# Analysis of heat and mass transfers in two-phase flow by coupling optical diagnostic techniques

## Authors

- First Online:

- Received:
- Revised:
- Accepted:

DOI: 10.1007/s00348-008-0519-z

- Cite this article as:
- Lemaitre, P. & Porcheron, E. Exp Fluids (2008) 45: 187. doi:10.1007/s00348-008-0519-z

- 6 Citations
- 173 Views

## Abstract

During the course of a hypothetical accident in a nuclear power plant, spraying might be actuated to reduce static pressure in the containment. To acquire a better understanding of the heat and mass transfers between a spray and the surrounding confined gas, non-intrusive optical measurements have to be carried out simultaneously on both phases. The coupling of global rainbow refractometry with out-of-focus imaging and spontaneous Raman scattering spectroscopy allows us to calculate the local Spalding parameter *B*_{M}, which is useful in describing heat transfer associated with two-phase flow.

### List of symbols

### Latin symbols

*B*_{M}Spalding parameter

*C*pheat capacity

*d*diameter

*D*diffusion coefficient

*k*conductivity

*L*latent heat

*M*molar mass

- \( \dot m \)
mass flow rate at the droplet interface

*n*number of moles

*P*TOSQAN absolute pressure

*R*_{TOSQAN}TOSQAN radius

*s*local saturation ratio

- \( \overline S \)
average saturation ratio

*t*Time

- \( \overline T _{{}}^{{_{{}}^{{}} }} \)
mean temperature

*T*temperature

*V*_{sump}sump volume

*V*_{TOSQAN}TOSQAN volume

- \( \bar V \)
mean radial velocity of gas

*X*volume fraction

*Y*mass fraction

### Greek symbols

*τ*_{phase}phase-change characteristic time

*τ*_{entr}entrainment characteristic time

*τ*_{conv}convective heat transfer characteristic time

*ρ*density

### List of subscripts

- air
relative to dry air

- drop
relative to the droplet

- int
relative to the droplet interface

- gas
relative to the gas mixture

- sat
relative to the saturation state

- steam
relative to steam

- sump
relative to the sump zone

- TOSQAN
relative to the TOSQAN entire volume

- Water
relative to water in liquid phase

### List of dimensionless numbers

*Sc*Schmidt number

*Sh*Sherwood number

*Nu*Nusselt number

*Pr*Prandlt number

*Re*Reynolds number

### List of acronyms

- GRR
global rainbow refractometry

- MS
mass spectroscopy

- PDA
phase Doppler anemometry

- PIV
particle imaging velocimetry

- SRSS
spontaneous Raman scattering spectroscopy

## 1 Introduction

The aim of this article is to study the heat and mass exchanges between spray droplets and the surrounding gas with thermal–hydraulic conditions representative of a hypothetical severe accident in a nuclear pressurized water reactor (PWR). In fact, a breach in the primary cooling circuit during a hypothetical severe accident in a nuclear power plant could lead to pressurization of the containment with steam. To maintain the integrity of the containment enclosure, spraying systems are used to reduce pressure and temperature, by means of steam condensation on the droplets.

There is an abundant literature of different models for taking account of the heat and mass transfers in multiphase flows; Sirignano (1993) drew up a comprehensive review of modelling approaches to provide an interfacial transfer expression that could be directly implemented in transport equations. Unfortunately, there is a lack of accurate measurements to validate these models for PWR applications. In fact, these models are established and validated in the fields of aerospace and combustion, and thus applied to the case of droplet vaporization (Ranz and Marshall 1952; Spalding 1952; Abramzon and Sirignano 1987). Thus, we still require a validation in the symmetrical case of steam condensation on droplets.

Moreover, the emergence of optical diagnostics over the last twenty years has enabled increasingly accurate non-intrusive measurements of parameters such as droplet temperature (van Beeck et al. 1999) and size (Glover et al. 1995) as well as gas composition (Cohen-Tanoundji et al. 1986); these measurements give a higher level of qualification of the previously established models.

Because of the typical size of containment buildings (60,000 m^{3}) experiments are often designed with scale different from prototypical. A variety of similarity and scaling methods is available to enable transferring such experimental results to real reactor conditions. The strategy adopted to study severe accidents is thus to perform experiments at different scales and to increase the level of instrumentation and its reliability with the decrease of the experimental scale.

In view of this state of the art, the TOSQAN experiment was created to simulate typical accidental thermal–hydraulic flow conditions in reactor containment. TOSQAN is a French acronym that means TONUS analytical qualification; TONUS is a CFD code developed by CEA and IRSN and dedicated to simulate severe accident in nuclear power plants. This experiment is small compared to containment buildings, but it is large compared to its level of instrumentation. A special scaling effort has been achieved to design the test presented in this article, in order to be representative to spray release during a severe accident (Fischer et al. 2002).

In the first part of this article, we describe the TOSQAN experiment and the associated optical instrumentation. The second part presents the implemented test and its associated measurements. Finally, we analyze the heat and mass exchanges on both phases while giving a detailed description.

## 2 Description of the facility and its instrumentation

### 2.1 TOSQAN facility

A single nozzle is used to spray injection into the upper part of the vessel. A special system was designed to allow vertical movements over a height interval of 90 cm, so as to carry out measurements at different distances from the nozzle but through the same porthole window. In addition, the nozzle is off-centred by 40 cm in relation to the axis of the enclosure.

The gas temperature and pressure are measured using a set of more than 150 thermocouples and two pressure sensors. Deflectors are fitted above the thermocouples to prevent the impact of droplets on the sensor, thus ensuring that the measurements are only related to the gas temperature.

For this particular test, the nozzle is off-centred by 40 cm so the different measurements—particularly the droplet temperature—can be carried out in the axis of the nozzle.

The optical measurements are carried out through 14 porthole viewing windows distributed at four levels. Each porthole window is composed of a double-glass screen, with hot air circulation in between to avoid steam condensation. To characterize the two-phase flow, we rely on two kinds of instrumentation.

On the one hand, we use conventional devices such as thermocouples, pressure sensors and flow meters, as well as commercially available techniques such as out-of-focus imaging for particle sizing (Lemaitre et al. 2006a, b) and particle imaging velocimetry (Porcheron et al. (2003)) to measure the droplet velocity field. In addition, sampling is performed to determine the gas composition by mass spectrometry. On the other hand, we developed two optical benches, dedicated for non-intrusive measurement of the local gas volume fraction (spontaneous Raman scattering spectroscopy) and droplet temperature (global rainbow refractometry). The principles of these two techniques are briefly presented below.

### 2.2 Global rainbow refractometry (GRR)

This technique was introduced by van Beeck et al. (1999) to measure the mean droplet temperature on a set of droplets located in a small volume of measurement. This optical diagnostic tool is based on the analysis of oscillations generated by the interference between the optical rays refracted by the droplets. This allows a determination of the refractive index, and, hence, the temperature of the droplets. However, this technique is hindered by problems linked to droplet non-sphericity that can produce a shift in the temperature measurements (van Beeck and Riethmuller 1995).

A special calibration has been established by Lemaitre et al. (2006a, b) to avoid these disturbances. This procedure is implemented in the TOSQAN experiment after each test with GRR measurements; it consists in injecting a spray at the same mass flow rate as during the test (30 g s^{−1}), and at exactly the same temperature as the gas when initially saturated with steam. Under these conditions, there is neither heat nor mass transfer between the droplets and the gas mixture; as a consequence, the average measured droplet temperature is equal to its injection temperature as well as the gas temperature. This allows us to correct the global rainbow temperature measurement and take into account the angular shift induced by the non-sphericity of the droplets.

Once the calibration is complete, we then need to validate the measurement, using exactly the same procedure as the calibration. A spray is injected into a gas mixture having the same temperature and saturated with steam, to check whether that the GRR measurement is close to its injection temperature. This method is reproduced at different temperatures close to the value expected to be measured in the test considered here.

As the difference between the GRR measurement and the gas temperature never exceeds 1°C, this method allows us to validate the GRR measurement obtained on the TOSQAN experiment (Lemaitre et al. 2006a, b).

### 2.3 Spontaneous Raman scattering spectroscopy (SRSS)

This non-intrusive technique was developed on the TOSQAN facility to characterize the gas composition. Spontaneous Raman scattering is an inelastic scattering process involving the interaction of a photon with a specific vibration-rotational state of a molecule. This scattering results from a process in which energy exchange occurs between the incident photons with frequency (υ_{0}) and a given molecule. The generated energy flux undergoes a frequency shift (±υ) that is characteristic of each polyatomic molecule depending on the occurrence of a Stokes (+υ) or anti-Stokes (−υ) transition. A full description of the Raman effect is presented by Cohen-Tanoundji et al. (1986).

Thus, if all the constituents of the gas mixture are active in Raman scattering, then the molar fractions of each constituent can be calculated using the Raman power scattered by each molecule. Porcheron et al. (2003) have presented a full description of the principle of this technique, along with the optical set up and its validation on the TOSQAN facility in the presence of droplets.

### 2.4 Evaluation of measurement uncertainties

Instrumental accuracy in the TOSQAN experiment

Measurement technique | Uncertainty | |
---|---|---|

Pressure | Pressure sensor | 0.01 bar |

Droplet temperature | GRR | 1°C |

Droplet and gas velocity | PIV | <5% |

Gas temperature | Thermocouples | 1°C |

Droplet size | Out-of-focus imaging | <5% |

Gas composition | SRSS | <2% |

## 3 Description of the test

Prior to the beginning of the actual test, while the gas in the TOSQAN enclosure is composed of dry air at 1 bar, the walls of the experiment are regulated at 120°C by circulating coolant fluid inside the double wall.

Then, the vessel is pressurized with 1.5 bar of dry steam, added to the 1 bar of dry air, by means of a steam injection pipe (Fig. 1).

The steam injection ceases when the pressure reaches 2.5 bar, whereas sprinkling starts with a water flow rate of 30 g s^{−1} at a temperature of 23°C. For this test, the spray injection is off-centred by 0.4 m. During the test, the water is drained accordingly to prevent water accumulation in the sump. To reproduce exactly the same initial conditions, dry air circulation is imposed before each test inside the vessel for 20 h, at a mass flow rate of 5 g s^{−1}, which represents a ventilation rate of two air changes per hour, and consequently a total of 40 air changes before a new test.

The spray is generated with a TG-3.5 nozzle (spraying systems). This starts a transient state of depressurization in the vessel, which is continued until an overall steady state is obtained where the pressure and the average temperature of the gas mixture present in the vessel remain stable. Throughout the test, the temperature of the vessel walls is regulated at 120°C, owing to the circulation of coolant fluid inside the double wall.

To perform the measurements for characterization of both phases (droplet size, velocity and temperature, as well as gas volume fraction), this test must be reproduced identically several times.

The test analysis consists of two parts. First, we carry out a global analysis of the heat and mass transfers, followed by a more detailed and local study of the exchanges between the droplets and the gas.

## 4 Heat and mass transfer analysis

### 4.1 Global analysis

For a more in-depth description of the test, we require thermodynamic parameters that fully describe the time course of the test. These parameters are the total number of moles inside the vessel *n*_{Total}, the average saturation ratio \( \overline S (t) \) and the average gas temperature \( \overline T _{{{\text{TOSQAN}}}} (t). \)

The total number of moles inside the vessel is computed according to equation (1) i.e., assuming an ideal gas.

^{−1}, the experimental enclosure is 4.8 m high, and the droplet velocity is more than 1 m s

^{−1}. As a consequence, the residence time of the droplets in the gas is less than 5 s, leading to a liquid volume in suspension smaller than 1.5 × 10

^{−4}m

^{3}, which is negligible compared to the entire TOSQAN volume (7 m

^{3})

*t*, unless otherwise specified)

Basing our analysis on the evolution of these three parameters, we can identify three characteristic times (*t*_{1}, *t*_{2} and *t*_{3}) from the test result.

For times after *t*_{2}, the uncertainty on the saturation ratio is smaller than 0.03, which represents 3% of the average saturation ratio. As a consequence, in cases where the values of average saturation ratio \( \left( {\overline S } \right) \) are sometimes slightly higher than unity (Fig. 5), this is attributed to the measurement uncertainty.

Throughout this study, the uncertainties on the different calculated parameters are evaluated using the guide to the expression of uncertainty in measurement (ISO 1993).

Therefore, we observe that, until time *t*_{1}, the total number of moles increases in the containment, which is correlated with a sharp decrease in temperature. Nevertheless, as the leakflow rate is negligible, the change in total number of moles corresponds to variations in the number of moles of vapor in the containment. The second characteristic time *t*_{2} corresponds to the time when the gas becomes totally saturated inside the vessel. Finally, time *t*_{3} corresponds to the time at which all the parameters of our analysis become stable.

These three characteristic times lead us to subdivide the test into four phases denoted A, B, C and D in Fig. 5.

Phase A of the test is a very short (100 s), corresponding to a strong droplet vaporization regime. This first phase is due to a low initial saturation ratio and an initial mass stratification, particularly in the bottom of the vessel (sump, Fig. 1). This last point is discussed in the section treating the local analysis of heat and mass transfers.

The variation in total number of moles highlights the fact that this second phase of the test (B phase) corresponds globally to condensation. The gas temperature decreases due to convective heat transfer between the gas and the droplets injected at 23°C. At the end of this phase, the atmosphere is globally saturated in the containment.

The convective heat transfer between the droplets and the gas drives this third phase (C phase). This heat transfer still cools the gas and, as a consequence, steam condenses on the droplets at a rate that keeps the vessel globally saturated.

On this graph, we observe that the total number of moles of condensed steam increases with decreasing temperature of the injected water. Hence, more steam can condense on colder droplets before the system reaches its equilibrium temperature: lowering the temperature of spray droplets is a very efficient means of condensing steam in the containment atmosphere.

Consequently, the local measurements presented in the next section correspond to tests performed with an injection temperature of 23 ± 0.5°C (uncertainty of thermocouple measurements).

### 4.2 Local measurements

Different local measurements were carried out during this test. As mentioned above, not all these measurements were obtained on the same day, but they were always carried out under exactly the same thermodynamic conditions. The results of these measurements are presented either as variations with time at a given distance from the nozzle, or as fields or vertical profiles during the equilibrium phase.

#### 4.2.1 Transient measurements

*X*

_{steam}) was recorded on the axis of the spray and at 25 cm from the nozzle orifice, using the SRSS technique (Fig. 7).

We can subdivide this time-evolution into three phases for this particular measurement. The first phase corresponds to the first 60 s of the test, and is associated with an increase in steam volume fraction. The second phase corresponds to a strong decrease of the steam volume fraction, and continues until its stabilization (third phase) after 3,500 s of spraying.

#### 4.2.2 Steady-state measurements

In this part of the paper, we present all the measurements needed for describing the test.

#### 4.2.3 Gas temperature measurement

As one of the thermocouples of the highest rod is in the axis of the nozzle orifice (Fig. 2), a gas temperature vertical profile is measured in the axis of the spray. To acquire a vertical profile of gas temperature, the nozzle is displaced during the equilibrium phase; as a consequence, each data point plotted on Fig. 9 corresponds to the temperature recorded by the thermocouple in the axis of the spray when the measurement is stabilized (600 s after the nozzle displacement). It is important to point out that displacement of the nozzle does not modify the equilibrium state; this is because the displacement (250 mm) is an order of magnitude lower than the height of the TOSQAN experiment (4,800 mm).

The gas cooling induced by the spray is observed on Fig. 8, with a minimum temperature in the axis of the nozzle. Moreover, this profile is dissymmetric with regard to the nozzle axis, because the nozzle is off-centred; as a consequence, the temperature gradient is steeper on the side towards the off-centred nozzle, due to the proximity of the heated wall.

Inputs of the simulation

Inputs of the simulation | Measurement techniques | Measurement results | Uncertainty |
---|---|---|---|

Droplet initial temperature | GRR measurement at 5 cm from the nozzle orifice | First measurement point of Fig. 10 | 1°C |

Droplet velocity vertical profile | PIV measurements | Fig. 13 | <5% |

Gas temperature vertical profile | Thermocouple measurements | Fig. 9 | 1°C |

Droplet initial size | Out-of-focus imaging measurements |
| <5% |

Gas saturation ratio | SRSS associated with thermocouple measurements | 97% | <2% |

#### 4.2.4 Droplet temperature measurement

#### 4.2.5 Droplet size measurement

The spray size distribution can be satisfactorily fitted by a log-normal distribution with a median diameter of 100 μm and a geometrical standard deviation of 1.65. The computed arithmetic mean diameter and the Sauter mean diameter are equal to 141 and 207 μm, respectively \( \left( {d_{{10}} = 141 \pm 7\,\upmu{\text{m}},d_{{32}} = 207 \pm 10\,\upmu{\text{m}}} \right). \) Measurements of droplet size at injection were performed at the same distance from the injection nozzle but at low gas temperature to prevent droplet vaporization, and with addition of air in order to reproduce the same gas density. As a consequence, we reproduced the same aerodynamic effects on the droplets. The resulting measurements match very closely with the result presented in Fig. 11.

#### 4.2.6 Measurement of droplet velocity

The droplet velocity fields are measured throughout the spray test using the PIV technique. Our PIV set up is a DANTEC commercial system composed of a double-cavity Nd YAG laser doubled in frequency to provide a laser beam at a wavelength of 532 nm. The flow is seeded with the spray droplets, whose displacement is measured between the two pulses of the laser. This displacement is measured with a Hisense camera (1280 × 1024), equipped with a 60 mm lens and an aperture of f22. Various tests were performed to determine the optimum setting for laser intensity and inter-pulse times (which vary with the distance from the nozzle due to droplet deceleration) and, finally, the interrogation size area and overlap. We chose an interrogation size of 32 × 32 for the images, without overlap. More details on the use of the PIV technique to measure droplet velocity are provided by Comer et al. (2001) and Porcheron et al. (2003) for the TOSQAN application.

We can observe a perfect continuity of the vertical component of the droplet velocity measured in the axis of the nozzle, which reflects the deceleration of the droplets as a function of the distance to the nozzle, due to the transfer of momentum to the gas. This observation is useful in the following (Table 2), for validation of the two-boundary-layer model.

### 4.3 Local analysis

*T*

_{gas},

*P*

_{steam},

*s*) are analogous to those of the global analysis \( \left( {\overline T _{{{\text{TOSQAN}}}} ,n_{{{\text{total}}}} ,\overline {S(t} )} \right), \) and thus allow us to highlight the similarities and differences between the local and global dynamics of the test.

First of all, we observe that, on the local scale, the initial droplet vaporization phase lasts only 60 s \( \left( {t_{1}^{{{\text{local}}}} = 60 \pm 3\,{\text{s}}} \right), \) whereas it corresponds to a 100 s phase on the global scale (of the whole vessel). This is due to the fact that, during the first 100 s of the test, the initial vaporization phase propagates from the near field of the nozzle to the far field. Moreover, we observe that, locally, the droplet vaporization phase ends when the gas nearly reaches saturation \( \left( {s = 0.97 \pm 0.03\quad {\text{for}}\,t > t_{1}^{{{\text{local}}}} } \right). \)

However, we observe that, globally, the first vaporization phase ends before the gas reaches overall saturation. This is because the spray does not extend throughout the entire containment volume. As a result, even if the spray zone is nearly saturated, the outer part of the spray zone is still far from saturation (but remains nearly at its initial water vapor content). Thus, we infer that the phase between times *t*_{1} and *t*_{2} corresponds to a phase of steam transfer from the spray zone to the non-spray zone.

^{−1}. Thus, using equation (4), we can derive the entrainment characteristic time (

*τ*

_{entr}) from this measurement

*τ*

_{phase}) is computed in the following from equation (17) using the droplet temperature measurement. We observe that the phase-change characteristic time is several orders of magnitude lower than the entrainment characteristic time (see Fig. 18)

This supports our hypothesis that the phenomena are not simultaneous. Vaporization processes drive the first phase and gas entrainment drives the second phase, called here respectively, A and B.

Measurements of gas and droplet temperature are performed during the equilibrium phase of the test by vertically displacing the nozzle (this change of nozzle position has no influence on the equilibrium phase during the test). Gas temperatures are measured by thermocouples (Fig. 9) and droplet temperatures by GRR (Fig. 10).

By coupling these measurements with steam volume fraction measured using SRSS (Fig. 7) at the same location, we can compute the Spalding parameter (*B*_{M}). Spalding (1952) showed that this latter parameter (Eq. 7) is needed to calculate the steam mass flow rate in equation (6) on the droplet.

*Y*

_{steam,∞}) is directly derived from SRSS measurements, whereas the steam mass fraction at the interface (

*Y*

_{steam,int}) is computed assuming the fluid is at thermodynamic equilibrium, and therefore saturated with steam according to equation (8)

*ρ*

_{steam}) and the steam diffusion coefficient in air (

*D*

_{steam,air}) at the mean temperature within the gaseous boundary layer around the droplet (

*T*

_{ref}), which is computed using equation (10)

This model developed by Lemaitre (2005) assumes a thermal boundary layer in the liquid phase and a non-isothermal diffusion boundary layer in the gaseous phase.

The heat transfer coefficient inside the liquid boundary layer is modeled using empirical correlations that depend on the flow regime inside the droplet as determined by Hendou (1992).

In these empirical correlations, the Reynolds numbers are computed using droplet velocity, which is derived from PIV measurements. Hence, we compute the heat and mass transfers at the interface in order to obtain its temperature (*T*_{int}).

We show that, for our experimental conditions, the temperature gradients inside the droplets are so weak that we can consider the interface and droplet temperatures as equal. This is mainly due to the high saturation ratio in the spray zone during the D phase. Studies were previously performed by Lemaitre (2005) to ensure that temperature gradients inside the droplet do not disturb the droplet temperature measurements performed with the GRR technique.

*B*

_{M}< 0, equation 4) from zones where the droplets are vaporizing (

*B*

_{M}> 0).

The uncertainty on the Spalding parameter, which is also evaluated according to the GUM, varies according to the distance from the injection pipe, from 2 × 10^{−2} at 5 cm to 3 × 10^{−2} at 25 cm.

As a consequence, we can identify four zones that are designated as α, β, δ and γ in Fig. 17. To understand the phenomena occurring in each of these zones, we need to compute characteristic times for convective heat transfer (*τ*_{conv}) and phase change (*τ*_{phase}).

*d*

_{GRR}(equation 22), as it is representative of GRR measurements. Moreover, it is close to the mean Sauter diameter (Fig. 11), which is a meaningful measurement of particle size since it represents the size of a droplet with the same surface-area to volume ratio as the entire spray. In addition, the Sauter diameter is known to characterize the heat and mass transfer of the droplets better than the arithmetic mean diameter (Lefebvre (1989)). These characteristic times are presented in Fig. 18.

Since the phase-change, characteristic time in the *α* zone is an order of magnitude greater than the characteristic time for convective heat transfer and, moreover, because the Spalding parameter is negative, we infer that the main process leading to droplet heating is steam condensation on the latter.

In the *β* zone, the Spalding parameter changes from a negative to a positive value, characterizing a transition from a condensing to a vaporizing zone. Nevertheless, since the Spalding parameter is very close to zero, the characteristic time for the phase change becomes orders of magnitude greater than for the convective heat transfer. Consequently, as the gas remains hotter than the droplets, the latter continue to heat up due to the heat transfer.

In the *δ* zone, we observe that the phase-change characteristic time becomes smaller than the convective heat transfer time. Thus, as the Spalding parameter is positive, the droplets start to vaporize in order to dissipate the energy surplus accumulated in the *β* zone.

Finally, in the *γ* phase, the characteristic times have nearly all the same order of magnitude, thus at any single moment, the convective heat flux collected by the droplet is compensated by the vaporization.

### 4.4 Sump singularity

Thus, the spray release will induce both gas mixing and strong droplet vaporization. As mentioned above, vaporization is predominant compared to mixing processes, since the characteristic times differ by several orders of magnitude.

*t*

_{0}and

*t*

_{1}

*n*

_{steam-sump}vaporized in the sump

The average sump temperature is computed from the measurements of 18 thermocouples located throughout the sump volume.

Thus, we calculate that 23 mol are vaporized over the whole vessel (Fig. 5), whereas 9.37 mol of steam are vaporized in the sump zone, which represents only 3.5% of the total volume of the experimental vessel. This highlights the density of mass exchanges in the sump zone.

### 4.5 Validation of two-boundary-layer model

Finally, we perform a validation of the two boundary layer model. For this simulation, the heat and mass transfer coefficients are computed using empirical correlations based on the Nusselt and Sherwood numbers, respectively. The mass flow rate at the interface is computed using a model developed by Spalding (1952).

*d*

^{7/3}, van de Hulst 1957). Therefore, it is important to determine an average diameter

*d*

_{GRR}that can be used as input for the droplet heating simulation, so we can make a pertinent comparison with the GRR measurement. Lemaitre (2005) used numerical simulations to show that the average droplet size representative of the GRR measurement (

*d*

_{GRR}) can be computed from equation (22). Global rainbow simulations are performed using the Airy theory: in the first type of simulation, we assume a correlation between droplet diameter and refractive index (

*m*=

*f*(

*d*)), while the second simulation takes all the droplets as having the same refractive index <

*m*

_{GRR}>, which is computed with equation (23)

By applying the second type of simulation, we obtain a perfect superimposition with the global rainbow refactrometry measurements. This shows that the mean diameter representative of the mean refractive index, as derived from GRR measurements, is equal to the *d*_{GRR} value computed from equation (23). Fortunately, this average diameter is very close to the Sauter mean diameter, which is very suitable for studying heat and mass transfer phenomena (Lefebvre (1989)).

In this figure, we observe a good consistency between the experimental results and the simulation. Thus, we conclude that the double boundary layer model is suitable for modeling the condensation on droplets under thermal–hydraulic conditions representative of a hypothetical nuclear accident.

Despite the satisfactory overall performance, we also observe that our model does not adequately describe the β and γ zones. Therefore, the model should be refined to achieve a better simulation of these previously defined zones.

In addition, supplementary measurements could be performed to improve the representativity of the simulation. Indeed, the simulation does not take into account any correlation between droplet size and velocity. Such a correlation could be investigated by coupling the out-of-focus imaging technique with a particle tracking velocimetry algorithm (Madsen et al. 2003).

However, in the case of our experimental arrangement with high droplet density, the out-of-focus imaging technique is hampered by overlapping interference fringes. These difficulties might be overcome using an advanced set-up developed by Kawaguchi et al. (2002), which involves an optical compression of the image. These measurements could be performed on smaller experiments with less uncertainties.

## 5 Conclusion

In this study, we present an analysis of heat and mass transfers between droplets and the surrounding gas during a test representative of thermal–hydraulic conditions during a hypothetical severe accident in a nuclear power plant. To carry out this study on the TOSQAN experiment, we coupled many innovative optical diagnostic techniques such as global rainbow refractometry, spontaneous Raman spectroscopy and out-of-focus imaging. An analysis of the local measurements enables us to explain the overall behavior of the test in terms of dimensionless numbers (Spalding parameter) and characteristic times. These measurements also highlight the efficiency of the Abramzon–Sirignano model in dealing with steam condensation on droplets, even though this model was initially established for the study of droplet vaporization.