Share

IIR document

Investigation of refrigerant leakage behavior prediction using machine learning.

Number: 1207

Author(s) : YOKONO R., OHAMA H., KAMADA M., HORI K.

Summary

This study explores the construction of a refrigerant leakage behavior prediction model with a sufficient level of accuracy for screening high-risk conditions and a light computational load using machine learning. Conventional prediction methods require labor-intensive CAD generation and computationally intensive CFD analysis performed by skilled operators. As a first step, this study attempted to reduce the training load by updating a DNN model trained on a large amount of CFD results for R32 with transfer learning using a small amount of new refrigerant CFD results. A transfer learning model with the input layer updated for R1234yf was able to capture the trends of the CFD results, while a similar model for R290 did not yield satisfactory results, which is estimated to be due to significantly different physical properties related to the output compared to those of the base model.

Available documents

Format PDF

Pages: 8 p.

Available

  • Public price

    20 €

  • Member price*

    Free

* Best rate depending on membership category (see the detailed benefits of individual and corporate memberships).

Details

  • Original title: Investigation of refrigerant leakage behavior prediction using machine learning.
  • Record ID : 30032720
  • Languages: English
  • Source: 16th IIR-Gustav Lorentzen Conference on Natural Refrigerants (GL2024). Proceedings. University of Maryland, College Park, Maryland, USA, August 11-14 2024
  • Publication date: 2024/08
  • DOI: http://dx.doi.org/10.18462/iir.gl2024.1207

Links


See other articles from the proceedings (135)
See the conference proceedings

Indexing