Prediction of the parameters affecting the performance of compact heat exchangers with an innovative design using machine learning techniques

Uguz S., İPEK O.

JOURNAL OF INTELLIGENT MANUFACTURING, vol.33, no.5, pp.1393-1417, 2022 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 33 Issue: 5
  • Publication Date: 2022
  • Doi Number: 10.1007/s10845-020-01729-0
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, IBZ Online, ABI/INFORM, Applied Science & Technology Source, Business Source Elite, Business Source Premier, Compendex, Computer & Applied Sciences, INSPEC
  • Page Numbers: pp.1393-1417
  • Keywords: Artificial neural networks, CFD analyses, Compact heat exchanger, Machine learning, Metal additive manufacturing, Multiple linear regression, Support vector machines
  • Süleyman Demirel University Affiliated: Yes


In this study, the innovative compact heat exchanger (CHE) newly designed and manufactured using metal additive manufacturing technology were numerically and experimentally investigated. Some experiments were carried out to determine the hot water (hw) and cold water (cw) outlet temperatures of CHE. As a result of the CFD analysis, the average outlet temperatures of the hw and cw flow loops on the CHE were calculated as 48.24 and 35.38 degrees C, respectively. On the other hand, the experimental outlet temperatures were measured as being 48.50 and 35.72 degrees C, respectively. The studies showed that the numerical and experimental results of the CHE are compliant at the given boundary conditions. Furthermore, it was observed that the heat transfer rate of the CHE with lower volume is approximately 47.7% higher than that of standard brazed plate heat exchangers (BPHEs) produced by traditional methods. More experiments conducted on the CHE will inevitably have a negative effect on its manufacture time and cost. Thus, various models were developed to predict the results of unperformed experiments using the machine learning methods, ANN, MLR and SVM. In the models developed for each experiment, the source and inlet temperatures of hw and cw, respectively, and the volumetric flow rate of cw were selected as input parameters for the machine learning methods. Thus, the hw and cw outlet temperatures of the CHE were estimated on the basis of these input parameters. The best performance was achieved by ANN. In addition, there is no significant performance difference between other methods.