Rule-based Mamdani-type fuzzy modelling of thermal performance of fin-tube evaporator under frost conditions

Frost formation brings about insulating effects over the surface of a heat exchanger and thereby deteriorating total heat transfer of the heat exchanger. In this study, a fin-tube evaporator is modeled by making use of Rule-based Mamdani-Type Fuzzy (RBMTF) logic where total heat transfer, air inlet temperature of 2 °C to 7 °C and four different fluid speed groups (ua1=1; 1.44; 1.88 m s-1, ua2=2.32; 2.76 m s-1, ua3=3.2; 3.64 m s-1, ua4=4.08; 4.52; 4.96 m s-1) for the evaporator were taken into consideration. In the developed RBMTF system, outlet parameter UA was determined using inlet parameters Ta and ua. The RBMTF was trained and tested by using MATLAB® fuzzy logic toolbox. R2 (%) for the training data and test data were found to be 99.91%. With this study, it has been shown that RBMTF model can be reliably used in determination of a total heat transfer of a fin-tube evaporator.


Introduction
An evaporator is a device operating by the vapor compression refrigerating cycle principle and is used for cooling purposes (like cooling the cabin of a refrigerator).If the temperature of the evaporator surface in a refrigerator is below the freezing temperature of the moisture within the air, then frosting forms on the evaporator surface.If the frost keeps staying below the freezing point, then frost will form a layer on the surface.The formed layer acts as an insulator by increasing thermal resistance and thereby leading to dropping the thermal performance of the evaporator [1].In the existing literature, there are many numerical and experimental studies on the performance of finned tube heat exchangers operating under frosting conditions.
Fuzzy logic is a mathematical discipline that we use every day and helps us to reach the structure in which we interpret our own behaviors.Fuzzy expert system is an expert system that uses a collection of fuzzy rules, instead of Boolean logic, to reason about data [2].Over the last few years, there have been many investigations on the application of fuzzy logic.Some are briefly mentioned below.Tasdemir et al. [3] studied artificial neural network (ANN) and fuzzy expert system (FES) and their comparison for prediction of performance and emission parameters on a gasoline engine.Tosun and Dincer [4] carried out a study on modelling of a thermal insulation system based on the coldest temperature conditions by using artificial neural networks to determine the performance of building for wall types in Turkey They noted that in this study the ANN approach has been applied accurately to model for the thickness of thermal insulation performance system on the coldest temperature basis for wall types that are mostly used for buildings in Turkey.Tosun et al. [5] investigated Rulebased Mamdani-type fuzzy modeling (RBMTF) of thermal performance of multi-layer precast concrete panels used in residential buildings in Turkey.They reported that prediction of RBMTF modeling approach can be successfully used for the modeling of thermal performance of multi-layer precast concrete panels used in residential buildings in Turkey.
In this study, the performance of a finned tube evaporator under frost forming conditions was modeled by making use of RBMTF technique.Input parameters taken were air entrance temperature and inlet speed while the output parameter was the total heat transfer of the evaporator.When the experimental data were compared with the data obtained from the RBMTF technique, it was found that the two are in correlation with each other.The UA values at different air operating conditions which were not studied experimentally were estimated with the RBMTF technique.

Experimental setup
The schematic diagram of the experimental system that shows evaporators' total heat transfer in a no-frost refrigerator operating under frosting conditions is given in Figure 1.The experimental setup consists of an air tunnel, refrigerating system and a data logger.The outside diameter of the test evaporator pipe inserted into the air tunnel is 8 mm, its thickness is 0.8 mm and the fin thickness is 0.12 mm.during the tests, the air was drawn by a fan from an aperture and circulated around the channel.In order to regulate the air operating conditions within the air tunnel, a demoisturizer, a cooler, a heater and a humidifier were used.The air amount sent to the evaporator was adjusted with a damper.Airstream straightener was used to maintain homogeneous air dispersion to the evaporator.During the experiments, air inlet and outlet temperatures and the relative humidity were measured with a temperature-humidity measuring device while the air speed was measured with the anemometer.

Air enthalpy at the evaporator inlet and outlet
The enthalpy of the moisture entering the evaporator is found by using Eq.(1) below: Here, cpa, Ta and wa represent specific heat capacity of air, temperature and specific humidity of the air respectively.

Heat transfer for air side
The air temperature on the medium where the channel exists is higher than that of the air within the channel.For this reason, there is a heat gain on the test zone of the channel.Air transfer on the air side is found with Eq. ( 2).
Where, ma, iai, iao and (UA)channel are respectively the inlet air mass flow, inlet air enthalpy, outlet air enthalpy and overall heat transfer coefficient of channel.Te and Ts are the environment and surface temperatures of the channel respectively.

Heat transfer for refrigerant side
The heat transferred from the refrigerant is calculated from the following equation.
Here r m  , iri and iro respectively, represent the mass flow rate of the refrigerant, evaporator inlet and outlet enthalpies.

Evaporator's overall heat transfer coefficient
Eq. ( 4) below, was obtained by making use of the arithmetic mean of the evaporator's heat capacity (qm), refrigerant's heat capacity (qr) and the heat capacity on the air The overall heat transfer coefficient of evaporator (UA), is found with the following equation: Here, ΔTm is logarithmic mean temperature difference and is presented in Eq. ( 6) below: where, Tai and Tao are the inlet and outlet temperatures of air to and from the heat exchanger respectively.Here Tria and Tro are the inlet and outlet temperatures of refrigerant to and from the heat exchanger respectively.

Developed fuzzy expert system for the performance of finned tube evaporator under frosting conditions
Fuzzy logic is a superset of Boolean-conventional logic that has been expanded to handle the concept of partial truth and truth values between ''completely true'' and ''completely false''.Fuzzy theory should be seen as a methodology to generalize any specific theory from crisp to continuous.Fuzzy modeling opens the possibility for straightforward translation of statements in natural language (verbal formulation) of the observed problem into a fuzzy system.Its functioning is based on mathematical tools [6,7].There are two types of fuzzy inference systems in the toolbox: Mamdani-type and Sugeno-type.These two types of inference systems vary somewhat in the way outputs are determined.Fuzzy  stimulus model is constructed into rule-based Mamdanitype fuzzy modeling, using input parameters as Ta, ua and output parameter as UA, described by RBMTF if-then rules (Figure 2).RBMTF has been designed using the MATLAB R2010b fuzzy logic toolbox.Classification of the variables into a specified number of subsets depends on the nature of the problem.The inputs Ta and ua together with the output UA were categorized into 7 subsets and their triangle membership functions were determined.The membership functions are presented in Figures 3 and  4. The membership functions were developed for the study, representing Very Low (L1), Low (L2), Negative Medium (L3), Medium (L4), Positive Medium (L5), High (L6), and Very High (L7) linguistic classes (Table 1).
In this study, the data set included 110 data (training+ test data).70 of them were chosen as training data (Table 2), whereas 40 of them were chosen for the test data (Table 3).Values of prediction consist of 50 sets (Table 4).Fuzzy membership functions in analytical form are expressed in Eqs. ( 7

Results and discussion
If there the surface temperature of evaporators in refrigerating systems, is lower than the freezing point of the water vapor in the air, then a frost film forms on the evaporator surfaces, first; and in case the frost surface temperature continues to be lower than the freezing point, then the frost accumulates over the surface and acts as an insulating agent.Since this situation will increase thermal resistance, the amount of heat absorbed by the refrigerant will decrease, the inter-fin distances will narrow due to the frost accumulation and ultimately, energy consumption will increase [1].In this study, RBMTF technique has been used to model the performance of a finned tube evaporator at the transient regime for 4 situations and with different ua values.The findings of this study are presented below:  Situation for ua=1; 1.44; 1.88 m s -1 : Ta=2-7°C.
The error during the learning session is called the rootmean-squared (RMS) value and is defined as follows [9]: In addition, the absolute fraction of variance (R 2 ) and mean absolute percentage error (MAPE) are defined as follows, respectively [9]: where t is target value, o is output value, and p is pattern [9].
The statistical values such as RMS, R 2 , MAPE, maximum error and minimum error given in Tables 9 and  10 for training and test values.R 2 for the training data is 99.91 % and R 2 for the test data is 99.91% (Figure 9).When Tables 7-8 and Figure 9 are studied, it is found that actual values and the values from fuzzy technique are very close to each other.

Conclusion
Modeling of performance of a finned-tube type evaporator was conducted in this study by making use of the Rule-Based Mamdani Type Fuzzy logic technique.
The evaporator was made to operate under frost forming conditions and the conclusions drawn in this paper are summarized as follows:  The RBMTF was trained and tested by means of the MATLAB software on a personal computer. Input parameters (Ta, ua) and output parameter UA, were described by RBMTF if-then rules.2) and 40 were for the testing session (Table 3). The UA values that were not considered with the experimental study were estimated by using the RBMTF (Figs. 5-8 and Table 4). The decrease in inlet air temperature has caused only a slight increase in the frost layer and that's why as the inlet air temperature falls the total heat transfer has increased.(Figs.5-8). The amount of water vapor in the air has decreased as a result of the decreasing air inlet speed.This has led to an increasing total heat transfer as the air inlet speed fell.(Figs.5-8).
The actual data and RBMTF results show that RBMTF can be successfully used for the modeling of performance of finned tube evaporator under frost conditions.Future studies may use the data obtained in this study and involve other techniques like artificial neural networks (ANN) and genetic algorithm (GA) under different operating conditions to calculate an optimum performance for total heat transfer of a system.

Figure 1 .
Figure 1.Schematic presentation of experimental system

Figure 2 .
Figure 2. Designed fuzzy modeling structure the performance of a finned tube evaporator operating under snowing conditions

Figure 3 .
Figure 3. Fuzzy membership functions for two input variables: (a) T a fuzzy sets graphics; (b) u a fuzzy sets graphics.

Figure 4 .
Figure 4. Fuzzy membership functions for one output variable: UA fuzzy set graphic.

Figure 6 .
Figure 6.Comparison of actual UA with the UA data obtained from fuzzy technique for u a =2.32 ms -1 , u a =2.76 ms -1

Figure 7 .
Figure 7.Comparison of actual UA with the UA data obtained from fuzzy technique for u a =3.2 ms -1 , u a =3.64 ms -1

Figure 8 .
Figure 8.Comparison of actual UA with the UA data obtained from fuzzy technique for u a =4.08 ms -1 , u a =4.96 ms -1

Figure 9 .
Figure 9.Comparison of the actual and RBMTF results


110 experimental data sets were used, out of which 70 were used in the training step (Table

Table 1 .
Fuzzy sets of input and output variables.

Table 9 .
The statistical error values for UA for training

Table 10 .
The statistical error values for UA for test