Arjun Gopichander Ravichander

Physics-Constrained Synthetic Data Generation For Industrial Machine Learning: Enforcing Conservation And Operating-Envelope Constraints, Validating Fidelity Against Sensor Records, And Measures

A method for generating industrial sensor data under explicit physical constraints is proposed, and how it should be evaluated is specified, including a fidelity validation design going beyond marginal-distribution testing to cover cross-sensor coupling, spectral content, direct measurement of constraint violation rate...