Abstract
Magnetic refrigeration technology (MRT) recently receives significant attention as a result of its environmental friendliness coupled with its superior cooling efficiency in comparison to the ubiquitous compression gas refrigeration system (CGRS). Discovery of manganite-based refrigerants results in lower cost of MRT and increases interest in the implementation of this technology at room temperature. However, the challenge of altering the magnetic ordering temperature (TC) of manganite based materials for ensuring MRT that operates at room temperature through doping of the parent manganite still remains one of the major setbacks that impede rapid progress in the implementation of this technology due to its extensive experimental procedures and routines. This present work aims at developing a robust model based on hybridization of neural network trained using sensitivity based linear learning method (SBLLM) and gravitational search algorithm (GSA) for determining the suitable dopant and the concentration that shifts the TC of manganite-based materials to ambient value. The developed SBLLM-GSA model is robust due to its capacity to incorporate up to four different dopants of different concentrations into parent manganite for TC estimation with excellent degree of accuracy. The estimated magnetic ordering temperatures using SBLLM-GSA model were validated with the experimentally measured values and excellent agreement was obtained. Technological implementation of the developed SBLLM-GSA model would definitely promote ambient MRT and significantly reduces the use of the harmful, ozone-depleting CGRS.