MUMBAI, India, Oct. 26 -- Intellectual Property India has published a patent application (202641123313 A) filed by Pinni Srinivasa Varma; Jyosthna D; G. Murali; and Koneru Lakshmaiah Education Foundation on October 14, 2026, for Physics-Guided Machine Learning Framework For Predicting Effective Thermal Conductivity In Synthetic Porous Media.
Inventors include Jyosthna D; G. Murali; and Chakali Bhaskar.
The application for the patent was published on October 23, 2026, under issue no. 43/2026.
Abstract: A simulation-driven, physics-guided machine-learning framework is disclosed for predicting effective thermal conductivity of two-phase porous media. Synthetic binary microstructures are generated using circular-pore and Gaussian-random-field morphologies, assigned solid and pore conductivities, and solved using a conservative finite-difference heat-conduction model with harmonic-mean interface conductivities. Effective conductivity obtained from the simulations is used to train feature-based regressors and image-based convolutional neural networks. Morphology descriptors include porosity, pore-size, connectivity, interface, and material- property features. A physics-guided convolutional model uses a composite loss comprising mean-squared error, a penalty for predictions outside constituent- conductivity bounds, and a penalty for predictions above an arithmetic-mixture trend. The framework provides rapid surrogate prediction while quantifying predictive accuracy, physical consistency, solver convergence, classical-model comparison, and out-of-distribution performance for porous-material thermal characterization and design screening.
Disclaimer: Curated by HT Syndication.
