Advanced big-data/machine-learning techniques for optimization and performance enhancement of the heat pipe technology – A review and prospective study. (15th July 2021)

• Record Type: Journal Article
• Authors: Wang, Zhangyuan; Zhao, Xudong; Han, Zhonghe; Luo, Liang; Xiang, Jinwei; Zheng, Senglin; Liu, Guangming; Yu, Min; Cui, Yu; Shittu, Samson; Hu, Menglong
• Is Part Of: Applied energy. Volume 294(2021)
• Publication Date: 2021-07-15
• Subjects: Heat pipe -- Big data -- Machine learning -- Optimization -- Prediction -- Algorithm; ANN Artificial Neural Networks -- ANFIS Artificial Neuro Fuzzy Inference System -- CRAC Computer Room Air-conditioning -- CLPHP Closed Loop Pulsating Heat Pipe -- DI Deionized -- FPMHP Flat-plate Micro Heat Pipe -- GA Genetic Algorithm -- HP Heat Pipe -- HPSCs Heat Pipe Solar Collectors -- LR Linear Regression -- NoSQL Not Only SQL -- OLPHP Open Loop Pulsating Heat Pipe -- RSM Response Surface Methodology -- SVM Support Vector Machine -- SVR Support Vector Regression -- TRN Thermal Resistance Network; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2021.116969
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: Critical review of the heat pipe (HP) technologies was undertaken. Existing HP simulation models are extremely time-consuming and impractical. A big-data-trained HP machine learning algorithm is a solution. Challenges for the big-data-trained HP machine learning technology was investigated. The future research directions of the HP machine learning were outlined. Abstract: A heat pipe (HP) is a passive heat transfer device able to transmit heat a few meters or several hundred meters away from the heat source without use of external energy. This paper presents a critical review of the HP technologies. It is found that the heat transfer performance of a HP is highly dependent upon its geometrical and operational conditions, whilst the existing computerized analytical and numerical models for the HP require a huge number of parametrical data inputs, and therefore is extremely time-consuming and impractical. Furthermore, the measurement results of the HPs vary time by time and show certain disagreement with the simulation prediction, giving a high uncertainty in characterisation of the HP. Development of a machine learning algorithm and associated models based on the structured HP database is a solution to tackle these challenges, which is able to provide the dimensionless and multiple-factors-considering solution for HP structural optimization and performance prediction. A review on big-date/machine-learning technology for HP application was undertaken, indicating that a database covering the HP parametrical data, operational variables and associated performance results has not yet been established. Challenges for the HP structural optimization and performance prediction using the big-data-trained machine learning technology lie in: (1) complex and unregulated HP data; (2) unidentified analytic algorithm for HP structural optimization; and (3) unidentified data-driven algorithm for HP performance prediction. This review-based study provides the potential future research directions for development of the big-data-trained machine learning technology for HP structural optimization and performance prediction.