Design of drain tube with cone to prevent frost accumulation of drainage hole in household refrigerators using artificial neural network
The drain tube in household refrigerators serves to discharge defrost water and equalize internal and external pressure. However, its open-ended design permits humid air ingress, accelerating frost buildup and reducing system efficiency. To address this, a simplified cone-shaped drain tube was propo...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-10-01
|
| Series: | Case Studies in Thermal Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2214157X25010615 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The drain tube in household refrigerators serves to discharge defrost water and equalize internal and external pressure. However, its open-ended design permits humid air ingress, accelerating frost buildup and reducing system efficiency. To address this, a simplified cone-shaped drain tube was proposed with a focus on manufacturability. Experiments and computational fluid dynamics (CFD) simulations were conducted to investigate frost formation patterns and the effects of absolute humidity. To reduce the cost of evaluating design variations, an artificial neural network (ANN) was developed as a surrogate model trained on CFD data. The ANN accurately predicted performance based on geometric parameters, capturing nonlinear interactions among pressure release, airflow, and frost accumulation. The optimal cone-type design, predicted with a 5.12 % error margin, was fabricated via 3D printing and validated experimentally through pressure and flow rate measurements. When applied in a commercial refrigerator, the cone-type tube delayed frost blockage by more than twice and reduced negative pressure peaks caused by door opening by approximately 85 % compared to conventional designs. These results highlight the effectiveness of the AI-assisted framework for data-efficient thermal optimization and demonstrate the potential of intelligent, learning-based approaches in thermal system design. |
|---|---|
| ISSN: | 2214-157X |