a wide range of Weber number (We) and Reynolds number (Re) values. The proposed scheme helps tune the parameters of the DCA model in a systematic and quick manner, thereby enabling one to explore the design space better and also reduce the time to design and develop novel fluids and devices for applications dealing with impinging droplets.

Research paper thumbnail of Mathematical Modeling of Coagulation and Flocculation of Colloidal Suspensions Incorporating the Influence of Surface Forces

Research paper thumbnail of Advances in bioreactor control for production of biotherapeutic products

Biotechnology and Bioengineering, Mar 1, 2023

Advanced control strategies are well established in chemical, pharmaceutical, and food processing... more Advanced control strategies are well established in chemical, pharmaceutical, and food processing industries. Over the past decade, the application of these strategies is being explored for control of bioreactors for manufacturing of biotherapeutics. Most of the industrial bioreactor control strategies apply classical control techniques, with the control system designed for the facility at hand. However, with the recent progress in sensors, machinery, and industrial internet of things, and advancements in deeper understanding of the biological processes, coupled with the requirement of flexible production, the need to develop a robust and advanced process control system that can ease process intensification has emerged. This has further fuelled the development of advanced monitoring approaches, modeling techniques, process analytical technologies, and soft sensors. It is seen that proper application of these concepts can significantly improve bioreactor process performance, productivity, and reproducibility. This review is on the recent advancements in bioreactor control and its related aspects along with the associated challenges. This study also offers an insight into the future prospects for development of control strategies that can be designed for industrial‐scale production of biotherapeutic products.

Research paper thumbnail of Comparison of multi‐column chromatography configurations through model‐based optimization

Biotechnology Progress, Jul 16, 2023

Research paper thumbnail of Symbolic Regression for PDEs using Pruned Differentiable Programs

arXiv (Cornell University), Mar 13, 2023

Physics-informed Neural Networks (PINNs) have been widely used to obtain accurate neural surrogat... more Physics-informed Neural Networks (PINNs) have been widely used to obtain accurate neural surrogates for a system of Partial Differential Equations (PDE). One of the major limitations of PINNs is that the neural solutions are challenging to interpret, and are often treated as black-box solvers. While Symbolic Regression (SR) has been studied extensively, very few works exist which generate analytical expressions to directly perform SR for a system of PDEs. In this work, we introduce an end-to-end framework for obtaining mathematical expressions for solutions of PDEs. We use a trained PINN to generate a dataset, upon which we perform SR. We use a Differentiable Program Architecture (DPA) defined using context-free grammar to describe the space of symbolic expressions. We improve the interpretability by pruning the DPA in a depth-first manner using the magnitude of weights as our heuristic. On average, we observe a 95.3% reduction in parameters of DPA while maintaining accuracy at par with PINNs. Furthermore, on an average, pruning improves the accuracy of DPA by 7.81%. We demonstrate our framework outperforms the existing state-of-the-art SR solvers on systems of complex PDEs like Navier-Stokes: Kovasznay flow and Taylor-Green Vortex flow. Furthermore, we produce analytical expressions for a complex industrial use-case of an Air-Preheater, without suffering from performance loss viz-a-viz PINNs.

Research paper thumbnail of Real-time Health Monitoring of Heat Exchangers using Hypernetworks and PINNs

arXiv (Cornell University), Dec 20, 2022

We demonstrate a Physics-informed Neural Network (PINN) based model for realtime health monitorin... more We demonstrate a Physics-informed Neural Network (PINN) based model for realtime health monitoring of a heat exchanger, that plays a critical role in improving energy efficiency of thermal power plants. A hypernetwork based approach is used to enable the domain-decomposed PINN learn the thermal behavior of the heat exchanger in response to dynamic boundary conditions, eliminating the need to retrain. As a result, we achieve orders of magnitude reduction in inference time in comparison to existing PINNs, while maintaining the accuracy on par with the physics-based simulations. This makes the approach very attractive for predictive maintenance of the heat exchanger in digital twin environments.

Research paper thumbnail of Application of Reinforcement Learning for Real-Time Optimal Control of the Pellet Induration Process

Transactions of the Indian Institute of Metals

Research paper thumbnail of Transfer Entropy-Based Automated Fault Traversal and Root Cause Identification in Complex Nonlinear Industrial Processes

Industrial & Engineering Chemistry Research

Research paper thumbnail of Modeling and Simulation of Ultrafine Grinding of Alumina in a Planetary Ball Mill

Transactions of the Indian Institute of Metals

Research paper thumbnail of Artificial Intelligence for Monitoring and Optimization of an Integrated Mineral Processing Plant

Transactions of the Indian Institute of Metals

Research paper thumbnail of System for Optimizing and Controlling Particle Size Distribution and Production of Nanoparticles in Furnace Reactor

Research paper thumbnail of On-line optimization of induration of wet iron ore pellets on a moving grate

Research paper thumbnail of Selective flocculation of fines

A number of factors that affect selective fkJCcu1ation of fines have been identified and the effe... more A number of factors that affect selective fkJCcu1ation of fines have been identified and the effect of ~e of pa-rameters on p~ behavior has been explained with the help of a few case studies. Experiments with francolite-montmo-rillonite and francolite-palygorskite mixtures indicated that francolite recovery depends on pH and the type of dispersant. The results showed that removal of multivalent ionic species on clay mineral surfaces seems to enhance flocculation. and separation efficiency increases as Caz+ ions are removed from the surface. When chalcopyrite and quartz are present to-gether. it is however necessary to clean the fIocs obtained to remove entrapped quartz.