Machine learning approach for predicting refrigerant two-phase pressure drop inside Brazed Plate Heat Exchangers (BPHE)

https://doi.org/10.1016/j.ijheatmasstransfer.2020.120450Get rights and content

Highlights

Abstract

This paper presents a Gradient Boosting Machines (GBM) model for predicting refrigerant two-phase frictional pressure gradient inside Brazed Plate Heat Exchangers (BPHE) based on an extensive database that includes 1624 boiling data-points, 925 condensation data-points, 16 different plate geometries, and 16 different refrigerants (including 4 natural refrigerants and 6 other low-GWP refrigerants). The model accounts for the effect of plate geometry, operating conditions and refrigerant properties. The model is able to reproduce the whole database with a Mean Absolute Percentage Error (MAPE) of 6.6%. The GBM model exhibits a better predictive performance than the state-of-the-art analytical-computational procedures for two-phase pressure drop inside BPHE available in the open literature. The characteristic parameters of the GBM model are thoroughly reported in the paper.

Introduction

The experimental and theoretical investigations on two-phase flow inside Brazed Plate Heat Exchangers (BPHE) available in the open literature are mainly focused on the measurements and the prediction of convective heat transfer coefficients, whereas only a relatively limited number of works refer to pressure drops. On the contrary, the availability of extensive experimental pressure drop databases and accurate pressure drop computational procedures is fundamental for the design of BPHE. This is even more true for refrigeration and air-conditioning application that involves the use of new low-GWP refrigerants which thermodynamic and thermophysical properties, in some cases, are not yet well established. In fact, the design of any heat transfer equipment must take into account two different concurrent issues: the increase of the heat transfer coefficients, which involves a reduction of size and direct costs of the heat exchanger for a set thermal effectiveness, and the reduction of pressure drops, which increases the global efficiency of the system reducing also pumping indirect costs. The final optimum solution derives from the balance between these two contrasting aspects.

Amalfi et al. [1,2] reviewed pressure drop correlations for refrigerant boiling in BPHE, while Tao and Infante Ferreira [3] surveyed pressure drop correlations for refrigerant condensation in BPHE. Both found that the existing computational procedures for two-phase flow pressure drop in BPHE exhibit poor performances, therefore they proposed new correlations based on huge databases and a dimensional analysis coupled with a multi-variable regression approach.

The Amalfi et al. [1,2] correlation is based on 4 non-dimentional variables:fTP=(ΔPfdeqρm2LG2)=15.698(0.955+2.125β*9.993)×We0.475Bd0.255ρ*0.571β*=ββmaxWe=G2deqρmσBd=(ρLρG)gdeq2σρ*=ρLρG

The Amalfi et al. [1,2] equation shows a MAPE of 21.5% with 90.9% of the predicted data within ± 50% against their boiling database consisting of 1513 pressure gradient data-points.

The Tao and Infante Ferreira [3] equation consists of 3 non-dimentional variables and a dimensional parameter:fTP=(ΔPfdeqρm2LG2)=(4.207+2.673β0.46)×(42005.41Bd1.2)Reeq0.95P*0.3Reeq=(GdeqμL)[(1Xm)+Xm(ρLρG)1/2]P*=PPcrwhere the inclination angle of the corrugation β must be expressed in radian.

The Tao and Infante Ferreira [3] correlation exhibits a MAPE of 31.2% with 87.5% of the predicted data within ± 50% against their condensation database consisting of 1590 pressure gradient data-points.

In the authors' opinion, it is difficult to obtain a substantially better predicting performance by using an analytical approach due to the intrinsic complexity of two-phase flow inside BPHE, which requires deep analysis of physical mechanisms, wide-ranging databases, and innovative non-linear regression techniques such as, for example, the Machine Learning techniques.

The aim of this paper is to present a wide database on refrigerant two-phase diabatic (boiling and condensation) flow pressure gradient inside BPHE and to develop a high performing machine learning model for predicting refrigerant two-phase pressure gradient inside BPHE.

Section snippets

Two-phase flow database

The authors of the present paper carried out in the past an extensive measurement campaign on refrigerant boiling [4], [5], [6], [7], [8], [9], [10] and condensation [11], [12], [13], [14], [15], [16], [17], [18] inside a commercial BPHE. Several traditional HFC refrigerants (R134a, R410A, R236fa, R404A), HydroCarbon (HC) refrigerants (R600a, R290, R1270), low-GWP HFC refrigerants (R152a, R32), HydroFluoroOlefin (HFO) refrigerants (R1234yf, R1234ze(E), R1234ze(Z)) and a HydroChloroFluoroOlefin

GBM model

This paragraph presents a Gradient Boosting Machines (GBM) model for predicting refrigerant two-phase diabatic (boiling and condensation) flow frictional pressure gradient inside BPHE based on the huge database illustrated in the previous paragraph. The variables used as input for the GBM model were corrugation enlargement ratio Φ, reduced inclination angle β/βmax, liquid Prandtl number PrL, specific kinetic energy number KEV, reduced pressure P/Pcr, and type of two-phase heat transfer process

Conclusion

This paper presents and analyses a database on refrigerant two-phase diabatic (boiling and condensation) flow inside BPHE consisting of 2549 data-points relative to 16 plate geometries and 16 refrigerants, including 4 natural refrigerants and 6 low-GWP refrigerants. A Gradient Boosting Machines (GBM) model for predicting refrigerant two-phase diabatic flow inside BPHE is presented. The model accounts for the effect of plate geometry, operating conditions and refrigerant properties. The model is

Declaration of Competing Interest

The authors wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome.

Acknowledgement

The research project was carried out within the International Network "Low GWP Refrigerants" which involves the Khyshu University, the Saga University, the Sangyo University, the Iwaki Meisei University, The Catholic University of America, the University of Padova, the US National Institute of Standard and Technology (NIST), the Construction Technology Institute of the National Research Council of Italy (CNR-ITC), and the Italian National Metrology Institute (INRIM) while the first author

References (46)

  • G.A. Longo

    Heat transfer and pressure drop during hydrocarbon refrigerant condensation inside a brazed plate heat exchanger

    Int. J. Refrig.

    (2010)
  • G.A. Longo et al.

    Condensation of the low GWP refrigerant HFC1234yf inside a brazed plate heat exchanger

    Int. J. Refrig.

    (2013)
  • G.A. Longo et al.

    Condensation of the low GWP refrigerant HFC1234ze(E) inside a brazed plate heat exchanger

    Int. J. Refrig.

    (2014)
  • G.A. Longo et al.

    Experimental assessment of the low GWP refrigerant HFO-1234ze(Z) for high temperature heat pumps

    Exp. Therm. Fluid Sci.

    (2014)
  • G.A. Longo et al.

    Condensation of the low GWP refrigerant HFC152a inside a brazed plate heat exchanger

    Exp. Therm. Fluid Sci.

    (2015)
  • G.A. Longo et al.

    HFC32, a low GWP substitute for HFC410A in medium size chillers and heat pumps

    Int. J. Refrig.

    (2015)
  • G.A. Longo et al.

    HFC404A condensation inside a small brazed plate heat exchanger: comparison with the low GWP substitutes propane and propylene

    Int. J. Refrig.

    (2017)
  • Y.Y. Hsieh et al.

    Saturated flow boiling heat transfer and pressure drop of refrigerant R-410A in a vertical plate heat exchanger

    Int. J. Heat Mass Transf.

    (2002)
  • Y.Y. Yan et al.

    Condensation heat transfer and pressure drop of refrigerant R134a in a plate heat exchanger

    Int. J. Heat Mass Transf.

    (1999)
  • W.S. Kuo et al.

    Condensation heat transfer and pressure drop of refrigerant R410A flow in a vertical plate heat exchanger

    Int. J. Heat Mass Transf.

    (2005)
  • D. Han et al.

    Experiments on the characteristics of evaporation of R410A in brazed plate heat exchangers with different geometric configurations

    Appl. Therm. Eng.

    (2003)
  • J. Huang et al.

    Heat transfer and pressure drop in plate heat exchanger refrigerant evaporators

    Int. J. Refrig.

    (2012)
  • J. Zhang et al.

    Flow boiling heat transfer and pressure drop characteristics of R134a, R1234yf and R1234ze in a plate heat exchanger for organic Rankine cycle units

    Int. J. Heat Mass Transf.

    (2017)
  • Cited by (22)

    • A universal correlation for predicting two-phase frictional pressure drop in horizontal tubes based on machine learning

      2023, International Journal of Multiphase Flow
      Citation Excerpt :

      The model could provide fair agreement with experimental results when the sizes of channels and the flow rate of each phase are used as inputs. Longo et al. (2020) presented that the gradient boosting model could be successfully employed for predicting two-phase FPDs of refrigerant diabatic (boiling and condensing) flowing inside plate heat exchangers. Moreover, the prediction accuracy for all 2549 data points outperformed the computational correlations available in the open literature.

    • Evaluation of machine learning approaches for estimating thermodynamic properties of new generation refrigerant R513A

      2023, Sustainable Energy Technologies and Assessments
      Citation Excerpt :

      It has been emphasized that the hybridization of neural networks with other expert systems outperforms the traditional Artificial Neural Network approach. Longo et al. [17] used machine learning to predict the pressure drop of the refrigerant inside brazed plate heat exchangers. In the study, a large database including 1624 boiling data points, 925 condensation data points, 16 different plate geometries, and 16 different refrigerants of two-phase diabatic flow pressure gradient in the plate heat exchanger was created and a high-performance Gradient Boosting Machines model was developed to predict the refrigerant pressure gradient in the heat exchanger.

    View all citing articles on Scopus
    View full text