Simulation of heat flow in welding using hybrid restarting FDM-SVR-Lanczos

This paper discusses the numerical simulations of heat transfer problem, particularly during welding materials processing. process highly depends on temperature to determine metallurgical properties, strength, and sureness joint welding; hence it is desirable study its flow. Finite difference method (FDM) used discretize partial differential equations (PDEs) obtain systems linear (SLEs), which can be solved by an iterative method. Here, we a variant Lanczos based methods, called Restarting-FDM-SVR-Lanczos solve two-dimensional flow in welding. The proposed combination machine learning under support vector regression (SVR) method, where SVR plays important role predicting non-existing solutions iterates, due breakdown issue Numerical results were presented showed that Hybrid Restarting FDM-SVR-Lanczos performed better solving problem compared with single FDM small residual norms.