Characterization of the ECN spray A in different facilities. Part 2: spray vaporization and combustion

. One of the objective of Engine Combustion Network (ECN), (https://ecn.sandia.gov/) is to provide experimental results with high accuracy in order to validate model and reach new steps in scienti ﬁ c understanding of spray combustion at conditions speci ﬁ c to engines. The ECN community de ﬁ nes different target conditions, experimental diagnostics and post processing methods to facilitate the comparison of experimental and simulations studies performed in different facilities or models. In this context two French laboratories pro-pose two new facilities, based on Rapid Compression Machines to reach the ECN spray A conditions. In this paper, the results of liquid and vapour spray penetration as well as Ignition Delay (ID) and Lift-Off Length (LOL) obtained with these Rapid Compression Machines are compared to the results obtained in the Constant Volume Preburn (CVP) vessel of IFPEN . The speci ﬁ cities of each experimental apparatus allow to bring com-plementary elements of understanding like con ﬁ nement effects. In non-reactive condition, the liquid and vapour sprays were characterized by Diffused-Back Illumination and Schlieren technique, and in reactive conditions, the LOL and the ID by OH* chemiluminescence. The analysis of the results with regard to the boundary conditions (temperature, velocity, con ﬁ nement) make it possible to validate these two new facilities and contribute to enhance the database of ECN, highlighting the con ﬁ nement effect typical of piston engine operation.


Introduction
Since 20 years, due to increasingly drastic emission standards, many studies have been focused on spray and combustion for Internal Combustion Engines to improve the efficiency and reduce the pollutant emissions. The development of Computational Fluid-Dynamic (CFD) simulation could allow to compute a wide range of chamber geometries and operating conditions for the optimization at substantially lower cost than experimental tests. However, the predictability of CFD remains depending on the degree of understanding of the physical phenomena of spray and combustion into the chamber [1,2]. In order to provide accurate data about diesel spray and combustion processes, several combustion chambers have been developed to reach High Pressure-High Temperature (HPHT) conditions, representing actual common-rail diesel conventional operating modes. As example, Constant-Volume Preburn (CVP) chamber, Constant-Pressure Flow (CPF) chamber, Rapid CYcling Machine (RCYM) [1,[3][4][5][6][7] have been developed to investigate fundamental phenomena of diesel spray and combustion by using optical techniques. The specificity of each one provides advantages and disadvantages.
In the case of CVP chamber, a premixed combustion is used to generate the HPHT condition, adjusted by varying the combustible-gas mixture [1,[3][4][5]7]. Then, the ambient gases pressure and temperature are slowly cooled down due to the heat losses until the target ambient condition at which the injection is started to study the spray and combustion developments, as detailed in [4,[6][7][8]. During the pre-combustion, the combustion products (CO 2 , H 2 O, Ar, N 2 and O 2 ) and some minor species (NO, NO 2 and OH) are present in the ambient gases when the injection starts [6]. This does not affect the spray process itself but certainly the combustion process, as the ignition delay, the lift-off Engine Combustion Network -France Christine Rousselle and Gilles Bruneaux (Guest editors) length, the soot production and oxidation [9][10][11]. For example, Nesbitt et al. [9] found that both major (CO 2 , H 2 O, O 2 ) and minor species (NO, NO 2 , OH) influence the ignition delay and the changes due to minor species are small relative to those from major species. Even if those species can be reduced by varying the mixture composition used to realise the pre-combustion, small quantities remain [12,13]. But, the CVP vessels are of interest due to the wide optical access provided by the number of optical windows (up to four). Moreover, it is worth highlighting that the CVP set-up can be used on a wide range of ambient temperature and pressure with an easy flexible switch between different conditions with a time between each test about 5 min Finally, the relatively low cost associated to develop these facilities is another advantage. This is why they are widely used in different research institutes (IFPEN [12], Sandia [4,6], TU/e [6], MTU [12,14]).
In the CPF chamber, the gases are compressed by a volumetric compressor in high-pressure tanks [6]. Then, the gases flow continuously through the electrical resistances to reach the target temperature before entering in the test chamber. The gas flow scavenges the injection fuel and/or combustion products to maintain the ambient temperature. A control system provides the desired test condition by measuring temperature and pressure to continuously adjust the power of heater and gas flow to supply in the upstream of the chamber. The operating condition of this chamber has been described in details by Payri et al. [1,6,13,14]. Generally, the global shape of this chamber is similar to a large pipe. Therefore, the spray and combustion evolution can be observed in a long distance thanks to several large windows, positioned at 90°angle around the circumference of the chamber. This kind of test rig is able to reach temperatures and pressures up to 1000 K and 150 bar [13]. As the gas flows continuously, this allows to reach a high injection repetition rate about 3 s. However, due to the requirement of compressed clean air and a continuous flow of gases, this facility is large, heavy and expensive.
The RCYM [3,4,15,16] is another interesting option to study spray in HPHT conditions, due to the absence of combustion products like in the CVP chamber, with the possibility to reach high pressure conditions to cover a full range of diesel engine working conditions (up to 100 bar and 900 K). The use of a 2-stroke engine allows the gases to be flowed and exchanged through the side-port-scavenging [3]. In [3] the original cylinder head was modified to support optical windows from different sides of combustion chamber. As it continuously operates and scavenges the injected fuel and combustion products every cycle (near Top Dead Center (TDC)), a high repetition rate can be reached and the ambient gases are those introduced with a controlled composition. However, the main drawback of this device is the small volume of its chamber at TDC even when using a large displacement engine, which can generate high vibration and noise levels during operation. Last, because of side-port-scavenging, high-velocity fields inside the chamber can be generated to affect spray and combustion processes [15,17].
The comparison of the boundary conditions, spray development and combustion parameters between CVP and CPF vessels and the different techniques of measurement have been extensively discussed in the Engine Combustion Network (ECN) [1,6,7,10] with some recommendations. Diffused Back-Illumination is now preferred over Mie scattering when possible to determine the liquid penetration length [18,19]. To follow the extension of the vapour phase, the bright or dark-field Schlieren or focused shadowgraph can be used [20]. To study the combustion process, the OH* chemiluminescence is recommended both to detect the lift-off-length and to determine the ignition delay thanks to a high frequency rate recording, even if it can also be determined by analysing pressure signals or Schlieren imaging [21]. These three techniques do not have the same sensitivity to the different ignition phases as the cool flame and high-temperature ignitions [22,23]. By using the Global Sensitivity Analysis (GSA) method [10,12] to investigate the influence of the different boundary conditions on the combustion parameters between the two kinds of chamber CVP and CPF, the most sensitive variables outside nozzle diameter are the fuel temperature for the liquid length, the initial turbulence intensity for the vapour penetration, the composition of the initial ambient gases (O 2, major and minor species) and their temperature for the ignition delay and the flame lift-off length.
The international Engine Combustion Network [2] has the objective to share accurate experimental data with high accuracy in order to reach new steps in scientific understanding of spray combustion at conditions specific to diesel and gasoline engines to improve and validate CFD models [4]. In order to extend this network, two new facilities were designed and improved to reach the Spray A thermodynamic conditions, which are recommended by the ECN to represent current common-rail light duty diesel engines. They are both based on Rapid Compression Machine concept but with different specificities. In the following, the first one is called "New One Shot Engine" (NOSE) [24][25][26] and the second is PPRIME_RCM [27][28][29][30].
Therefore, the objective of this paper is to introduce these new devices and to validate their use at ECN Spray A conditions. Different single-hole injectors were used in these set-ups as the characterization work were conducted simultaneously. All injectors were first extensively tested in IFPEN_CVP where the spray characterizations are used as reference for comparison to avoid discrepancies related to uncertainties about the injectors' nozzle geometry. Using the best ECN practices for the measurement techniques and data processing, the Liquid Length (LL), the spray Vapour Penetration (VP), the Ignition Delay (ID) and the Lift-Off Length (LOL) are quantified and compared. The effects of confinement, highlighted by the smaller ambient gas volume of RCYM compared to CVP and CPF, will be discussed with regard to the ambient gas composition.  [2]. The initial and the boundary conditions are described in [29] and the main characteristics of the different facilities are summarized in Table 1. The NOSE and PPRIME_RCM facilities are more flexible to control the ambient gases composition [31] but due to the need of high compression ratio, the clearance volume is smaller than the one of the pre-burn vessel inducing more confinement. Each of the measurements presented in the following section has been repeated 5-10 times to provide statistical information.

Injection system
The entire fuel injection system, including the common-rail, the injector and the high-pressure tube, is identical for all three set-ups to fulfil Spray A specifications [2]. A 22 cm 3 volume and 28 cm length common-rail is used, and is connected to the injector with a 24 cm tube. The injection pressure is measured by a high response piezo-electric pressure sensor (KISTLER Type 6533A11), positioned 7 cm from the injector inlet, and connected to a charge amplifier (KISTLER Type 4618A2).
For the NOSE set-up, a high pressure pneumatic pump (MAXIMATOR M189 DVE-HD), driven by 7 bar compressed air, maintains the pressure around 1500 bar (±50 bar) before the Start Of Injection (SOI). For PPRIME_RCM, a similar hydro-pneumatic pump (MAXIMATOR GSF300) is used but the injector return line is connected to a backpressure regulator. For the reference set-up, CVP chamber has already been extensively described in [6]. The same injector driver-settings were used as specified in [1].
Three different Bosch CR2.16 injectors, with the same nominal orifice outlet diameter of 90 lm are used in this work: Injector #14 in PPRIME_RCM, Injector #16 and Injector #19 in NOSE. These injectors are single hole, axially oriented. The nozzle shape was done by hydroerosion with a 1.5 k-factor and the mini-sac volume was 0.2 mm 3 . The specificities of these 3 injectors are compared in Table 2.

Diffused Back-Illumination (DBI)
As in [1,2], the liquid penetration length is defined as the maximum distance between the injector orifice and the tip of the liquid spray. DBI is recommended by ECN as the standard technique to measure the Liquid Length (LL) [18,19]. The set-up of this technique, which consists of a diffused light source a camera, is shown in Figure 1.
In NOSE, a white LED light plate 100 Â 100 mm 2 in size was used to illuminate the liquid phase of the spray. The light intensity through the optical windows was collected by a high-speed camera (Phantom V1611), equipped with a NIKON 60 mm f/2.8 lens, as shown in Figure 1a. A frame rate of 49 kHz with an exposure time of 3 ls, and an image resolution of 512 Â 512 pi 2 provide the most suitable values to capture high-quality images at a high-frequency rate.
Two different setups were implemented in PPRIME_ RCM to check the sensitivity to using different type of illumination and different exposure times. The first setup, Figure 1b is composed of an ion laser (Spectra Physics Stabilite 2017) and a combination of Bragg cell, iris diaphragm and concave mirror to generate a pulsed parallel beam. This beam is projected on an engineered diffuser placed at 60 mm from the injector plane. On the opposite side, a fast camera (running at 72 kHz frame rate) is installed and focused on this plane. In this configuration, the exposure time is set on both the camera and the Bragg cell to 1 ls. In the second setup, the parallel beam is generated by using a laser diode and an aspheric lens, therefore the exposure time is only fixed by the camera. More details about DBI set-ups are summarized in Table 3. The post-processing was performed using the code available on the ECN website [2]. The images acquired during the steady-state period of the spray (500 and 3000 ls) are used to compute the time averaged image intensity I avg . This allows to avoid the initial and final transients of the spray. I avg is then normalized by the time averaged background image I bg , obtained by averaging about fifteen images before the Start Of Injection (SOI). Then, the light extinction factor along the spray core (s) is calculated using the Beer-Lambert law as shown in equation (1). To avoid the effect of the beam steering phenomenon due to the variation in the refraction index on the medium at the liquid spray tip, the distance where the linear fit line crosses the X-axis was determined as the LL of spray as shown in Figure 2, In Figure 3, the accuracy of the values obtained with the standard post-processing was evaluated by changing the DBI setups with related exposure time in PPRIME_RCM set-up. The good agreement between the results confirms their robustness on the type of light.

Schlieren imaging
The Schlieren imaging is a well-known technique to visualize the refractive index gradients in transparent mediums [32]. In the case of vaporizing diesel sprays, this technique is able to capture the line-of-sight boundary between vaporized fuel and ambient gases [33][34][35]. A pinhole, also called "aperture Schlieren stop", was used to increase the contrast between the vapor and the background.
In NOSE experiment, the Z configuration Schlieren setup is composed of the continuous white LED with a 1 mm pinhole, to represent the one point-light source and two parabolic mirrors with 108 mm diameter and 837 mm focal length, as schematized in Figure 4a. A 6 mm pinhole is used to record Schlieren light after the test section with the high-speed camera (Phantom V1611) and a collecting lens of 30 mm focal length. The image resolution was 1024 Â 400 pi 2 for 39 kHz frame rate with 5 ls of exposure time and 12.3 pix/mm of magnification.
In RCM_PPRIME, an ion laser (Spectra Physics Stabilite 2017) is used as light source, pulsed up to 1 ls   Figure 4b. A combination of an iris diaphragm and a concave mirror is used to generate a parallel beam, steered through the combustion chamber.
A first convergent lens is used to focus the light beam on a knife edge and a second one to project the refracted beam on a diffuser screen. A fast camera (Photron SA-Z) is used for image recording at 42 kHz, with an image resolution of 640 Â 600 pi 2 . Its 1 ls exposure timing is synchronized with the Bragg cell actuation system. The images are post-processed using the same Matlab script shared by the ECN [36]. The vapor phase edges are detected by comparing the projected density gradients from successive images. In Figure 5, examples of Schlieren image and of spray boundary (in red) determined from the processed image are given. The vapor penetration length is determined at the crossing point between the limit of the vertical spray front limit and the spray axis from injector tip (Tab. 4).

OH* chemiluminescence
Follow the fuel and thermodynamics conditions, two ignition phases can be identified during the combustion process: the first one corresponds to the cool flame, i.e. low temperature combustion and the second one to the main flame, i.e. "hot flame" [2]. The cool flame corresponds to the moment where the parent fuel molecules are broken down which slightly increases the ambient temperature. After that, the main combustion phase occurs, when OH* radicals are produced and can be detected from their UV chemiluminescence at 310 nm [23,33,37]. This signal is used to characterize both the Lift-Off Length (LOL) and the Ignition Delay (ID), which is defined as the time between the Start Of Injection (SOI) and the beginning of the high-temperature combustion phase. The LOL represents the distance between the orifice of the injector and the axial location where the flame is stabilized [2,12,13,38]. To compute the LOL, intensity profiles around the spray axis are extracted from the recorded chemiluminescence images. The intensity is normalized to maximum values located at the flame lobes region: red and blue profiles around the spray centreline as shown in Figure 6. The LOL is defined as the average of the distances between the injector tip and the distance corresponding to 50% of the maximum intensity, following the ECN method [38].
In NOSE, the intensified CMOS Photron Fastcam APX I2 camera was used with a 60 mm f/3.5 UV lens, equipped with a 310 nm (FWHM 10 nm) Band-Pass Filter (BFP) with a long gating time of 449 ls to average LOL during a quasi-steady phase after start of ignition (1500-3000 ls) [39]. Ten images were recorded but only 5th-7th images, i.e. 2.0 ms after SOI were considered to calculate the steady-state LOL. A Newport Oriel Photomultiplier, sideon with a PMT 70705 high voltage power supply and a BPF of 307 nm (FWHM 10 nm) were used to record temporal OH* chemiluminescence signal during 4 ls to determine ID as the time where OH* intensity reaches its maximum value.
In PPRIME_RCM, two different setups were used: first, a combination of a fast CMOS camera (Photron SA-Z), an intensifier (Hamamatsu), a UV lens and a short band-pass (SBP Asahi 310 nm) filter to record OH* chemiluminescence images at 45 k fps (2.5 ls), to determine both ID and LOL and second, an ICCD camera (Princeton PI-MAX4) to record longer exposure OH* chemiluminescence images at 1 ms after the start of ignition.
All specificities are summarized in Table 5.

Vapor phase penetration
In Figure 7, the averaged evolution of Vapor Penetration (VP) is plotted as a function of time after SOI (obtained  with an accuracy on 10 ls by analysing the mass flow rate or the DBI images with very high frame rate). The results obtained on NOSE and on PPRIME_RCM are compared to IFPEN CVP ones, and the absolute difference are also plotted in Figure 7. The vapor penetration measured on PPRIME_RCM is lower than CVP one for early timings, but after approx. 500 ls the difference drops below 3%, thus showing a good agreement. For injectors #19 and #16, NOSE experiments provide 4% lower VP, in comparison to CVP experiments. This may indicate that the air entrainment is slightly higher in NOSE. By considering identical thermodynamic condition, the Naber and Siebers model [40] can be used to estimate the spreading angle which enables to match the measured penetration in the two cases. The angle obtained is respectively 21°and 22.5°for CVP and NOSE. This would mean that both sprays #16 and #19 are wider compared to CVP experiments, and that they have evolved in an identical manner. Otherwise, the lower vapor penetration may be caused by slight differences in ambient density. In NOSE, the protrusion of the injector is slightly greater than in CVP and the wall temperature is lower than IFPEN_CVP (Tab. 1), thus leading probably to more entrainment of the colder and denser gases from the boundary layer [29].

Liquid phase penetration
The Liquid Phase Penetration obtained in the two new facilities (NOSE and PPRIME_RCM) are compared to the values obtained in CVP in Figure 8. For all three  institutions, a temperature increase leads to a decrease of liquid length, as expected. However, the magnitude of this decrease is significantly lower for PPRIME_RCM compared to CVP and NOSE. Besides, the absolute values of the liquid length are higher for PPRIME_RCM compared to NOSE and CVP, beyond the uncertainty of the measurements.
In the following discussion, to understand the effect of the different factors that can affect liquid length, 1D models will be used. In [40,41], Naber and Siebers pointed out that the vaporization is mainly governed by the air entrainment rather than the interphase transport. Using several simplistic assumptions coupled to the mass, momentum and energy balances, a Liquid Length (LL) scaling law was introduced in [41]. The model was validated on a wide range of experimental conditions performed on a CVP vessel [42]. The experiments showed a clear dependence of the LL to the nozzle diameter d, the entrained air density q a and temperature T a , the fuel temperature T f and the fuel volatility, as in the LL model: with B the term derived from the energy equation as, where a and b are constants, q is the density, C a is the orifice area contraction coefficient, h is the full cone angle of the real spray, Z is the compressibility factor, T s and P s are the temperature and pressure at the saturated fuel vapor condition and at the maximum penetration distance of liquid-phase fuel, h is the enthalpy, M is the molecular weight, the subscripts "a", "f" and "lf" indicate respectively for ambient gases, fuel and liquid fuel.  In your case, the non-ideal gas effect is neglected, B can be rewritten as: where x i is percentage of each gas, the enthalpies h ia of each gas are estimated with an interpolation of NIST data [43], lv is the latent vaporization heat of dodecane corrected with the Watson law and cp lf is the thermal capacity estimated with an interpolation of NIST data at T s þT f 2 . The pressure is calculated with Antoine's law [43], log 10 P s ð Þ ¼ 4; 1 À 1625:9 T s À 92:8 :  The unknown, T s , can be solved iteratively, given the fuel and ambient gas properties and initial fuel and ambient gas conditions. Once determined, T s defines B, as well as the pressures, temperatures, and enthalpies of the fuel and ambient gas at the liquid length location. For injector #19, liquid length measured in CVP is approx. 5% higher than in NOSE at 800 K, and 6% higher at 900 K. This is the order of magnitude of the variation of liquid length (7%) computed with the 1D spray model using the spray angles of 21°for CVP and of 22.5°for NOSE estimated from model to fit the vapor penetration, as described in the previous section (cf. Fig. 10).
The differences observed in liquid length for injector #19 between NOSE and CVP could thus be explained by a variation of the air entrainment. Also, taking as a reference the liquid length at 900 K, the 1D spray model gives an increase of 10% of the liquid length at 850 K. This increase should be 20% at 800 K. These are the order of magnitude of the increase of liquid length when temperature decreases for NOSE and CVP. But the increase is much lower for PPRIME_RCM: the liquid length increases only by 2% when the temperature decreases from 900 K to 850 K. Also, the differences in absolute values of the liquid length between PPRIME_RCM and CVP are higher in magnitude. The main factors that may explain this discrepancy in the LL are: 1. The DBI experimental set-up and post processing of the images may induce variations in the measured liquid length. But the Figure 3 shows that the measurement is robust on different type of light. 2. Since the injector tip protrusion is not exactly the same: the injector protrudes further inside the combustion chamber of NOSE, the fuel temperature can be slightly higher, inducing a reduction of LL. However, a 60 K decrease in fuel temperature would lead only to 15% increase of liquid length at 900 K, which is well below the 40% increase measured on PPRIME_RCM. So the fuel temperature alone cannot explain the differences in liquid length. 3. The different spreading angle: But since the vapor penetrations are similar, this hypothesis should be discarded.
4. The ambient temperature: in PPRIME_RMC it was demonstrated [29] that during the injection, the colder gases from the boundary layer are entrained in the spray region, which decreases the effective ambient temperature. However, 40% increase in liquid length would require a decrease of the ambient temperature from 900 K to 750 K, which is unlikely. So the ambient gas temperature alone cannot explain the differences in liquid length. 5. The differences in ambient gas composition: since the properties of these gases may affect the mixing and the fuel evaporation. This point will be discussed below. Table 6 presents the composition, pressure, C p and molar weight of the ambient gases at injection timing.
During PPRIME_RCM experiments, the liquid length has been measured for variations of the pressure and C p of ambient gases, keeping the density constant. This has been done by adjusting the proportions of N 2 , Ar, CO 2 and He (Tab. 6). The results are presented in Figure 9. The LL shows a decrease with increased P a . Most likely, the decreased LL observed with varying the gas composition is explained by the decreased molecular weight and increased thermal conductivity of the gas. Helium has a significantly higher thermal conductivity and lower molecular weight compared to the other ambient gases. Indeed the Siebers Model 41] explains this trend by a change in Spalding number at equilibrium. The saturated temperature is estimated at equilibrium (see Eq. (4)). The Spalding number increases with the decrease in molecular weight despite an increase in pressure. The liquid length is then calculated with a spreading angle of 21°± 1°and with a ratio b a chosen in order than model matches with IFPEN_CVP data at 900 K, 60.1 bar. The calculated results are also presented in Figure 9 and show the same trend. However, the decrease in the liquid length is greater. But at 48 bar, the estimate is very close to the measurement. Thus, the model does predict a 21% difference between the CVP and RCM case. At 48 bars, by correcting for the effect of gas composition and back pressure, the difference between CVP and RCM is reduced to less than 4%. Thus between 850 K and 900 K the magnitude of this decrease is explained. The estimate shows a variation of only 2% as measured (cf. Fig. 10).
As a conclusion, the main effect which explained the difference between NOSE and IFPEN_CVP is the difference spreading angle and between PPRIME_RCM and IFPEN_CVP is a combination of effect of different ambient gases composition and back-pressure. Figure 11 presents the ID values determined from OH* chemiluminescence signal as a function of the ambient temperature. To add a reference for the case at 850 K, the average of the results obtained from other ECN contributors have been added with the related uncertainty [12,44,45].

Ignition delay and lift-off length
As expected, all the laboratories detected a decrease in ID with the increase of the ambient temperature, due to the effect on the chemical reactions and in agreement with other observations [12,45]. The measured values in different facilities are substantially similar. The differences observed are generally smaller than the experimental uncertainties, which increase at lower temperature.
The ID measurement obtained with NOSE are in good agreement with the reference measurement at IFPEN_ CVP, with differences well below the test to test variations at the tested temperature levels. The results obtained in PPRIME_RCM have significantly higher discrepancy of ID at 900 K. The ID is 30% longer than the one obtained by IFPEN_CVP using the same injector. However, at 850 K the result is in better agreement with ECN average reference data. This behaviour is partly consistent with the liquid length results and the estimated temperature at the saturated fuel vapor condition. As for LL results at 900 K, the ID is significantly higher than the reference and the sensitivity to the temperature variation is lower than observed in other facilities. As discussed in [29], the decrease of temperature during the vaporisation phase can cause both of these effects. However, at 850 K the ID value measured at PPRIME_RCM is rather close to the reference average value, while LL was still significantly longer than the reference. The ID results from PPRIME_ RCM showed generally lower sensitivity to temperature compared to the other facilities. Most likely, such behaviour is explained by the significantly different gas composition in this facility. Figure 12 shows the experimental result of LOL as a function of ambient temperatures. As expected, by increasing ambient temperature LOL is shortening due to the combined effects of faster evaporation and mixing processes and local reaction rate, allowing the flame to stabilize closer to the injector [46]. The deviations between the different set-ups remain below the standard variations from test to test. Also in this case, reference data from the ECN database, has been added for the case at 850 K.
More in details, all the measurements are very close to each other at 900 K. The measurements from NOSE well follow the temperature effect. As for the ID, the values at 800 K are slightly shorter than the values obtained at IFPEN_CVP. Even if this difference is smaller than the  experimental uncertainty, the consistency of these two results might indicate some effect of difference in ambient temperature at 800 K or some impact of the gas composition [29].
The results presented by PPRIME-RCM at 850 K indicate that in this case the results better follow the temperature sensitivity detected by in other labs. To better understand this aspect, the temporal evolution of LOL from OH* chemiluminescence has been investigated as plotted in Figure 13 for 3 different ambient temperature conditions. Unlike the other ECN facilities, the LOL is not constant and decreases progressively. At the end of injection, the LOL is approximately 20% lower than at the start of combustion. In [29] it was demonstrated that due to the smaller chamber volume of PPRIME_RCM, the spray combustion induced a substantial increase of the ambient temperature. Most likely, this increase of ambient temperature causes the progressive decrease of the LOL presented in Figure 13.

Summary and conclusion
Two new experimental set-ups based on rapid compression machine were designed and improved to study the atomization processes at spray A conditions. The results obtained were compared to values measured at IFPEN with a CVP and to other reference data from the ECN database. In most cases, the observed differences in the measured spray characteristics are comparable to the standard deviations from repeatability tests. These are likely caused by the levels of accuracy and precision achievable to control the boundary conditions: such as the injection pressure, the ambient temperature, the fuel temperature, the ambient density and the ambient gases composition.
To reach the two temperature conditions (900 K and 850 K) for the same density, PPRIME varies the pressure and the gas composition. Thus the observed trends differ concerning the liquid length and ignition delay. The sensitivity to the two conditions tested are very close (variation of 2%). When only the temperature varies with a small modification of pressure, PRISME and IFPEN for one observe the same sensitivity to temperature decrease for all experiments.
So the current study particularly highlights that some of the discrepancies can be explained by: 1. The composition of the ambient gases: The properties of the gases in the ambient environment such as, molecular diffusivity, heat capacity, thermal conductivity and compressibility influence the evaporation phenomenon and thus the ignition delay.  2. The confinement of the spray (variation of the ambient conditions during the spray event): with smaller chamber volume and colder boundary layers, the entrainment of ambient gas may be different and thus modify the liquid and vapor penetration. During the combustion process, it is demonstrated that the significant increase of temperature and pressure affected the length of the Lift-off. 3. The differences in ambient conditions and particularly the temperature of the boundary layer which affects the density homogeneity as also highlighted in Pei et al. [10].
The current work as in its first part [29], suggests to consider the specificities of each experimental device to improve the comparison of the spray characteristics from different facilities. This highlight that the validation of simulations should also consider these specificities to further improve the accuracy of the models.