Advertisement

Journal of Thermal Analysis and Calorimetry

, Volume 135, Issue 2, pp 1269–1283 | Cite as

Application of response surface methodology in optimization of automotive air-conditioning performance operating with SiO2/PAG nanolubricant

  • A. A. M. Redhwan
  • W. H. AzmiEmail author
  • G. Najafi
  • M. Z. Sharif
  • N. N. M. Zawawi
Article

Abstract

The effect of compressor speed, initial refrigerant charge and volume concentrations of SiO2/PAG nanolubricant on the performance of automotive air-conditioning (AAC) system are investigated in this study. Response surface method (RSM) was used in designing the experimental work and is based on face composite design. The developed quadratic models from RSM were helpful to envisage the response parameters namely heat absorbs, compressor works, and coefficient of performance (COP) to identify the significant relations between the input factors and the responses. The results depicted that adding SiO2 nanoparticle into PAG lubricant will enhance the COP of AAC. Optimization of independent variables was performed using the desirability approach of the RSM with the goal of maximizing the heat absorb and COP, consequently, minimizing the compressor work. The results revealed that the optimal condition with a high desirability of 73.4% for the compressor speed of 900 rpm, refrigerant charge of 95 g and volume concentration of 0.07%. At this condition, the AAC system operated with 193.99, 23.28 kJ kg−1 and 8.27, respectively, for heat absorb, compressor work and COP. DoE based on RSM was capable of optimizing the significant parameters which affect AAC performance.

Keywords

Nanolubricant Heat absorb Compressor work COP Response surface method 

List of symbols

AAC

Automotive air-conditioning

ANOVA

Analysis of variance

CCD

Central composite design

COP

Coefficient of performance

EER

Energy efficiency ratio

FCD

Face-centered design

PAG

Polyalkylene glycol

QL

Heat absorb (kJ kg−1)

RAC

Resident air-conditioning

rpm

Revolution per minute

RSM

Response surface method

Win

Compressor work (kJ kg−1)

Greek symbols

ϕ

Volume concentration (%)

ρ

Density (kg m−1)

Introduction

Although air-conditioning is one of the auxiliary components in an automotive system, it has become a necessary part of providing thermal relieve inside car passengers’ compartment, particularly, in nations experiencing hot and humid atmospheres. On the other hand, the compressor of air-conditioning becomes a singular major auxiliary load on an automotive engine. The additional load engaged by compressor will cause a reduction in efficiency, increase in fuel consumption and release of greenhouse gases. The consequences of employing air-conditioning systems in terms of energy reduction and other refrigerant restraint have forced scholars to consider new ways and technologies to improve its efficiency [1]. One of the recent techniques to increase the efficiency of automotive air-conditioning (AAC) is by introducing nanoparticles into the system. The nanoparticle was dispersed directly into a refrigerant base to form nanorefrigerant or into a refrigerant lubricant to create nanolubricants. Redhwan et al. [2] thoroughly reviewed the development of nanorefrigerant and nanolubricant for different refrigerant bases and performance improvement. In another paper, Azmi et al. [3] summarized the potential of nanorefrigerant and nanolubricant on energy saving in a refrigeration system.

Nanorefrigerant researchers have focused their effort on improving the refrigeration performance in domestic refrigerators [4, 5] resident air-conditioning (RAC) system [6] and vapor compression refrigeration (VCR) system [7]. The researchers concluded that the addition of nanoparticles in the refrigeration system reduces the energy consumption up to 26% [5], increase its coefficient of performance (COP) up to 33% [8] and cooling of energy efficiency ratio (EER) up to 6% [6]. In addition, Nair et al. [9] emphasized that employing nanoparticle into the refrigerant base will enhance the efficiency of compressor power, hence, reduce the energy consumption of refrigeration system.

Sabareesh et al. [10] have studied the effect of operating parameters by utilizing an individual approach. The previous research by Frey and Wang [11] of altering a control factor for one time which fit well only in specific conditions should not be considered in the present study. This is due to the performance of AAC since it is influenced by the collective effect of various parameters such as compressor speed, refrigerant charge, concentration of nanolubricant and other factors. Hence, an organized multifactor analysis could offer a comprehensible and detailed understanding of the performance characteristics of the AAC system in comparison with an individual approach. In such multifactor problems, the employment of nonlinear method such as Design of Experiments (DoE) is appropriate to investigate the interaction effects of experiment variables. DoE is regarded as one of the most efficient and cost-effective methods to assess the individual and collective effects of experiment factors on output responses [12]. Several techniques, for instance, factorial design, Taguchi method, and response surface method could be employed for planning the experiments.

In the present paper, response surface methodology (RSM) is employed to investigate the influence of input factors on the response parameters. RSM is a compilation of mathematical and statistical techniques which was employed to ascertain a mathematical representation between factors and responses, and identify the cause of factors affecting a response in a specific process [13]. RSM is also renowned as a practical technique to analyze engineering problems based on both modeling and optimizing the response surface which is affected by the experiments inputs [14]. As the key benefit, the use of RSM in designing the experiment entailed lesser tests and less time-consuming as compared to the full factorial design experimentation. Much of the computation resources are reduced. Consequently, the time needed to resolve the objective problem could be minimized by using RSM. While Taguchi’s method is tool for robust design, it offers a simple and systematic approach to optimize design for performance, quality and cost. Taguchi designs help determining, the parameter settings for experiments that give optimal settings. However, this optimal setting value depends on the number of experiment, since experimentation itself involves great cost. RSM on the other hand/could achieve closer to global optimum with lesser of experiment.

The works related to the application of RSM in the evaluation of the performance of refrigeration system were limited in the literature. One of the recent papers was presented by Costa and Garcia [15]. They used RSM in optimizing the efficiency of a refrigeration cycle demonstration unit using a multi-response optimization method. In their work, Costa and Garcia [15] considered experimental variable parameters such as evaporator temperature, condenser temperature, condenser mass flow rate and evaporator mass flow rate. Statistically designed experiments were carried out to concurrently maximize the refrigeration effect and minimize energy consumption of a compression refrigeration cycle (CR). However, the employment of RSM techniques in AAC system by using nanolubricant is yet to be explored further.

Hence, the present study utilizes RSM approach to investigate the influence of compressor speed, refrigerant charge and volume concentration on the AAC performance operated with SiO2/PAG nanolubricants. Design Expert software is used in the present analysis and the experiments are planned using face-centered design (FCD) procedure. Based on the FCD, 20 experiments were conducted. The heat absorb (QL), compressor work (Win) and coefficient of performance (COP) are used as the response factors in the RSM evaluation.

Methodology

The response surface method (RSM) and desirability approach in achieving the optimal performance are deliberately explained in this section. Prior to that, the preparation and stability of SiO2/PAG nanolubricants is briefly discussed. The experimental setup of AAC system was designed and developed for the performance analysis of SiO2/PAG nanolubricants in the previous study [16, 17]. For further evaluation, the experimental data presented by Sharif et al. [16] are used in the present study.

Material and Formulation of SiO2/PAG Nanolubricant

In the present study, SiO2 (amorphous) nanoparticles with 99.9% purity with an average size of 30 nm were used and procured. Field emission scanning electron microscopy (FESEM) micrograph is used in attaining the morphology, size and characterization of SiO2 nanoparticles. Figure 1a depicted the magnification of 200,000× FESEM image. Figure 1a shows that the nanoparticles are approximately 30 nm in size and the shape is observed to be in spherical shapes. Polyalkylene glycol (PAG) is used as the AAC lubricant. PAG lubricant demonstrates better tribology performance over mineral oils when run together with hydrofluorocarbon (HFC) refrigerant. PAG show little solubility in the gaseous refrigerant and offer excellent lubrication at elevated temperatures and pressures. Further, the compatibility characteristic of PAG with most of the elastomers, this lubricant has been favorably chosen to be used in the automotive air-conditioning (AAC) system [18]. Further properties of SiO2 nanoparticle and PAG lubricant are depicted in Table 1.
Fig. 1

a FESEM image of dry SiO2 nanoparticle at × 200,000 magnifications b TEM image of SiO2 nanoparticle dispersed in PAG lubricant at × 88,000 magnifications

Table 1

Properties of SiO2 nanoparticles [34, 35] and PAG 46 lubricant [36, 37]

Property

Nanoparticle SiO2

Lubricant PAG 46

Purity/%

99.9

Molecular mass/g mol−1

60.08

Average particle diameter/nm

30

Density/kg m−3

2220

995.4

Thermal conductivity/W (m K)−1

1.4

Specific heat/J (kg K)−1 K−1

745

Flash point/°C

174

Kinematic viscosity, cSt @ 40 °C

41.4–50.6

Pour point/°C

− 51

SiO2/PAG nanolubricant was then to be produced by the two-step method process as recommended by Yu and Xie [19]. No surfactant was introduced throughout the preparation process of the nanolubricants. The same approach was done in preparing their nanolubricant [16, 20, 21, 22, 23, 24, 25, 26, 27]. Then, stability analyses are done by visual sedimentation and UV–Vis spectrophotometer [16]. Figure 2 depicted the visual sedimentation test of 0.2–1.0% volume concentration SiO2/PAG. Dispersion stability was observed visually and compared between samples after preparation (Fig. 2a) and after a month of preparation (Fig. 2b). It is found that very minimum sedimentations occurred in the samples through that observation. The outcome of the visual sedimentation was then validated by the UV–Vis spectrophotometer data. UV–Vis spectrophotometer absorbance measurement is a significant technique to evaluate the nanoparticle dispersion in lubricant in terms of colloidal stability [19]. The result of the SiO2/PAG nanolubricant for 2 h sonication is depicted in Fig. 3. From the graph, it shows that the SiO2/PAG nanolubricant is stay stable up to 80% event after 2 weeks. Complementary, transmission electron microscopy (TEM) investigation was carried out to verify the state of SiO2 nanoparticle dispersion and agglomeration in the lubricants. Figure 1b shows the TEM photo of SiO2 nanoparticles dispersed in the PAG lubricant for magnification of 88,000×. Critical observation on the TEM photo, the SiO2 nanoparticle is well dispersed in the PAG lubricant. Nevertheless, small portion of agglomeration and minimum clustering of SiO2 nanoparticles are observed in the PAG lubricant as depicted in Fig. 1b.
Fig. 2

SiO2/PAG nanolubricant samples after a month of preparation. a Samples after a day of preparation, b Samples after a month of preparation

Fig. 3

SiO2/PAG nanolubricant with different sonication time

Experimental design employing RSM

Three experiment variables in the present study including the compressor speed, refrigerant charge, and volume concentration of SiO2/PAG nanolubricant were regarded as effective factors on the heat absorb (QL), compressor work (Win) and coefficient of performance (COP) as responses. Designs that can fit model must have at least three different levels in each factor. This is fulfilled by central composite designs (CCD) which has three levels per variable. Since the region of concern and area of operability is almost identical, the face-centered design (FCD) which consider the eight corners of the cube are centered and scaled to (+ 1, + 1, + 1) and α = 1. FCD was engaged for the present study to attain the experimental data, which would suit full second-order polynomial models representing the response surfaces above a comparatively wide range of parameters. In CCD, the number of experiment point is determined by using Eq. (1).
$$N = 2^{\rm{n}} + 2n + n_{0}$$
(1)
where N is the number of running test of experiment, n is the number of factors and n0 is the number of central points. In Eq. (1), the “2n” term is identified as factorial experiment points. These points permit apparent approximates of all major causes and 2-factors interrelations. Meanwhile, “2n” term is known as axial points which permit the pure quadratic effects estimations. Finally, “n0” represents the center point and can be designed to be run simultaneously both as axial and factorial points.
Every variable was categorized into three levels: the high level (+ 1), the low level (− 1), the center points (coded level as 0). In the recent study, FCD with three factors was noted to have a total of 20 runs of experiments which consist of eight factorial points, six axial points, and six central points. It was employed to evaluate the data attained from the experimental work. The maximum and minimum values of speed, refrigerant charge and volume concentration were considered and the full experimental plan with coded and actual value is shown in Table 2. A multiple regression analysis was employed to attain the coefficients. In RSM method, a model is determined for each dependent variable that represents the main and interaction effects of factors on each variable separately [28]. The multivariate model can be written as [28].
$$Z = \beta_{0} + \sum\limits_{i = 1}^{3} {\beta_{\text{i}} x_{\text{i}} + } \sum\limits_{i = 1}^{3} {\beta_{\text{ii}} x_{\text{i}} x_{\text{i}} + } \sum\limits_{i = 1}^{3} {\beta_{\text{ij}} x_{\text{i}} x_{\text{j}} }$$
(2)
where β0 is the intercept, βi is the linear regression coefficient of the ith factor, βii is the quadratic regression coefficients of the ith factor, βij is the interaction of the ith and jth factors, and Z is the dependent variable [28, 29]. Later, the equations were used to envisage the responses. The correlation between the factors and responses was attained by applying a statistically significant model.
Table 2

Independent variables and levels for central composite design

Independent variables

Code

Variable levels

− 1

0

1

Speed

A

900

1500

2100

Refrigerant charge

B

95

110

125

Volume concentration

C

0

0.05

0.1

Desirability approach

The realistic issues situations entailed optimization with varying responses of concerns. Constrained optimization problems, overlaying the contour plots for each response and desirability approach are being employed. Desirability approach is a preferred choice among optimization methods since it has beneficial characteristic such as unfussiness, accessibility in the software, suppleness in weighting and giving importance ranking for individual response [14]. The same approach was also done by Khoobbakht et al. [30]. The researcher specifies the maximum, minimum, and/or preferred values acceptable for each of the original fitted response functions together with the weights that govern the rate at which the desirability of a function increases or decreases over the range of acceptable values. This method transforms each of the fitted functions into a desirability value and then the desirability values for each response are combined using the geometric mean into a single objective to be optimized. Therefore, the present research work employed desirability approach in optimizing experiment variables, namely, compressor speed, refrigerant charge and volume concentration for the response properties of heat absorb, compressor work and COP. The optimization step was done using Design Expert software where each response is changed to a dimensionless desirability value in the range between 0 and 1. The dimensionless value of d = 0 suggested that the response is totally unacceptable while d = 1 suggests that the response is most desirable. In the present study, each response has its own goal, either to maximize or minimize. The response desirability is then collectively combined using the geometric approach which eventually presents the total general desirability, D. Desirability approach proposed by Derringer and Suich [31] can be summarized in mathematical form as shown in Eq. (3).
$$D = \left( {d_{1} \times d_{2} \times \cdots \times d_{\text{n}} } \right)^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 n}}\right.\kern-0pt} \!\lower0.7ex\hbox{$\rm{n}$}}}} = \left( {\prod\limits_{i = 1}^{n} {d_{\text{i}} } } \right)^{{{\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 n}}\right.\kern-0pt} \!\lower0.7ex\hbox{$\rm{n}$}}}}$$
(3)
The Di ranges from 0 to 1 (least to most desirable, respectively), represents the desirability of each (i) response, and n is the number of responses being optimized.

Results, analysis, and discussion

Experimental results

In the present study, the experimental run was conducted using face-centered design (FCD). The experimental responses, namely, heat absorb, compressor work and coefficient of performance (COP). With three responses (n = 3) and six central point (n0 = 6), the experiment requires 20 runs of experiments. These parameters are nominated at three levels; also, the statistical model table (with 20 runs) is defined based on the number of the parameters and their levels as discussed by Shirvan et al. [29]. The design matrix along with their corresponding points on fitted models based on RSM is summarized in Table 3.
Table 3

The experimental design, result and prediction based on RSM

Run

Process parameter settings

Response

Speed/rpm

Refrigerant charge/g

Concentration/%

Experimental

Predicted

QL/kJ kg−1

Win/kJ kg−1

COP

QL/kJ kg−1

Win/kJ kg−1

COP

1

1500

110

0.05

189.69

31.20

6.08

189.40

31.34

6.03

2

900

125

0.00

188.81

25.85

7.30

188.91

25.94

7.39

3

2100

110

0.05

188.40

38.90

4.84

187.83

40.48

4.96

4

2100

125

0.10

188.60

38.70

4.87

188.38

38.25

4.97

5

1500

95

0.05

193.23

34.30

5.63

192.33

32.55

5.86

6

1500

125

0.05

188.06

30.60

6.15

188.65

30.13

6.20

7

1500

110

0.05

188.69

30.95

6.10

189.40

31.34

6.03

8

1500

110

0.00

188.98

38.15

4.95

188.62

37.67

4.91

9

2100

95

0.00

190.13

49.81

3.82

190.52

49.41

3.67

10

1500

110

0.05

189.19

32.00

5.91

189.40

31.34

6.03

11

900

110

0.05

191.36

22.50

8.50

190.96

22.20

8.35

12

2100

125

0.00

185.92

46.83

3.97

185.77

47.00

4.00

13

900

95

0.10

194.17

24.50

7.93

194.15

25.19

8.02

14

900

125

0.10

191.84

22.70

8.45

191.52

22.77

8.35

15

2100

95

0.10

190.79

40.40

4.72

191.01

40.67

4.63

16

1500

110

0.05

188.90

30.40

6.21

189.40

31.34

6.03

17

1500

110

0.05

190.01

31.30

6.07

189.40

31.34

6.03

18

900

95

0.00

193.35

27.74

6.97

193.66

28.35

7.05

19

1500

110

0.10

189.53

32.30

5.87

190.17

31.72

5.87

20

1500

110

0.05

189.20

31.25

6.05

189.40

31.34

6.03

Analysis of data

The analysis of response data (heat absorb, compressor work and COP) by using analysis of variance (ANOVA) and response surface method were explained in the next subsection. The ANOVA is performed using Design Expert software. The results obtained are used in the software for statistical accuracy and to develop a model equation for the responses. The normal probability plots presented in Fig. 4a–c are the results of normality testing for the experimental results. The figures show the predicted versus actual values for the design matrix. For any ANOVA, the normal probability plot should be checked for the range of residuals which should lie close to the mean line. As a result, Fig. 4a–c shows that the values of residuals are very small and closely fitted to the mean line depicted in the graph since it is normally fitted for all responses.
Fig. 4

Normal probability plot of a heat absorb, b compressor work, c coefficient of performance

Heat absorb (Q L)

The model summary of quadratic model and the analysis of variance (ANOVA) for heat absorb is carried out and represented in Table 4. The fit of the model was also articulated as the coefficient of determination R2 as shown in Table 4. The value point out 0.9488 (94.88%) of the variability in response that can be elucidated by the model for the heat absorb. The closer the R2 value to 1, the better the models fits the experimental data [32]. Table 4 also shows the predicted R-squared value of 0.9076, which is in reasonable agreement with adjusted R-squared value of 0.9305 in which the difference between these values is less than 0.2 as desired. The adequate precision value for heat absorb is 28.839 as depicted in Table 4 where this parameter reflects the signal-to-noise ratio. A value greater than 4 is sought. In addition, the value achieved above 28.839 shows that it is an adequate signal and can be used to navigate the design space. As shown in Table 4, the model sum of squares and f values are 73.30 and 23.37, respectively, which show that the model is significant. The p value was provided as an instrument to scrutinize the significance of each coefficient. The values of “Prob > f” less than 0.05 are preferable and indicate that the model terms are significant. If the value of “Prob > f” is more than 0.1, it shows that the model terms are not significant. In this case, the p value is less 0.0001 which means that there is a maximum of 0.01% chance that the “Model f value” is occurring due to noise. In terms of factors, the Speed (A), refrigerant charge (B), volume concentration (C) and a combination of BC and B2 show significance with “Prob > f” values to be less than 0.05. Other combinations of factors models are not significant with “Prob > f” values of more than 0.100. Therefore, the heat absorb is, \(Q_{\text{L}} = f(A,B,C,BC,B^{2} )\). Contrary to models values, the “Lack of Fit” value required to be not significant and desire the “Prob > f” value of more than 0.100. Model reduction is done by removing the not significant factors and the act will improve the model. For heat absorb, the f value is 1.89 which indicates that that the “Lack of Fit” is not significant relative to the pure error. There is a 25.16% chance that a “Lack of Fit” of f value could occur due to noise. Non-significant “Lack of Fit” is desired in order to make certain model fits. The developed actual quadratic models of heat absorb (QL) as fitted based on RSM in terms of the experimental factors is shown in Eq. (4).
$$Q_{\text{L}} = 268.84589 - \left( {2.61508 \times 10^{ - 3} } \right)A - 1.22905B - 62.16028C + 0.70570BC + \left( {4.86729 \times 10^{ - 3} } \right)B^{2}$$
(4)
where QL is heat absorb (kJ kg−1), A is the speed (rpm), B is the refrigerant charge (g), and C is the volume concentration of SiO2/PAG nanolubricant (%). A positive sign of the coefficient in the equation reflects a synergistic outcome, while a negative sign indicates that it has an antagonistic effect toward the analyzed response [30]. The predicted value of QL determined by Eq. (4) was adequately close to the experimental values.
Table 4

Model summary and ANOVA for heat absorb response surface quadratic model

Source

Model summary

Sum of squares

df

Mean squares

f value

p value prob > f

Remarks

R 2

0.9488

Adjusted R2

0.9305

Predicted R2

0.9076

Closed to Adj. R2

Adequate precision

28.839

> 4

Model

73.30

9

8.14

23.37

< 0.0001

Significant

A-Speed

24.62

1

24.62

70.63

< 0.0001

Significant

B-Ref. charge

34.02

1

34.02

97.60

< 0.0001

Significant

C-Concentration

5.98

1

5.98

17.16

0.0020

Significant

AB

0.027

1

0.027

0.076

0.7879

Not significant

AC

0.032

1

0.032

0.093

0.7665

Not significant

BC

2.24

1

2.24

6.43

0.0296

Significant

A2

0.28

1

0.28

0.81

0.3897

Not significant

B2

3.24

1

3.24

9.29

0.0123

Significant

C2

0.25

1

0.25

0.73

0.4139

Not significant

Residual

3.49

10

0.35

Lack of fit

2.28

5

0.46

1.89

0.2516

Not significant

Pure error

1.21

5

0.24

Based on ANOVA, refrigerant charge shows a significant interaction effect with the rest of variables. The heat absorbs gradually decreased with the increment of refrigerant charge. This is mainly due to the reduction of the superheat as the increase of refrigerant charge [33]. The influence of refrigerant charge and volume concentration is shown in Fig. 5. The curvatures of plots indicate the interaction between the variables as shown in Fig. 5. At 0% volume concentration, the increasing of refrigerant charge effect has caused a significant decrease in heat absorb. Furthermore, increasing the volume concentration will reduce the rate of heat absorb decrement when refrigerant charge is increased. This can be seen at 0.10% volume concentration, where across the refrigerant charge, the heat absorb value is slowly decreasing.
Fig. 5

Influence of volume concentration and refrigerant charge on heat absorb for SiO2/PAG nanolubricant a 3D surface plot, b contour plot

Compressor work (W in)

The model summary of a quadratic model and the analysis of variance (ANOVA) for compressor work was carried out as presented in Table 5. The coefficient of determination R2 for compressor work was found to be 0.9909 as shown in Table 5 The value revealed 99.09% of the variability in response can be explained by the model. Furthermore, Table 5 shows the predicted R-squared value of 0.9826, which is in reasonable agreement with the adjusted R-squared value of 0.9877. The adequate precision value for compressor work is 61.110 as shown in Table 5 which is greater than 4. As shown in Table 5, the f value is 198.54 which shows that the model is significant. In terms of factors, speed (A), refrigerant charge (B), volume concentration (C) and a combination of AC and C2 show significance with “Prob > f” values of less than 0.05. Other combinations of factors models are not significant with “Prob > f” values to be more than 0.100. Therefore, the compressor work is, \(W_{\text{in}} = f(A,B,C,AC,C^{2} )\). Furthermore, for the “lack of fit” value for compressor work, the f value is 1.86 which means that the “lack of fit” is not significant relative to pure error. The developed actual quadratic model of compressor work (Win) as fitted based on RSM in terms of the experimental factors corresponded to Eq. (4). The predicted value of Win determined by Eq. (5) was adequately close to the experimental values.
$$W_{\text{in}} = 20.20367 + 0.017548A - 0.080461B - 124.15907C - 0.046458AC + 1342.97840C^{2}$$
(5)
Based on ANOVA, volume concentration reflects a significant interaction effect with some of the factors. The influence of refrigerant charge and volume concentration, speed and refrigerant charge, volume concentration with speed on compressor work are shown in Fig. 6. The curvatures of plots indicate the interaction between the variables. The interaction between refrigerant charge and volume concentration, and the relationship between speed and refrigerant charge show no significant relation between these two factors as no curvature was displayed. This is also supported by Eq. (5). The speed and volume concentration factors show good influence to work compressor and interaction between these two factors are shown in Fig. 6.
Table 5

Model summary and ANOVA for compressor work response surface quadratic model

Source

Model summary

Sum of squares

df

Mean squares

f value

p value prob > f

Remarks

R 2

0.9909

Adjusted R2

0.9877

Predicted R2

0.9826

Closed to Adj. R2

Adequate precision

61.110

> 4

Model

1013.18

9

112.58

198.54

< 0.0001

Significant

A-Speed

834.48

1

834.48

1471.73

< 0.0001

Significant

B-Ref. charge

14.57

1

14.57

25.69

0.0005

Significant

C-Concentration

88.65

1

88.65

156.35

< 0.0001

Significant

AB

0.13

1

0.13

0.22

0.6487

Not significant

AC

15.54

1

15.54

27.41

0.0004

Significant

BC

0.23

1

0.23

0.41

0.5348

Not significant

A2

2.80

1

2.80

4.93

0.0506

Not significant

B2

1.51

1

1.51

2.67

0.1336

Not significant

C2

34.03

1

34.03

60.02

< 0.0001

Significant

Residual

5.67

10

0.57

Lack of fit

4.32

5

0.86

3.19

0.1144

Not significant

Pure error

1.35

5

0.27

Fig. 6

Influence of speed and volume concentration on compressor work for SiO2/PAG nanolubricant a 3D surface plot, b contour plot

Coefficient of performance

Table 6 shows the ANOVA of coefficient of performance attained when SiO2/PAG nanolubricant is used in the compressor of AAC system. It summarizes the consequence of individual influences and their interactions fitted as a second-order quadratic model for coefficient of performance (COP). The model is developed for 95% confidence level with suitable model reduction. The model f value for SiO2/PAG nanolubricant is 275.05. The p value of < 0.0001 means that the model is significant with negligible influence of noise for SiO2/PAG nanolubricant. From the further assessment of f and p values, it can be seen that factor C (volume concentration) has the most significant effect on COP. It was followed by factor A (speed) and finally factor B (refrigerant charge) with less significance compared to other factors. The combinations of different factors also show no significance as depicted in Table 6. Nevertheless, the squared factors of speed (A2) and concentration (C2) show that the model is significant. Therefore, the coefficient of performance, \({\text{COP}} = f(A,B,C,A^{2} ,C^{2} )\). The statistical accuracy is also been examined to make certain the predictive power of the model, which is shown in Table 6. The estimated R2 value for COP is 0.9932 for SiO2/PAG nanolubricant. This implies that the adequate representation (99.32%) of the actual relationships among the various experimental factors for the model. Adequate precision which quantifies signal-to-noise ratio is found to be 67.386 as shown in Table 6. This value reaffirms the accuracy of the model since the ratio is greater than 4. Hence, this model can be employed to navigate the design space. Table 6 also shows the predicted R-squared value of 0.9853, which is in reasonable agreement with adjusted R-squared value of 0.9907 in which the difference between these values is less than 0.2 as desired.
Table 6

Model summary and ANOVA for coefficient of performance (COP) response surface quadratic model

Source

Model summary

Sum of squares

df

Mean squares

f value

p value Prob > f

Remarks

R 2

0.9932

Adjusted R2

0.9907

Predicted R2

0.9853

Closed to Adj. R2

Adequate precision

67.386

> 4

Model

32.96

9

3.66

275.05

< 0.0001

Significant

A-Speed

28.66

1

28.66

2152.63

< 0.0001

Significant

B- Ref. charge

0.28

1

0.28

21.07

0.0010

Significant

C-Vol. concentration

2.33

1

2.33

174.77

< 0.0001

Significant

AB

0.038

1

0.038

2.88

0.1203

Not significant

AC

0.011

1

0.011

0.81

0.3907

Not significant

BC

4.548 × 10−3

1

4.548 × 10−3

0.34

0.5719

Not significant

A2

1.22

1

1.22

91.50

< 0.0001

Significant

B2

0.039

1

0.039

2.92

0.1185

Not significant

C2

0.98

1

0.98

73.84

< 0.0001

Significant

Residual

0.13

10

0.013

Lack of fit

0.087

5

0.017

1.86

0.2569

Not significant

Pure error

0.047

5

9.325 × 10−3

The developed quadratic model of the coefficient of performance (COP) as fitted based on RSM in terms of the experimental factors corresponded to Eq. (6). The predicted value of COP determined by Eq. (6) was adequately close to the experimental values.
$${\text{COP}} = 11.79213 - (7.99688 \times 10^{ - 3} )A + \, 0.011165B + 35.34651C + (1.72508 \times 10^{ - 6} )A^{2} - 256.98633 \, C^{2}$$
(6)
The interactions of speed, refrigeration charge and volume concentrations affecting the coefficient of performance (COP) are shown in Fig. 7. Although there is curvature shown at an actual speed of 1500 rpm, the influence of volume concentration and refrigerant charge on the coefficient of performance (COP) shows less significance. Further, the influence of refrigerant charge and speed on COP shown the influence of these two factors showed less significance on COP. Interestingly, a significant relationship between speed and volume concentration influence on COP is observed in Fig. 7. Initially, at low speed (900 rpm), the increment of volume concentration shows a small effect on the rate of change of COP. Furthermore, the rate of change of COP becomes more significant by increasing the compressor speed. As shown in Fig. 7b for contour graph, the optimum COP becomes narrower at high speed and it is between 0.06 and 0.07% of volume concentration.
Fig. 7

Influence of speed and volume concentration on COP for SiO2/PAG nanolubricant a 3D surface plot, b contour plot

Optimization

Referring to Table 3, the best experimental results for each response (heat absorb, compressor work and coefficient of performance) were at run no. 11 and run no. 13, respectively. Unfortunately, these data show only the best result for each individual response. For example, experimental run no. 13 gives the best result for heat absorb which is 194.17 kJ kg−1, but it is not applicable for compressor work (24.5 kJ kg−1) and coefficient of performance (7.93). Hence, a desirability investigation was carried out in order to identify the optimal conditions by considering all responses. Using design expert software, numerical optimization was executed to find the design space by employing the mathematical models to identify the factor settings that suit the defined targets. The target for speed, refrigerant charge and volume concentration is set as “in range” which stated a range for tolerable variable within upper and lower limit. The target for heat absorb and coefficient of performance (COP) is set as “Maximize” which sets the upper limit as desired best result while compressor work is set as “Minimize” which sets the lower limit as desired best result. The optimization investigation was carried out based on desirability analysis. The total desirability (D) is the geometric (multiplicative) mean of all individual desirability ranging from 0 (least) to 1 (most). The best combination of parameters was chosen based on the highest desirability value as shown in Table 7. The highest desirability value is 0.734 which reflected the maximum heat absorb and coefficient of performance (COP) and minimum value of compressor work as depicted in Fig. 8. The predicted results of desirability were then validated through experiments as depicted in Table 8. From Table 8, it can be concluded that the predicted result with certain desirability is compared. These are the outcome of the optimum conditions taking into account all responses. The results show a close agreement between the predicted and experimental results with a maximum error of 3.69%.
Table 7

Optimum heat absorb, compressor work and COP under different volume concentration

No.

Factors

Response

Desirability

Speed/rpm

Refrigerant charge/g

Concentration/%

QL/kJ kg−1

Win/kJ kg−1

COP

1

900

95

0.07

193.99

23.28

8.27

0.734

2

900

95

0.05

193.92

23.29

8.22

0.720

Fig. 8

Desirability plot considering the heat absorb, compressor work and COP responses

Table 8

Comparison of predicted and experimental result

No.

Factors

Desirability

Q L

W in

COP

Pred./kJ kg−1

Exp./kJ kg−1

Error/%

Pred./kJ kg−1

Exp./kJ kg−1

Error/%

Pred.

Exp.

Error/%

1

(Refer Table 7)

0.734

193.99

196.15

1.11

23.28

23.98

3.01

8.27

8.18

1.09

2

0.720

193.92

198.97

2.60

23.29

24.15

3.69

8.22

8.24

0.23

Conclusions

The optimization of operating parameters for an automotive air-conditioning system (AAC) was performed in the present work by varying the compressor speed, refrigerant charge and volume concentration of SiO2/PAG nanolubricant. The design of experiments (DOE) based on response surface methodology (RSM) was helpful in designing the experiment and the statistical analysis in order to distinguish the significant variables which will contribute to the coefficient of performance of AAC system. This design of experiment drastically reduced the time required by reducing the number of experiments to be carried out and represent statistically proven models for all the responses. Desirability method of the response surface methodology (RSM) was believed to be the most efficient and simplest optimization technique in the present study. A high desirability of 73.4% was achieved at the compressor speed of 900 rpm, refrigerant charge of 95 g and nanolubricant volume concentration of 0.07%. This situation was regarded as the optimum parameter for the AAC system having heat absorb (QL) of 193.99 kJ kg−1, compressor work (Win) of 23.28 kJ kg−1 and COP of 8.27.

Notes

Acknowledgements

The authors are grateful to the Universiti Malaysia Pahang (www.ump.edu.my) for financial supports given under RDU160395 and PGRS170374. The authors also thank to the research team from Automotive Engineering Centre (EAC) and Advanced Automotive Liquids Laboratory (A2LL), who provided insight and expertise that greatly assisted in the present research work.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.

References

  1. 1.
    Di Battista D, Cipollone R. High efficiency air conditioning model based analysis for the automotive sector. Int J Refrig. 2016;64:108–22.  https://doi.org/10.1016/j.ijrefrig.2015.12.014.CrossRefGoogle Scholar
  2. 2.
    Redhwan AAM, Azmi WH, Sharif MZ, Mamat R. Development of nanorefrigerants for various types of refrigerant based: a comprehensive review on performance. Int Commun Heat Mass Transf. 2016;76:285–93.  https://doi.org/10.1016/j.icheatmasstransfer.2016.06.007.CrossRefGoogle Scholar
  3. 3.
    Azmi WH, Sharif MZ, Yusof TM, Mamat R, Redhwan AAM. Potential of nanorefrigerant and nanolubricant on energy saving in refrigeration system—a review. Renew Sustain Energy Rev. 2017;69:415–28.  https://doi.org/10.1016/j.rser.2016.11.207.CrossRefGoogle Scholar
  4. 4.
    Bi S, Guo K, Liu Z, Wu J. Performance of a domestic refrigerator using TiO2-R600a nano-refrigerant as working fluid. Energy Convers Manag. 2011;52(1):733–7.CrossRefGoogle Scholar
  5. 5.
    Bi SS, Shi L, Zhang LL. Application of nanoparticles in domestic refrigerators. Appl Therm Eng. 2008;28(14):1834–43.CrossRefGoogle Scholar
  6. 6.
    Wang R, Wu Q, Wu Y. Use of nanoparticles to make mineral oil lubricants feasible for use in a residential air conditioner employing hydro-fluorocarbons refrigerants. Energy Build. 2010;42(11):2111–7.CrossRefGoogle Scholar
  7. 7.
    Kumar DS, Elansezhian RD. Experimental study on Al2O3-R134a nano refrigerant in refrigeration system. Int J Mod Eng Res. 2012;2(5):3927–9.Google Scholar
  8. 8.
    Subramani N, Prakash MJ. Experimental studies on a vapour compression system using nanorefrigerants. Int J Eng Sci Technol. 2011;3(9):95–102.Google Scholar
  9. 9.
    Nair V, Tailor PR, Parekh AD. Nanorefrigerants: a comprehensive review on its past, present and future. Int J Refrig. 2016;67:290–307.  https://doi.org/10.1016/j.ijrefrig.2016.01.011.CrossRefGoogle Scholar
  10. 10.
    Sabareesh RK, Gobinath N, Sajith V, Das S, Sobhan CB. Application of TiO2 nanoparticles as a lubricant-additive for vapor compression refrigeration systems—an experimental investigation. Int J Refrig. 2012;35(7):1989–96.CrossRefGoogle Scholar
  11. 11.
    Frey DD, Wang H. Adaptive one-factor-at-a-time experimentation and expected value of improvement. Technometrics. 2006;48(3):418–31.  https://doi.org/10.1198/004017006000000075.CrossRefGoogle Scholar
  12. 12.
    Pandian M, Sivapirakasam SP, Udayakumar M. Investigation on the effect of injection system parameters on performance and emission characteristics of a twin cylinder compression ignition direct injection engine fuelled with pongamia biodiesel–diesel blend using response surface methodology. Appl Energy. 2011;88(8):2663–76.  https://doi.org/10.1016/j.apenergy.2011.01.069.CrossRefGoogle Scholar
  13. 13.
    Myers RH, Montgomery DC, Anderson-Cook CM. Response surface methodology: process and product optimization using designed experiments. 3rd ed. New Jersey: Wiley; 2016.Google Scholar
  14. 14.
    Karimi F, Rafiee S, Taheri-Garavand A, Karimi M. Optimization of an air drying process for Artemisia absinthium leaves using response surface and artificial neural network models. J Taiwan Inst Chem Eng. 2012;43(1):29–39.  https://doi.org/10.1016/j.jtice.2011.04.005.CrossRefGoogle Scholar
  15. 15.
    Costa N, Garcia J. Using a multiple response optimization approach to optimize the coefficient of performance. Appl Therm Eng. 2016;96:137–43.  https://doi.org/10.1016/j.applthermaleng.2015.11.080.CrossRefGoogle Scholar
  16. 16.
    Sharif MZ, Azmi WH, Redhwan AAM, Mamat R, Yusof TM. Performance analysis of SiO2/PAG nanolubricant in automotive air conditioning system. Int J Refrig. 2017;75:204–16.CrossRefGoogle Scholar
  17. 17.
    Redhwan AAM, Azmi WH, Sharif MZ, Hagos FY, editors. Development of nanolubricant automotive air conditioning (AAC) test rig. MATEC web of conferences; 2016.Google Scholar
  18. 18.
    Matlock PL, Brown WL, Clinton NA. Polyalkylene glycols. New York: Chemical Industries-Marcel Dekker; 1999. p. 159–94.Google Scholar
  19. 19.
    Yu W, Xie H. A review on nanofluids: preparation, stability mechanisms, and applications. J Nanomater. 2012;2012:1.Google Scholar
  20. 20.
    Redhwan AAM, Azmi WH, Sharif MZ, Zawawi NNM, editors. Thermal conductivity enhancement of Al2O3 and SiO2 nanolubricants for application in automotive air conditioning (AAC) system. MATEC web of conferences; 2016.Google Scholar
  21. 21.
    Sharif MZ, Azmi WH, Redhwan AAM, Mamat R. Investigation of thermal conductivity and viscosity of Al2O3/PAG nanolubricant for application in automotive air conditioning system. Int J Refrig. 2016.  https://doi.org/10.1016/j.ijrefrig.2016.06.025.Google Scholar
  22. 22.
    Sharif MZ, Azmi WH, Redhwan AAM, Zawawi NMM, editors. Preparation and stability of silicone dioxide dispersed in polyalkylene glycol based nanolubricants. MATEC web of conferences; 2016.Google Scholar
  23. 23.
    Sharif MZ, Azmi WH, Redhwan AAM, Zawawi NNM, Mamat R. Improvement of nanofluid stability using 4-step UV–vis spectral absorbency analysis. J Mech Eng. 2017;4(2):233–47.Google Scholar
  24. 24.
    Redhwan AAM, Azmi WH, Sharif MZ, Mamat R, Zawawi NNM. Comparative study of thermo-physical properties of SiO2and Al2O3 nanoparticles dispersed in PAG lubricant. Appl Therm Eng. 2017;116:823–32.  https://doi.org/10.1016/j.applthermaleng.2017.01.108.CrossRefGoogle Scholar
  25. 25.
    Abdolbaqi MK, Sidik NAC, Aziz A, Mamat R, Azmi WH, Yazid MNAWM, et al. An experimental determination of thermal conductivity and viscosity of BioGlycol/water based TiO2 nanofluids. Int Commun Heat Mass Transf. 2016;77:22–32.  https://doi.org/10.1016/j.icheatmasstransfer.2016.07.007.CrossRefGoogle Scholar
  26. 26.
    Azmi W, Hamid KA, Mamat R, Sharma K, Mohamad M. Effects of working temperature on thermo-physical properties and forced convection heat transfer of TiO2 nanofluids in water–ethylene glycol mixture. Appl Therm Eng. 2016;106:1190–9.CrossRefGoogle Scholar
  27. 27.
    Khdher AM, Sidik NAC, Hamzah WAW, Mamat R. An experimental determination of thermal conductivity and electrical conductivity of bio glycol based Al2O3 nanofluids and development of new correlation. Int Commun Heat Mass Transf. 2016;73:75–83.  https://doi.org/10.1016/j.icheatmasstransfer.2016.02.006.CrossRefGoogle Scholar
  28. 28.
    Shirvan KM, Mamourian M, Mirzakhanlari S, Ellahi R. Numerical investigation of heat exchanger effectiveness in a double pipe heat exchanger filled with nanofluid: a sensitivity analysis by response surface methodology. Powder Technol. 2017;313:99–111.CrossRefGoogle Scholar
  29. 29.
    Shirvan KM, Ellahi R, Mirzakhanlari S, Mamourian M. Enhancement of heat transfer and heat exchanger effectiveness in a double pipe heat exchanger filled with porous media: numerical simulation and sensitivity analysis of turbulent fluid flow. Appl Therm Eng. 2016;109:761–74.CrossRefGoogle Scholar
  30. 30.
    Khoobbakht G, Najafi G, Karimi M, Akram A. Optimization of operating factors and blended levels of diesel, biodiesel and ethanol fuels to minimize exhaust emissions of diesel engine using response surface methodology. Appl Therm Eng. 2016;99:1006–17.  https://doi.org/10.1016/j.applthermaleng.2015.12.143.CrossRefGoogle Scholar
  31. 31.
    Derringer G, Suich R. Simultaneous optimization of several response variables. J Qual Technol. 1980;12(4):214–9.CrossRefGoogle Scholar
  32. 32.
    Oh BR, Seo JW, Choi MH, Kim CH. Optimization of culture conditions for 1,3-propanediol production from crude glycerol by Klebsiella pneumoniae using response surface methodology. Biotechnol Bioprocess Eng. 2008;13(6):666–70.CrossRefGoogle Scholar
  33. 33.
    Choi JM, Kim YC. The effects of improper refrigerant charge on the performance of a heat pump with an electronic expansion valve and capillary tube. Energy. 2002;27(4):391–404.CrossRefGoogle Scholar
  34. 34.
    Vajjha RS, Das DK, Kulkarni DP. Development of new correlations for convective heat transfer and friction factor in turbulent regime for nanofluids. Int J Heat Mass Transf. 2010;53(21–22):4607–18.  https://doi.org/10.1016/j.ijheatmasstransfer.2010.06.032.CrossRefGoogle Scholar
  35. 35.
    Azmi WH, Sharma KV, Sarma PK, Mamat R, Najafi G. Heat transfer and friction factor of water based TiO2 and SiO2 nanofluids under turbulent flow in a tube. Int Commun Heat Mass Transf. 2014;59:30–8.CrossRefGoogle Scholar
  36. 36.
    Brown WL. Polyalkylene glycols. CRC Handb Lubr Tribol. 1993;3:253–67.Google Scholar
  37. 37.
    Dow. Safety data sheet: UCON™ refrigeration lubricant 213. USA: The Dow Chemical Company. 2013. p. 1–11.Google Scholar

Copyright information

© Akadémiai Kiadó, Budapest, Hungary 2018

Authors and Affiliations

  • A. A. M. Redhwan
    • 1
    • 3
  • W. H. Azmi
    • 1
    • 2
    Email author
  • G. Najafi
    • 4
  • M. Z. Sharif
    • 1
  • N. N. M. Zawawi
    • 1
  1. 1.Faculty of Mechanical EngineeringUniversiti Malaysia PahangPekanMalaysia
  2. 2.Automotive Engineering Centre, Universiti Malaysia PahangPekanMalaysia
  3. 3.Faculty of Manufacturing Engineering TechnologyTATI University CollegeKemamanMalaysia
  4. 4.Tarbiat Modares UniversityTehranIran

Personalised recommendations