MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641064330 A) filed by Amrita Vishwa Vidyapeetham on May 21, 2026, for Cyclic Consistency Based Vorticity Field Prediction In Fluid Flows.
Inventors include M. S. , Vishnu Mohan; Mohan Prasanna, Hariprasad; and Menon, Vivek.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: The present disclosure provides a system (100) for predicting vorticity fields in fluid flows and a method for vorticity field prediction using cyclic consistency-based training. The system (100) includes a forward generator (118) having a transformer encoder (118-2) and an embedding network (118-4), a backward generator (120), a discriminator (122), and a loss unit (124) computing a hybrid loss (316) including cyclic consistency loss (306), adversarial loss (308), reconstruction loss (310), structural similarity loss (312), and gradient loss (314). The system (100) enables single-step and multi-step prediction conditioned on temporal offset value (106), kinematic viscosity value (108), and inlet velocity value (110). Unlike conventional autoregressive approaches that exhibit cumulative error propagation, this approach achieves stable long-horizon forecasts through cyclic consistency regularization while enabling transfer learning from higher-offset models to lower-offset models for robust generalization across diverse flow regimes.
Disclaimer: Curated by HT Syndication.
