Prediction of Heat Transfer during Condensation of Ammonia Inside Tubes and Annuli

Abstract

Ammonia has been used as a refrigerant since the beginning of the refrigeration industry and it continues to be an important industrial refrigerant. However, there is no well-verified method for predicting heat transfer during condensation in tubes and annuli. Available information is often contradictory, especially about the effect of oil. In this paper, available test data and predictive methods are reviewed. Reliable test data are identified and compared to well-known general correlations. The effect of oil on heat transfer is investigated. The results of this research are presented, and recommendations are made for design calculations.

1. Introduction

Ammonia was among the first refrigerants to be used in refrigeration systems. Its use declined with the development of halocarbon refrigerants in the 1930s. Its use has increased drastically in recent years due to environmental damage concerns about halocarbon refrigerants, while ammonia does not have such adverse effects. It is therefore important to be able to correctly predict heat transfer during its condensation to ensure the optimum design of condensers. Condensers for ammonia include those with condensation on the outer surface of tubes such as shell and tube condensers as well as those with condensation inside channels such as atmospheric condensers and double-pipe condensers. Surprisingly, there are no well-verified published methods to determine heat transfer during the condensation of ammonia in any type of condenser. Especially in question is the effect of oil on heat transfer. Some authors consider oil to have a profound effect, such as Mazukewitch [1]. Others assert that it has no effect, such as Abdulmanov and Mirmov [2]. This research was undertaken to try to identify reliable methods for the design of condensers involving condensation inside tubes and annuli. To do so, the literature on this subject was reviewed, and test data were collected and compared to leading general correlations. The results of this research are presented and discussed. Recommendations are made for application to design.

2. Previous Work

2.1. Experimental Work

Early work on ammonia condensation has been reviewed by McAdams [3] and Hoffman [4]. More recent works and the effect of oil on heat transfer are reviewed in Shah [5].

Kratz et al. [6] performed tests on a horizontal double-pipe condenser consisting of twelve double pipes, each 6.1 m long, arranged in a vertical column and connected in series by U-bends. The outer diameter of the outer tube was 50.8 mm (2 inch) and the outer diameter of the inner tube was 31.75 mm (1.25 inch). Ammonia flowed in the annular space, while cooling water flowed in the inner tube in counterflow to ammonia. Ammonia vapor was supplied from a typical refrigeration compressor. Hence, ammonia would have contained lubricating oil; it would have been immiscible as only immiscible oils were available at that time. Overall heat transfer coefficients were calculated from the measurements for each set of two adjacent tubes and were given in tables. Kratz et al. [6] also tested an atmospheric condenser and a vertical shell and tube condenser.

Mazukewitch [1] condensed ammonia on vertical tubes 16 mm in OD, surrounded by a glass tube 46 mm in diameter. Using a tube with a polished surface, he found heat transfer coefficients to be close to those predicted by Nusselt’s theory. On a tube with a rusty surface, the heat transfer coefficient was about 25% lower. When the rusty tube was smeared with oil, the heat transfer coefficient dropped further to about half of that for the polished tube. Information about the source of ammonia vapor was not provided; it appears that the ammonia vapor entering the test section was oil-free.

Tests with oil-free ammonia have been reported by Vollrath et al. [7] and Komandiwirya et al. [8]. These were carried out at the same test facility, and two of the authors were involved in both reports. Hence, these should be regarded as a single source. In these tests, ammonia was condensed in horizontal tubes. These authors compared their test data to a number of correlations and reported poor agreement. Cavallini et al. [9] compared their general correlation to the data of Vollrath et al. [7] and reported that its predictions were 2 to 3 times the measured heat transfer coefficients. It may be noted that the data for the saturated boiling of ammonia from Zurcher et al. [10] and the subcooled boiling of ammonia from Noel [11] in tubes are in excellent agreement with Shah’s [12,13] general correlations, respectively; these correlations agree well with data for many fluids over very wide ranges. The mechanism of heat transfer during condensation is similar to that during evaporation without bubble nucleation. Therefore, one will not expect the heat transfer of condensing ammonia to be different from that of other fluids. Therefore, these data appear to be very inaccurate and hence unreliable.

Fronk and Garimella [14] condensed oil-free ammonia in horizontal tubes with 0.98, 1.44, and 2.16 mm diameters. These data were compared by Shah [15] with his general correlation as well as other correlations. The MAD (mean absolute deviation) of the Shah correlation was 34.2% and the AD (average deviation) −24.2%. Other correlations also predicted low AD, ranging from −16.2% to −54.1%. According to these authors, the maximum uncertainty in the reported heat transfer coefficients is ±36%. If the reported heat transfer coefficients are reduced by 25%, all correlations will be in adequate agreement with them. These data may therefore be considered fairly accurate, though with a large margin of uncertainty.

Most recently, Ruzaikin et al. [16] reported tests on the condensation of oil-free ammonia in 8 mm and 11 mm tubes in horizontal and vertical downflow orientations, as well as other orientations. They report good agreement with the Shah [17] and Dobson and Chato’s [18] correlations. They have also given a correlation of their own data. The agreement of these data with well-verified general correlations suggest that they are accurate.

The range of analyzable test data that are considered reliable is given in

Table 1. Range of available test data.

Source	Test Section	DHYD (DHP) mm	Flow Direction	Oil Content	TSAT, °C	G, kg/m2s	x	Note
Kratz et al. [6]	Annulus	12.2 (29.2)	H	Immiscible oil	27	27.5	0.40 0.81 (mean)	Overall heat transfer coefficients reported.
Fronk and Garimella [14]	Tube	0.98	H	None	40–50	75–100	0.20–0.71	Local heat transfer coefficients reported.
		1.44			30–60	75–225	0.2–0.82
		2.16			50–60	75–100	0.20–0.86
Ruzaikin et al. [16]	Tube	8	H, VD, VU	None	35–65	40–160	0.08–0.80	Local heat transfer coefficients reported.
		11	H, VD, VU		35–65	20–120	0.10–0.80
		8, 11	Inclined 15° to 82° downward		55	80–120	0.15–0.80

.

2.2. Methods for Predicting Heat Transfer

2.2.1. General Correlations for Horizontal and Vertical Downflow

Many correlations have been published, which are stated to be applicable to both horizonal and vertical downflow. Among these are the various versions of the correlation given by Shah, from Shah [19] to Shah [15]. According to Shah’s correlations, heat transfer in the two orientations can be different, especially at lower flow rates. It gives different formulas for the two orientations. All other correlations consider heat transfer to be unaffected by flow directions and predict the same heat transfer for horizontal and vertical downflow. Among the other correlations that have been verified with extensive databases are Cavallini et al. [9], Moser et al. [20], Moradkhani et al. [21], Hosseini et al. [22], Marinheiro et al. [23], Nie et al. [24], Dorao and Fernandino [25], and Kim and Mudawar [26]. The correlations of Akers et al. [27] and Ananiev et al. [28] were based on limited data but have been widely cited and compared to test data by many researchers.

2.2.2. General Correlations for Inclined Tubes

There have been comparatively few correlations for inclined tubes. The ones verified with the most data The ones verified with most data are the Shah [29] correlation and its improved version Shah [30]. These modify the predictions of the Shah correlations for horizontal and vertical downflow.

Correlations have been presented by Mohseni et al. [31], Yang et al. [32], Xing et al. [33], and Moghadam et al. [34]. These were based entirely on their own test data and were not verified with data from other sources. Adelaja et al. [35] presented a correlation which showed good agreement with data from four sources.

The correlation of Mohseni et al. [31] requires Δx/L to be inserted, where Δx is the change in quality in tube length L. This information is not available for the present database. It was therefore not possible to evaluate this correlation. The correlation of Adelaja et al. [35] requires the insertion of ΔT = (T_SAT − T_w). This is unknown, and hence, iterative calculations with assumed ΔT are required. This can be carried out only if heat flux is known. However, heat flux is not known for these data. Therefore, this correlation could not be evaluated.

The correlations of Yang et al. [32], Xing et al. [33], and Moghadam et al. [34] have given correlations based entirely on their own data and they do not involve the insertion of any unknowns. The correlation of Xing et al. requires the insertion of heat transfer coefficients in horizontal and vertical downward flow but does not provide a method to predict them. Hence, it cannot be used for analyzing data from other sources.

2.2.3. Correlations Specifically for Ammonia

Ruzaikin et al. [16] have given a correlation which gives good agreement with their own data from their tests. They have not compared it with any other data for any fluid.

3. Data Analysis

3.1. Oil-Free Ammonia Data

3.1.1. Horizontal and Vertical Downflow Data

Data for oil-free ammonia for horizontal and vertical downflow were compared to the general correlations of Shah [15], Kim and Mudawar [26], Ananiev et al. [28], Hosseini et al. [22], Moradkhani et al. [21], Moser et al. [20], Akers et al. [27], Marinheiro et al. [23], and Nie et al. [24]. Properties of ammonia were obtained from REFPROP 9.1, Lemmon et al. [36]. The deviations of various correlations for the data considered reliable are listed in

Table 2. Results of comparison of data for oil-free ammonia in horizontal and vertical downflow with various correlations.

Source	D, mm	Flow Direction	pr	G, kg/m2s	x	N	Deviations % of Various Correlations, MAD (Upper Row), AD (Lower Row)
							Shah [15]	Kim and Mudawar [26]	Ananiev et al. [28]	Dorao and Fernandino [25]	Hosseini et al. [22]	Moradkhani et al. [21]	Moser et al. [20]	Traviss et al. [37]	Akers et al. [27]	Marinheiro et al. [23]	Nie et al. [24]
Fronk and Garimella [14]	0.98	H	0.1367 0.1788	75 100	0.20 0.71	18	43.8 −43.8	49.7 −49.7	59.7 −59.7	62.7 −62.7	72.8 62.5	27.4 −27.1	62.5 −62.5	34.2 34.2	36.0 36.0	58.5 −58.5	99.3−99.3
	1.44		0.1025 0.2300	75 225	0.2 0.82	49	29.5 −29.5	28.6 −28.6	37.1 −37.1	43.2 −43.2	54.0 52.6	21.1 −20.7	45.3 −45.3	20.0 20.0	47.4 47.4	40.7 −40.7	98.9−98.9
	2.16		0.1788 0.2300	75 100	0.20 0.86	12	39.1 28.0	23.1 11.7	31.5 −11.0	24.2 −2.5	28.8 28.7	32.8 18.7	30.0 −18.3	57.8 57.8	109.7 109.7	27.9 2.4	95.5−95.5
Ruzaikin et al. [16]	8	H	0.1187 0.2593	40 160	0.12 0.81	116	10.3 5.4	20.6 −3.6	20.6 −12.4	19.7 18.2	15.0 15.0	15.7 −7.1	24.6 −23.4	71.7 71.7	16.9 16.9	13.4 −2.7	67.9−67.9
		VD	0.203	40 160	0.13 0.79	25	14.7 7.8	25.6 17.9	13.3 0.4	40.5 39.5	16.2 15.6	14.9 8.0	15.9 −11.2	114.4 114.4	36.5 36.5	17.2 14.1	59.7−59.7
	11	H	0.1187 0.2593	20 120	0.07 0.79	101	13.3 −12.2	43.6 −13.3	39.5 −39.5	10.2 0.4	16.9 16.7	27.7 −27.6	45.7 −45.7	37.7 37.7	20.3 20.3	26.4 −26.4	55.0−55.0
		VD	0.203	40 120	0.05 0.79	38	18.0 −6.9	26.1 8.5	20.2 −13.3	41.9 40.6	12.9 13.4	13.5 2.8	23.0 −21.7	78.4 78.4	14.6 14.6	18.0 5.7	42.6−42,6
All sources	0.98 11.0		0.1025 0.2593	20 225	0.05 0.86	359	17.5 −7.2	30.6 −8.8	30.0 −24.9	26.3 4.0	24.1 −11.4	20.7 −12.8	34.7 −33.7	56.4 44.4	27.2 13.1	24.3 −15.1	67.7−67.7

. These deviations are defined as below.

The mean absolute deviation (MAD) is defined as follows:

$$MAD={\displaystyle \frac{1}{N}}{\sum }_1^{N}ABS\left\{\left(h_{predicted}-h_{measured}\right)/h_{measured}\right\}$$

(1)

The average deviation (AD) is defined as

$$AD={\displaystyle \frac{1}{N}}{\sum }_1^{\mathrm{N}}\left\{\left({\mathrm{h}}_{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}}-{\mathrm{h}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{e}\mathrm{d}}\right)/{\mathrm{h}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{e}\mathrm{d}}\right\}$$

(2)

Considering all data, the Shah [15] correlation has the lowest MAD of 17.1%. Other correlations also give reasonable deviations, except the correlations of Akers et al. [27], Traviss et al. [37], and Nie et al. [24]. These three were also found to be erratic by Shah [30] upon comparison with data for many fluids over a wide range of parameters. Hence, these are ignored in further discussions. The agreement of the other correlations is fair to good with the data of Ruzaikin et al. [16]. The deviations are larger with the data of Fronk and Garimella [14]. As noted in Section 2.1, these authors stated that there was large uncertainty in their data.

Figure 1, Figure 2, Figure 3 and Figure 4 show the comparison of some data of Ruzaikin et al. [16] for horizontal and vertical downflow. Figure 5 compares the data of Fronk and Garimella [14] for a horizontal tube with some correlations.

The data of Vollrath et al. [7] and Komandiwirya et al. [8] were also analyzed. Some of the data were in reasonable agreement with the more reliable correlations, while some were very low. Figure 6 shows the results for some of the data, which are very low. It is seen that the predictions of all correlations are much higher than the measured heat transfer coefficient. It is clear that the data from these two studies are very erratic and no conclusions can be drawn from them. These are therefore not included in Table 2 and are not discussed any further.

3.1.2. Inclined Tubes

Besides the test in horizontal and vertical downflow, Ruzaikin et al. [16] also carried out tests with vertical upflow as well as with downward flow at inclinations of −82 to −6 degree. These were compared to the correlations of Shah [30], Moghadam et al. [34], and Yang et al. [32]. The results are listed in

Table 3. Results of comparison of data for inclined tubes with various correlations.

Source	D, mm	θ°	pr	G, kg/m2s	x	N	Deviations % of Various Correlations, MAD (Upper Row), AD (Lower Row)
							Shah [30]	Yang et al. [32]	Moghadam et al. [34]
Ruzaikin et al. [16]	8	+90	0.2032	40 160	0.08 0.76	70	20.1 18.1	82.0 −82.0	19.3 −18.1
		−82 −6	0.2032	80 120	0.10 0.78	92	10.9 5.3	83.2 −83.2	29.8 −6.4
	11	+90	0.2032	40 120	0.07 0.78	72	18.9 9.5	82.2 −82.2	35.9 −35.9
All data	8.0 11.0	−82 to −6 and +90	0.202	40 160	0.07 0.78	234	16.1 10/4	82.6 −82.6	28.5 −19.0

. The Shah correlation is seen to give good agreement, with an MAD of 16.1%. The correlation of Moghadam et al. [34] is also in fairly good agreement, with an MAD of 28.5%. The Yang et al. [32] correlation predicts very poorly.

Figure 7 and Figure 8 show the comparison of the three correlations with data for vertical upflow for the 8 mm and 11 mm diameter tubes. Figure 9 and Figure 10 show the comparison of heat transfer coefficients during downward inclined flow with the correlations of Shah and Moghadam et al. [34]. The Shah correlation is seen to be in good agreement with the data, while that of Moghadam et al. [34] has higher deviations but not too high.

3.2. Ammonia with Oil

As described in Section 2.1, Kratz et al. [6] have provided data for condensation in a horizontal double-pipe condenser, which received ammonia vapor discharged from a compressor lubricated with immiscible oil. Ammonia condensed in the annular space, while cooling water flowed inside the inner tube. They have provided overall heat transfer coefficients. To determine the heat transfer coefficient of ammonia, the following equation was used:

$${\displaystyle \frac{1}{U}}={\displaystyle \frac{1}{h_{TP}}}+{\displaystyle \frac{A_o\delta }{{A_{m}k}_{tube}}}+{\displaystyle \frac{A_o}{A_i{h}_{water}}}$$

(3)

U is the overall heat transfer coefficient, A_o is the outside area of the inner tube, A_i is the inside area of the inner tube, A_m = (A_o + A_i)/2, and k_tube is the thermal conductivity of tube material. h_water is the heat transfer coefficient of water flowing inside the inner tube. It was calculated by the following equation given by McAdams [16]:

$$h=0.023{Re}_{LT}^{0.8}{Pr}_L^{0.4}{k}_L/D$$

(4)

The material and thickness of the tubes were not stated by Kratz et al. [6]. The tubes were assumed to be of carbon steel as is usual and of 12 BWG (Birmingham Wire Gauge) as these appeared to be thick enough for the pressure. This gives the inside diameter of the outer tube to be 31.8 mm and that of the inside tube as 26.2 mm. The thermal conductivity of the tubes was taken to be 45 W/m K, which is typical for carbon steel. Only the data for the upper four rows of tubes were analyzed as the ammonia to water temperature differences in the lower tubes were too small to allow an accurate calculation of ammonia side heat transfer coefficient.

Calculations were also carried out assuming the tube to be 10 BWG. The results were close to those for the 12 BWG tubes.

These data were compared to all the correlations that were used to evaluate the oil-free data. The Shah correlation requires the use of D_HP in calculating the single-phase heat transfer coefficient and use of D_HYD for all other parameters. The same was also carried out with other correlations, except those of Kim and Mudawar [26], Dorao and Fernandino [25], Hosseini et al. [22], and Moradkhani et al. [21]. For these correlations, D_HYD was used as the diameter in all calculations because that was specified by these authors. The definitions of the two equivalent diameters are below.

$$D_{HP}={\displaystyle \frac{4 x Flow area}{Perimeter with heat transfer}}$$

(5)

$$D_{HYD}={\displaystyle \frac{4 x Flow area}{Wetted Perimeter}}$$

(6)

All of the more reliable correlations were found to grossly overpredict these data. Figure 11 shows the results with some of the correlations. This discrepancy appears to be due to the formation of oil films on the tube surface, which reduce heat transfer due to their low thermal conductivity. This is discussed in Section 4.2.

4. Discussion

4.1. Oil-Free Ammonia

As seen in Table 2, the data of Ruzainik et al. for horizontal and vertical downflow is in close agreement with the Shah correlation, its MAD being only 17.2%. The agreement is also fairly good with other general correlations. For example, the MAD of the Moradkhani et al. [21] correlation is 20.7% and that of the Dorao and Fernandino [25] correlation is 26.1%. The MADs of other general correlations are within this range. These data include diameters of 8 and 11 mm and a wide range of pressures and flow rates. The deviations are higher with the data of Fronk and Garimella [14], but this can be attributed to the large (±36%) uncertainty in these data reported by these authors. If the actual heat transfer coefficients were 20% lower than those reported, they would be in adequate agreement with most correlations.

The data of Ruzaikin et al. [16] for inclined tubes are in close agreement with the Shah [30] correlation. Their agreement with the correlation of Moghadam et al. [34] is also fairly good. These results indicate that well-verified general correlations are applicable to oil-free ammonia.

4.2. Ammonia Containing Immiscible Oil

As seen in Section 3.2, the data of Kratz et al. [6] for oil-containing ammonia are much lower than the predictions of the Shah correlation, as well as other general correlations. The reason for this could be the insulating effect of oil films as discussed below.

As was mentioned in Section 2.1, Mazukewitch [1] found that the heat transfer coefficient from the outer surface of a rusty tube smeared with oil was only about half of that with a clean polished tube. Abdulmanov and Mirmov [2] and Mirmov and Yemelyanov [38] report that design calculations in the USSR for ammonia condensers usually assume oil films to be 0.05 to 0.08 mm thick. However, these authors assert that the low heat transfer coefficients present in ammonia condensers are low because of the presence of air, not because of oil films. As seen in Section 4.3, the data of ammonia with miscible oil are in satisfactory agreement with general correlations. The presence of air drastically reduces heat transfer, irrespective of which fluid is being condensed, with or without oil. If air is the cause of low heat transfer, one may question why air enters the condenser when ammonia with immiscible oil is being condensed but does not enter when pure ammonia or ammonia with miscible oil is being condensed. Further, why are low heat transfer coefficients not present during the condensation of pure refrigerants such as R-12 and R-134 and other halocarbon refrigerants? If air can enter during the condensation of ammonia, it should also be able to enter during the condensation of other fluids. Pressure is above atmospheric during the condensation of ammonia and during condensation of halocarbon refrigerants. Data for the condensation of halocarbon refrigerants from numerous studies have been shown to be in good agreement with the general correlations such as those of Shah [15] and Cavallini et al. [9]. It is therefore clear that the reduction in heat transfer during the condensation of immiscible oil-containing ammonia is caused by the resistance of oil films not due to the presence of air.

Some insight into the effect of immiscible oil may be gained from experience with tests on ammonia evaporators. Shah [39] performed tests on a 26.2 mm diameter evaporator, which included visual observations. He observed oil films during single-phase flow as well as two-phase flow. The measured heat transfer coefficients during liquid flow were found to be much lower than those predicted by Equation (4). He attributed it to the resistance of oil films. He calculated the oil film thickness using the following equation.

$${\displaystyle \frac{1}{h_{measured}}}={\displaystyle \frac{1}{h_{LT}}}+{\displaystyle \frac{\delta }{k_{oil}}}$$

(7)

where δ is the thickness of oil film and h_LT is the heat transfer coefficient predicted by Equation (4). He correlated the calculated oil film thickness by the following equation:

$${\displaystyle \frac{\delta }{D}}={\displaystyle \frac{0.028}{{Re}_{LT}^{0.23}}}$$

(8)

Chaddock and Buzzard [40] performed tests on a horizontal evaporator that was 13.4 mm in diameter. Tests were carried out with pure ammonia as well as immiscible oil-containing ammonia. They also observed oil films. The heat transfer coefficients of oil-containing ammonia were much lower than those of pure ammonia.

Boyman et al. [41] studied the heat transfer of ammonia flowing in a 14 mm diameter horizontal tube. Test were carried out with 0 to 3% immiscible mineral oil. They found that 0.1% oil reduced average heat transfer coefficients by up to 50% compared to oil-free ammonia. Further reductions occurred as oil concentration increased up to 1%.

The thickness of oil film δ, which will reconcile the heat transfer coefficients measured by Kratz et al. [6] with the predictions of the Shah [15] correlation, were calculated with the equation below.

$${\displaystyle \frac{1}{h_{measured}}}={\displaystyle \frac{1}{h_{Shah}}}+{\displaystyle \frac{\delta }{k_{oil}}}$$

(9)

The thermal conductivity of oil k_oil used by Kratz et al. [6] is not known. It was taken to be 0.13 W/m K, the same as that of the oil used in the tests of Shah [39].

The calculated oil film thicknesses are shown in Figure 12. These are seen to vary from 0.015 to 0.05 mm, decreasing with increasing quality. An increase in quality increases the velocity and hence the shear force on the tube wall. This will cause the thinning of the oil film with increasing quality. The calculated oil film thicknesses are mostly lower than the 0.05 to 0.08 mm used in the USSR. This may be because these seem to be for shell and tube condensers, and in them, vapor velocities are likely to be lower.

Considering all the information in this section, it may be concluded that when immiscible oil is present in ammonia vapor, oil films are formed on the tube surface, which cause a decrease in the heat transfer coefficient.

4.3. Ammonia with Miscible Oil

Komandiviriya et al. [8] performed tests with pure ammonia as well as tests with up to 5% oil. As was stated in Section 2.1, their data for pure ammonia were very low and were therefore not analyzed. Nevertheless, a comparison of their data for pure and oil-containing ammonia is of interest. Upon examining their data, these authors concluded that “oil does not affect the heat transfer coefficient at low mass flux and low quality. For ammonia oil mixtures, the heat transfer coefficient decreases at high mass fluxes and quality due to oil effect. In general, as oil concentration increases, the heat transfer coefficient decreases”. To the present author, it appears from their data that a significant lowering of heat transfer occurs only at high mass fluxes and qualities above 0.7. The other data do not show any clear trend. Overall, the behavior of ammonia–miscible oil mixtures is similar to that of mixtures of other refrigerants with miscible oil.

It is also interesting to study the results for evaporators. Gao et al. [42] studied the boiling of ammonia mixed with miscible oil. Tests were carried out in an 8 mm diameter horizontal tube with an oil concentration from 0 to 5.78%, mass flux from 51 to 99.5 kg/m²s, heat fluxes of 9 and 21 kW/m², and temperature from −5.5 to −5 °C. Heat transfer coefficients were based on the bubble point temperature and were found to decrease with increasing oil content at all qualities. The decrease was generally in the range of 5 to 20%.

There have been many experimental studies on the evaporation of miscible oil-containing halocarbon refrigerants. For up to 1% oil content, the heat transfer coefficient differs from pure refrigerant value by ±10%. Most studies reported a decrease in the heat transfer coefficient with 5% oil content. Hence, the results of Gao et al. [42] for ammonia with miscible oil are similar to those of other refrigerants with miscible oils.

The literature on the effect of oil on boiling and condensation heat transfer has been reviewed in Shah [5]. These include experimental studies and proposed methods for the calculation of the effect of oil on heat transfer.

4.4. Recommendations for Design

• For pure ammonia, heat transfer coefficients can be calculated with general correlations applicable to all fluids. The most verified among them is Shah [15] for horizontal and vertical flow and Shah [30] for other inclinations.
• For ammonia containing miscible oil, heat transfer can be calculated using the same methods as for other fluids. As the amount of oil in normally operating systems is small, assumption of 5–10% reduction in the heat transfer coefficient due to oil is suggested.
• For ammonia containing immiscible oil, heat transfer coefficients calculated for pure ammonia should be reduced by 50 to 60 percent in a normally operating system. Poorly operating systems may have large amounts of oil, and then heat transfer coefficients could be even lower.

5. Conclusions

• The literature on the condensation of ammonia inside tubes and annuli was reviewed to identify the available data and reliable prediction methods.
• Test data for pure ammonia were compared to general correlations for condensation heat transfer. Satisfactory agreement was found. It was concluded that the heat transfer coefficient of pure ammonia can be calculated by reliable general correlations applicable to all fluids.
• The effect of oil on heat transfer was investigated. It was concluded that for ammonia containing miscible oil, its effect on heat transfer can be calculated by the same methods as for other fluids. For normally operating systems, a 5 to 10% reduction in heat transfer due to the effect of oil is recommended.
• For ammonia containing immiscible oil, heat transfer coefficients are much lower than those of pure ammonia due to the formation of insulating oil films on the tube surface. These usually reduce heat transfer coefficients by 50 to 60%.