Data mining techniques for thermophysical properties of refrigerants

https://doi.org/10.1016/j.enconman.2008.09.002Get rights and content

Abstract

This study presents ten modeling techniques within data mining process for the prediction of thermophysical properties of refrigerants (R134a, R404a, R407c and R410a). These are linear regression (LR), multi layer perception (MLP), pace regression (PR), simple linear regression (SLR), sequential minimal optimization (SMO), KStar, additive regression (AR), M5 model tree, decision table (DT), M5’Rules models. Relations depending on temperature and pressure were carried out for the determination of thermophysical properties as the specific heat capacity, viscosity, heat conduction coefficient, density of the refrigerants. Obtained model results for every refrigerant were compared and the best model was investigated. Results indicate that use of derived formulations from these techniques will facilitate design and optimize of heat exchangers which is component of especially vapor compression refrigeration system.

Introduction

Refrigerants are one of the most important parts of refrigeration and heat pump systems. For many years, CFCs and HCFCs have been used successfully as refrigerants, blowing agents, cleaning solvents, and aerosol propellants. CFCs seemed to be an ideal choice due to their unique combination of properties; however, the exceptional stability of these compounds, coupled with their chlorine content, has linked them to the depletion of the earth’s atmosphere protective ozone layer. By international agreement (Montreal Protocol), signed in 1987 and later amended several times, this group of refrigerants, consisting primarily of R-11, R-12, R-113, R-114, R-115, R-500 and R-502 were scheduled to be phased out of production by 1 January 1996, in the developed countries and by the year 2000 in the developing countries [1], [2].

The objective of this study is to obtain thermophysical properties depending on temperature and pressure values of different new refrigerants used in the vapor compression refrigeration system. The calculations were made for alternative refrigerants R134a, R404a, R407c and R410a, which do not damage to ozone layer. The thermophysical properties of refrigerants were taken from Solkane Software. All data mining analyses were performed using WEKA software and thermophysical property formulations of refrigerants were derived using linear regression (LR), multi layer perception (MLP), pace regression (PR), simple linear regression (SLR), sequential minimal optimization (SMO), KStar, additive regression (AR), M5 model tree, decision table (DT), M5’Rules techniques within data mining. WEKA software contains a suite of advanced knowledge discovery algorithms that extract knowledge from an investigated database and present this knowledge in symbolic rules that can be interpreted by a user. Development of a model requires the user to input a set of dependent and independent variables, either directly or by importing from an existing file (e.g., an excel spreadsheet). The user must simply specify a dependent variable, the independent variables, a time limit for running the algorithm, and a desired standard error. Use of the package seems quite appropriate for many users.

Data mining is applied in a wide variety of fields for prediction, customer behavior, and production control. In addition, data mining has also been applied to other types of scientific data such as bioinformatical, astronomical, and medical data [3].

In the literature, available studies on analysis with data mining approach of energy systems are very limited. Şencan has used data mining process to determine specific volume values of methanol/LiBr and methanol/LiCl used in absorption heat pump systems. Mathematical formulations were found to be in good agreement with the experimental data [4]. Tso and Yau have used regression analysis, decision tree and neural networks models in the data mining approach for the prediction of electricity energy consumption [5]. Kusiak et al. was applied data mining approach to analyze relationships among parameters of a circulating fluidized-bed boiler. The efficiency could be predicted to the same degree of accuracy with and without the data describing the fuel composition or boiler demand levels in study. Authors have determined that data mining approach is applicable to different types of burners and fuel types [6]. Figueiredo et al. have presented an electricity consumer characterization framework based on a knowledge discovery in databases procedure, supported by data mining techniques. This framework consists of two main modules: the load profiling module and the classification module. The load profiling module creates a set of consumer classes using a clustering operation and the representative load profiles for each class. The classification module uses this knowledge to build a classification model able to assign different consumers to the existing classes [7]. The electricity price forecast framework, which can predict the normal price as well as the price spikes based on data mining approach by Lu et al. has carried out. The proposed model is based on a mining database including market clearing price, trading hour, electricity demand, electricity supply and reserve. This proposed model is able to generate forecasted price spike, level of spike and associated forecast confidence level [8]. Şencan et al. have used different methods such as linear regression (LR), pace regression (PR), sequential minimal optimization (SMO), M5 model tree, M5’Rules, decision table and back propagation neural network (BPNN) for modelling the absorption heat transformer [9].

Section snippets

Data mining methodologies

Data mining comprises numerical and statistical techniques that can be applied in a wide variety of fields for prediction. A functional general definition of data mining is the use of numerical analysis, visualization or statistical techniques to identify non-trivial numerical relationships within a dataset to derive a better understanding of the data and to predict future results [10].

Data mining is a complex process. Typically it involves the following procedures [11]:

  • Developing an

Application of data mining for thermophysical properties prediction of refrigerants

In this study, different data mining techniques (LR, MLP, PR, SLR, SMO, KStar, AR, M5 model tree, DT, M5’Rules) were used for determining thermophysical properties as the specific heat capacity, viscosity, heat conduction coefficient, density relating to liquid and vapour phase of the refrigerants. Furthermore, comparison of these different modeling techniques was presented. The thermophysical properties of refrigerants used in data mining analysis were taken from Solkane Software. In Table 1,

Results and discussion

In this work, data mining methods were used for determining thermophysical properties as the specific heat capacity, viscosity, heat conduction coefficient, density of R134a, R404a, R407c and R410a refrigerants which do not cause ozone depletion for vapor compression refrigeration systems. The best approaches which have minimum errors were obtained. Therefore, equations of thermophysical properties were derived.

In Table 6, a comparison is given between the Solkane Software results and obtained

Conclusions

In this work, data mining techniques was used for determining the thermophysical properties of R134a, R404a, R407c and R410a refrigerator which do not cause ozone depletion for vapor compression refrigeration systems. Used data mining techniques are LR, MLP, PR, SLR, SMO, KStar, AR, M5 model tree, DT, M5’Rules models. The best approach for specific heat capacity relating to liquid and vapour phase of R134a is SMO model. The best approach for viscosity relating to liquid phase of R134a is PR

References (31)

Cited by (21)

  • Insights into modelling and evaluation of thermodynamic and transport properties of refrigerants using machine-learning methods

    2023, Energy
    Citation Excerpt :

    It should be mentioned these refrigerants, and their mixtures, e.g., HFO-1234yf, have low GWP [32], small compressor size, high coefficient of performance (COP) [33], and maximum cooling capacity [34]. To estimate the thermophysical specifications of refrigerating agents (R134a, R404a, R407c, and R410a), Küçüksille et al. [35] proposed ten modelling procedures within the data mining process. A novel approach was utilised by Mohebbi et al. [36] for auto-designing a neural network according to the genetic algorithm (GA) for forecasting saturated liquid density for 19 pure, and six admixed refrigerating agents.

  • Natural and synthetic refrigerants, global warming: A review

    2018, Renewable and Sustainable Energy Reviews
    Citation Excerpt :

    Several refrigerants screenings and optimization studies are available focusing thermo-physical, heat transfer environmental and economic aspects. Saleh and Wendland [15] proposed physical BACKBONE equation; Kucuksille et al. [102] suggested data mining technique by using WEKA software. Other studies on the topic include neural networks [103], hybrid formulae for thermodynamic properties [104], cost based methods [105], entrainment ratio techniques [106], and comparison of operating performance [3].

  • Prediction of surface tension of binary refrigerant mixtures using artificial neural networks

    2018, Fluid Phase Equilibria
    Citation Excerpt :

    Hydrofluorocarbons (HFCs) and their mixtures have been increasingly considered as the working fluid in heat-pump and air-conditioning systems because of the low ozone layer depletion, great refrigeration performance, and low global warming potentials. And they are considered as the alternative refrigerant instead of chlorofluorocarbons (CFCs) and hydrochlorofluorocarbons (HCFCs) [5,6]. Therefore, the experimental data and correlations of the surface tension are vital to design refrigeration cycles.

  • Cooling mechanism of a solar assisted air conditioner: An investigation based on pressure–enthalpy chart

    2017, International Journal of Refrigeration
    Citation Excerpt :

    Since the evaporator is not the focus of this paper, only information such as cooling capacity and inlet and outlet enthalpies are needed to determine the required refrigerant mass flow rate. Thermodynamic properties, such as density, thermal conductivity, viscosity, isobaric heat capacity and specific enthalpy for both liquid and vapor states, of R-134a and R-410A can be measured using specified correlations as those by Kabelac (1991), Cleland (1994), Lemmon (2003), Küçüksille et al. (2009) and Küçüksille et al. (2011). It must be noted that R-410A has a very slight temperature glide (<1 °C), in which the dew point and bubble point temperatures may be ignored.

View all citing articles on Scopus
View full text