Relative cooling power modeling of RE2TM2Y ternary intermetallic rare-earth-based magnetocaloric compounds for magnetic refrigeration application using extreme learning machine and hybrid intelligent method

Abstract

Ternary intermetallic rare-earth-based magnetocaloric compounds (RE₂TM₂Y, where RE = Gd, Tb, Dy, Ho, Er, Tm, TM= Ni, Cu, Co and Y = Sn, In, Cd, Ga, Al) have attracted attention lately as magnetic refrigerants in addressing major concerns of the conventional system of refrigeration. Assessment of the amount of heat transferred between cold and hot reservoirs at varying applied magnetic fields through relative cooling power (RCP) determination is costly and experimentally intensive which calls for predictive computational techniques with characteristic precision. In this contribution, intelligent-based predictive models are developed through sine activation function-based extreme learning machine (SELM) and genetically optimized support vector regression (GSVR) with Gaussian (GU) and polynomial (PY) kernel functions for data mapping and transformation using applied magnetic field and ionic radii of the constituent elements as descriptors. The GU-GSVR model exhibits superior performance compared to both the PY-GSVR and SELM models when validated using a testing set of ternary intermetallic rare-earth-based magnetocaloric compounds with improvement of 10.55% and 2.28%, respectively using correlation coefficient (CC) as assessment parameter. During model validation, the developed GU-GSVR also showcases enhanced performance across additional performance metrics, including root mean square error (RMSE) and mean absolute error (MAE). The impact of externally applied magnetic field on the RCP of different ternary intermetallic rare-earth-based magnetocaloric compounds was examined by utilizing the developed GU-GSVR model. The characteristic precision and accuracy of the developed computational intelligent models would enable adequate as well as comprehensive investigation of ternary intermetallic rare-earth-based magnetocaloric compounds for a clean system of refrigeration.