Article Preview
TopIntroduction
Magnetic refrigeration has been used for a long time in different technological applications by using solid material and magnetic field. (Smith, 2013; Brown; 1976Warburg; 1881). In order to deal with the drawbacks of the global warming and the negative impact of synthetic refrigerant chlorofluoro carbon (CFC), hydro chlorofluoro carbon (HCFC) and hydro fluoro carbon (HFC) on the environment (Balli, 2017; Gómez, 2013; Yu, 2010), several alternative cooling technologies were proposed (Balli, 2017; Chiba, 2017). In this context, the magnetic refrigeration based on the magnetocaloric effectis currently considered as one of the best alternative for conventional systems (Kitanovski, 2015). The magnetocaloric effect can be defined as the thermal response of a magnetic material wen subjected to an external magnetic field which is due interplay between the magnetic moments and the phonons lattice (Bouchekara, 2014). Aiming to predict the performance of active magnetic refrigerator thermodynamic cycles, a one and multi-dimensional numerical models was proposed (Roudaut, 2011). On the other hand, new designs and innovative prototypes working with active magnetic refrigerator cycles were recently reported in the literature (Kamran, 2015; Chiba, 2015).
Recently, intelligence artificial method has been used for prediction and optimisation for several industrial process behavior dedicated for engineering as well fuzzy logic, neural networks and multiobjective optimization (Russell, 2010; Ross, 2004; Bouchekara, 2014).
Neural networks have broad applicability to scientific problems. In fact, they have already been applied successfully in many domains. Since a neural network has been considered a good method for identifying models or trends in data, they are well suited for simulation or prediction needs including (Haykin, 1999; Christopher, 1995); performance predictions, industrial process control, optimization systems and data validation.
The aim of this paper is to develop two models with good performances based on Artificial Neural Networks and to predict coefficient of performance and temperature span values of active magnetic refrigerator cycle based on gadolinium materiel.