Prediction of 2G HTS Tape Quench Behavior by Random Forest Model Trained on 2-D FEM Simulations
Dmitry Sotnikov, M. Lyly, Tiina Salmi
Abstract
Detailed Finite Element Method (FEM) based simulations for 2G HTS tapes return high quality results, but the computation takes a long time due to the non-linearity of superconducting properties and they needed high mesh density. This work describes a method for prediction of quench behavior in a long 2G HTS tape based on a series of 2D FEM model simulations for short length of tape in many different conditions. The random forest model is trained by the set of results from the short-pieces FEM calculations. Subsequently the model can be applied to any length of HTS tape with similar thermal characteristics. Comparison of quench simulation in 10 cm long HTS tape between a detailed FEM model and a fully trained random forest model show that the predicted temperatures are within 0.68%, while the computation time is significantly faster: The random forest model ran in less than 1 s, while the run time of the FEM model was 5:30 min.