|
Based on the design of our magnetic refrigeration test machine we have developed a numerical model of Active Magnetic Regenerator (AMR) based on parallel plates. An illustration of the design of the AMR is shown below. Further below is shown an actual photograph of a closeup of the test machines AMR geometry. Currently the numerical model is able to replicate the central and important aspects of the physics going on in an AMR to some extent. The points in modeling the parallel-plate AMR are primarily to investigate if our understanding of the physics behind regeneration in magnetic cooling is satisfactory and most importantly to be able to predict the performance of certain buildable geometries and operation conditions of the current as well as future designs. If we can model the AMR sufficiently and thus be able to predict how to build the optimal magnetic refrigeration device, we can save a lot of money and time by not physically building many different concepts, perhaps even with nothing or little in common.
It is important to evaluate whether the model actually does predict the behavior of the experimental setup without mingling too much with “unknown” parameters such as fictitiously including losses and so on. This is a very general concern when doing numerical modeling that one should always be aware of; how good does the model predict the real world and how much has one “tuned” it to reproduce the experimental values. Currently, our model is somewhat idealized in the sense that it is 2-dimensional and makes extensive use of assumed symmetries.
The results we have so far, which are comparable to the experimental setup as far as possible with the current model design, are generally quite good. They show to a large extend the same tendencies but do not reproduce the same absolute values. The key parameter to directly compare between experiment and model is the no-load temperature difference. This is the temperature difference between the hot and cold end of the system in cyclic steady-state when no cooling is done. Here we see the same tendencies for various situations but the temperature span is generally over-estimated by the model.
One of the next steps we aim at is to make a 3-dimensional model with realistic losses, i.e. simply to model the experimental geometry in much higher detail. This is doable within a short time frame since the physics going on are the same and in general the equations to solve are the same. The only big difference is the significant computation time needed to solve a 3-dimensional model compared to a 2-dimensional one.
It should also be noted that it is very nice to know that the model in its current state reproduces quite the same tendencies as the experiment. This means that if the model predicts an enhanced performance for a certain setup within the limits of the experiment, then the experiment should show an enhancement in performance as well. This can be used as a pointer towards the direction the design of an efficient AMR based on parallel plates should follow. |