AI-Driven Monte Carlo Uncertainty Analysis of Curie Temperature Effects on Active Magnetic Regenerator Performance

In the paper AI-Driven Monte Carlo Uncertainty Analysis of Curie Temperature Effects on Active Magnetic Regenerator Performance, published in the International Journal of Refrigeration (IF = 3.8), researchers from the Laboratory for Refrigeration and District Energy revisited the methodology for analyzing the impact of statistical variations in Curie temperature on the performance of multilayer magnetocaloric regenerators, as previously established in a recent publication [link].

The study was conducted in collaboration with colleagues from the Federal University of Santa Catarina (Brazil), this time using their numerical model of a multilayer active magnetic regenerator (AMR) combined with machine learning. Instead of second-order materials, the study focused on first-order LaFeSiH magnetocaloric materials, which are even more sensitive to Curie temperature distributions.

Once again, the study revealed that Curie temperature deviations above 1 K drastically reduce the probability of achieving the target cooling power. Given current margins provided by MCM suppliers (standard deviations between 1.5 and 2 K) and a typical AMR layer count (10 to 15), the cooling power would need to be oversized by 30% to 80% to ensure a 90–95% probability of meeting performance targets. In practice, this would require oversized magnets and regenerators, significantly increasing production costs.

Thus, large-scale production of magnetic refrigeration devices with current LaFeSiH materials is unfeasible, as quality assurance standards would require much tighter control over Curie temperature uncertainty than currently offered by manufacturers. For a successful market deployment of magnetocaloric technology, MCM manufacturers will need to prioritize reducing Curie temperature variability in their materials.

Figure: a) Schematic representation of a multilayer AMR and the Curie temperature deviation. Solid lines represent the original curves and dashed lines represent the ones achieved by the manufacturing process, b) Effect of the uncertainty of the Curie Temperature on the performance of the 10 layer AMR, c) Cumulative Distribution Function for the 10 layer AMR, d) Effect of the certainty level on the achieved performance of the 10 layer AMR.