.png){kind=icon}
.png){kind=icon}
Return
A data-driven framework for energy-efficient design optimization of semiconductor cleanroom HVAC systems
.png){kind=icon}
OA
AI
DOI：10.1016/j.csite.2026.108351.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Semiconductor cleanroom HVAC systems consume substantial energy due to stringent temperature and humidity requirements, yet previous studies have largely focused on predefined configurations or individual components, leaving the coupled effects of component arrangement and operating variables underexplored. This study proposes a data-driven optimization framework integrating computational fluid dynamics (CFD), surrogate modeling, and a genetic algorithm to identify feasible HVAC designs satisfying target outlet air conditions. Using Latin Hypercube Sampling, over 10,000 CFD cases were generated for each of the summer and winter conditions, and Random Forest, LightGBM, and Multi-Layer Perceptron (MLP) surrogates were compared. Among them, the MLP achieved the best predictive performance, with RMSE values of 0.11 K and 4.65 × 10−5 kgw/kgair under summer conditions and 0.20 K and 1.06 × 10−4 kgw/kgair under winter conditions. The trained MLP achieved a computational speed-up of approximately 4.4 × 106 for single-case evaluation. The optimization results showed that the target outlet air conditions could be achieved through multiple feasible design solutions rather than a single unique optimum. These findings demonstrate that the proposed framework can support energy-efficient cleanroom HVAC design by clarifying the coupled effects of component arrangement and operating variables and by guiding the selection of practical design alternatives under energy and installation cost trade-offs.
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
6.4
Papers:
8.0K
Citations:
2.6W