Projection Based and Data-Driven Reduced Order Models for Annealing Furnace
Please login to view abstract download link
The present work demonstrates the application of a projection-based and data-driven reduced order model for the annealing furnace with the variation of the gas composition at the inlet of the furnace, annealing temperature and annealing duration. Annealing is a heat treatment process that involves heating steel to a specific temperature and holding it at that temperature for a predetermined time, followed by controlled cooling. The current application involves a multi-species problem where the gas composition in the inlet contains water vapour with air flow with a dewpoint of -30 to 10^0 C and volume percentage for hydrogen level of 10-30 %. With a certain annealing surface temperature and annealing duration, the steel grades are adjusted inside the annealing furnace. We will focus on finding out the optimum gas inlet compositions, annealing temperature and duration for a desired decarburization depth. We decouple the present multi-physics (considering the lower concentration of water vapour and Carbon Monoxide) problem in two stages – first obtain the solution of the flow field considering a fixed viscosity fluid assumption, buoyancy and turbulence effect. In this stage, We have used intrusive methods such as stabilized Proper Orthogonal Decomposition (POD)-Galerkin Projection based reduced order model [1] and a non-intrusive Reduced Order Model based on POD-LSTM [2] for obtaining the flow physics which is furthermore coupled with the reaction model and species transport equation to obtain the decarburization depth in the second stage. We will consider the temporal and parametric variation and different stabilization techniques in the reduced-order models.