Some characteristics of multi-jet flares

Results of a test program with a multi-jet, high momentum burner simulating many of the features found in offshore flare systems are described. Previous work (Pagot et al., 2003) has given background on a hot and cold model study for this type of flare. The features of the flare flames are described in more detail in this paper. The burner consisted of twenty-four nozzles mounted on eight supply lines with firing rates in the range of 268 - 523 kW, corresponding to jet velocities of 39.3 - 78.5 m/s with a natural gas fuel. The flare model was tested with different nozzle fittings and jet to cross-flow momentum flux ratios ranging from 65 to 268. The flame structure was identified through the use of still pictures and CCD imaging techniques. Gas temperature and composition measurements along the flare trajectory were also obtained. This multi-jet configuration provides good flare stability allowing firing rates higher than single jets in a cross-flow. Over the range of firing rates investigated, the flame could be divided in two regions, a momentum interaction zone where there is an initial mixing between the fuel jets and the cross flow, followed by a zone where there was a straight flame trajectory characteristic of a strongly buoyant flame. Near the transition point between these regions there was a higher gas temperature, minimum O2 concentration and maximum CO concentration. The flare efficiency, expressed in terms of the unburned hydrocarbon fuel, was in typically in excess of 99%.

A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 S o m e C h a r a c te ris tic s o f M u lti-J e t F la re s P.R. Pagot1, E.W. Grandmaison* and A. Sobiesiak2 Department of Chemical Engineering Queen's University Kingston, ON K7L 3N6 CANADA 1 Petrobras, Petroleo Brasileiro S.A. 2 Department of Mechanical, Automotive & Materials Engineering, University of Windsor, C orresponding author: Department of Chemical Engineering Queen's University Kingston, ON K7L 3N6 Canada TEL: [phone] FAX: [phone] Email: [e-mail] ABSTRACT Results of a test program with a multi-jet, high momentum burner simulating many of the features found in offshore flare systems are described. Previous work (Pagot et al., 2003) has given background on a hot and cold model study for this type of flare. The features of the flare flames are described in more detail in this paper. The burner consisted of twenty-four nozzles mounted on eight supply lines with firing rates in the range of 268 - 523 kW, corresponding to jet velocities of 39.3 - 78.5 m/s with a natural gas fuel. The flare model was tested with different nozzle fittings and jet to cross-flow momentum flux ratios ranging from 65 to 268. The flame structure was identified through the use of still pictures and CCD imaging techniques. Gas temperature and composition measurements along the flare trajectory were also obtained. This multi-jet configuration provides good flare stability allowing firing rates higher than single jets in a cross-flow. Over the range of firing rates investigated, the flame could be divided in two regions, a momentum interaction zone where there is an initial mixing between the fuel jets and the cross flow, followed by a zone where there was a straight flame trajectory characteristic of a strongly buoyant flame. Near the transition point between these regions there was a higher gas temperature, minimum O 2 concentration and maximum CO concentration. The flare efficiency, expressed in terms of the unburned hydrocarbon fuel, was in typically in excess of 99%. Introduction Flares are used to burn waste and vent hydrocarbons from petrochemical production and processing facilities. Research in the problem o f flaring technology has increased in recent years in support of more stringent regulations related to combustion efficiency, pollutant emissions and radiation levels from flaring operations. An early review of the flaring problem has been presented by Brzustowski (1976) and more current topics are discussed by Bandaru and Turns (2000). These references provide background dealing with a single jet and relatively low momentum flares. 1 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 The problem of interest in this paper deals with the structure and properties of multi-jet flares used in offshore oil production systems. The flaring process for these units pose problems similar to land-based operations but with added complications due to the rather isolated nature of these production units. Some of the concerns include emission levels and the radiation levels incident on the working area of oilrigs at current and proposed increased production rates. High capacity offshore flare systems able to burn up to 4 x 106 m3/day (~1.9 GW) are commonly used. A typical ship-based facility is shown in Fig. 1, with the flare system mounted on a large boom (~80 m long) angled from the main deck of a FPSO (Floating Production Storage and Offloading) ship. This geometry helps maintain flame radiation at an acceptable level for the working environment on the ship. A top view of a flare system is shown in Fig. 2, depicting a series of smaller burners mounted on eight arms emanating from a central fuel supply manifold. The smaller burners consist of another series of eight arms, each with multiple fuel ports constituting the multi-jet arrangement commonly used in this flare configuration. This burner system is intended to provide consistent combustion efficiency and flare performance independent of wind conditions. In the present work (Pagot, 2002), a flare model was developed incorporating three different nozzle attachments to investigate the effect of these mixing-altering devices on the resulting flare system. An overview of some cold- and hot-model features of this work has been presented by Pagot et al. (2003). In the present paper, more details of the flare performance are described in terms of emissions (O2, CO, unburned hydrocarbons and NOx), gas temperature and geometrical features of the flame. Figure 1. An offshore FPSO ship with a flare boom shown in the left picture and an operating flare shown in the right picture. 2 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Figure 2. Top view of a typical multi-jet flare system showing forty-one individual burners mounted on a primary manifold and eight arms supplying fuel for the burners. Experim ental The flare model consisted of a series of eight arms attached to a central manifold with three jet nozzles on each arm as shown in Fig. 3. The diameter of each nozzle was dj = 3.0 mm and the spacing between the nozzles on each arm was S = 48 mm. This separation distance was selected in order to minimize interference between the jets in the near field region of the jets (Menon, 1984; Menon and Gollahalli, 1988). This burner configuration with 24 plain nozzles constituted the reference case for comparison with three other cases examined in this work. The three additional cases, Fig. 4, had different mixing altering devices positioned above the nozzle ports - (i) cone attachments simulating a bluff body stabilizer, (ii) a sudden expansion nozzle creating a precessing jet and (iii) a ring attachment mounted above the nozzle ports. 3 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Gas Supply Figure 3. Side view (left diagram) of the burner model, all dimensions in mm - the 87 mm dimension corresponds to the elevation for the model above the wind tunnel floor in cold model studies (described by Pagot et al., 2003). Top view (centre diagram) showing the position of the jet nozzles in the eight arms originating from the supply manifold for the model. Partial top view (right diagram) showing dimensions for the position of the jet nozzles. Figure 4. Nozzle configurations employed in this experimental study: (i) cone attachments (top), precessing jet nozzle (middle) and (iii) ring attachments (bottom). 4 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 For the hot model studies, a special test facility was constructed outside the main building of the CAGCT (Centre for Advanced Gas Combustion Technology, Queen's University). The flare model was mounted on a fixed base with probes for temperature and gas composition measurements on a three-dimensional traverse system downstream from an open wind tunnel shown schematically in Fig. 5. This facility provided a mean cross flow velocity of 3.9 m/s and an average turbulence intensity of ~13% just upstream from the flare model. Commercial natural gas was supplied to the flare through a series of regulators, needle valves and an orifice meter at five different flow rates as shown in Table 1. Gas temperatures were measured in the flare flames with thermocouples manufactured from 50 |j,m diameter Pt-10%Rh (type S) thermocouple wire; a sampling probe (Turnbull, 1995) was used to withdraw and monitor gas samples from the flame for concentrations of O2, CO, NOx and CH4. The mean radiation flux was measured with a calibrated thermopile radiometer with a water-cooled aluminum block. These quantitative measurements were supplemented by flame-image recordings with CCD, infrared, VHS and single shot still cameras. Tests with the hot model system were only performed during relatively calm night time conditions. 800 C r o s s flow air s u p p ly 1000 y I - m . 150 Natural gas supply Figure 5. Side and front view of the hot model test facility showing the wind tunnel supply for the cross-flow, mounting assembly for the flare model and the three-dimensional traverse system for flame probing. Table 1. Hot model (combustion) experimental conditions. Nozzle conditions s S uj Test conditions NRe Flaring conditions m f , kg/h Heat release, kW Cross flow conditions U, m/s P f f Pou 2 A 78.5 16,800 35.8 523 3.9 261 B 70.9 15,100 32.3 471 3.9 213 C 55.0 11,700 25.0 384 3.9 128 D 39.3 8,400 17.9 268 3.9 65.4 E 17.8 3,800 8.1 140 3.9 13.4 5 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Results Hot model tests were performed with the flare model fitted with the three burner attachments shown in Fig. 4 and the reference case nozzles. The bulk of the tests were performed at flow conditions A - D with each of the burner attachments and flow condition E for the reference and ring attachments. Flame images were obtained in CCD and still picture format to provide an indication of the flame shape and trajectory. Gas temperature and species concentration were measured along the trajectory of the flames for these operating conditions. The radiation heat flux was also measured for each nozzle fixture at firing rates up to 523 kW. The flare model produce stable, lifted flames as shown by the photographs (1/30s exposure) in Fig. 6. The left picture for each case shows the side view of the flame and the right picture shows the front-view observed from a downstream position. The individual jets from each nozzle remained distinct and separate up to a distance of about 2S (S denotes the distance between adjacent jets in the flare model). Subsequently, the flames merge into a luminous region that is more evident on the downstream side of the flare. The visible flames for the flare models with cone attachments and precessing jets were larger than the reference case and the ring attachment nozzles. The visible flames for the ring attachments nozzles were the smallest of the four cases. CCD images of the flames were also recorded at 1/500s and 300 samples were used to generate black and white images (100s time period) to estimate the flame trajectory, length, area and volume. Images for the four flare cases are shown in Fig. 7 for test condition A. The top photographs for each flare case show the black and white rendering of the average image for the flame and the bottom graphs illustrate the extent of the flame in terms of a 50% intermittent region for the edge of the flame. More detailed intermittency profiles (10, 50 and 90%) based on the side-view of the flames for the reference case and the cone attachments are also shown in Fig. 8. These data indicate that the general shape of the flames are quite similar at the different intermittency levels, but with a broader distribution on the downstream side of the flare model. Based on these observations the 50% intermittency level was chosen to provide an estimate of the flame length, flame volume, and more importantly, the buoyant flame volume at the different operating conditions. The flame trajectories depicted by the solid symbols in Fig. 7 were estimated from the midpoint of the flame wide (50% intermittency) at any elevation, z, above the flare source. The front-view of the flames (right photographs in Fig. 6 and right pictures and graphs in Fig. 7) shows a good vertical alignment of the flare models. The front-view also shows a relatively smooth ellipsoidal shape for the reference case and ring attachments, but a more bulbous shape in the lower part of the flare for the cone and precessing jet attachments. The flare trajectories for the side view images in Fig. 7 follow a smooth trend through the initial region of the flame followed by a relatively straight trajectory further downstream. An explanation for this behaviour is shown in Fig. 9 where the side-view 50% intermittency profile is shown for the reference case at flow condition A. Superimposed on this profile is the trajectory for a 3 mm diameter single jet flame in a cross flow (for the flow conditions used in the present study, over a trajectory distance of 300mm) based on the correlation of Brzustowski (1976, 1980). Two regions are also shown in this figure - Zone 1, where there is a strong jet/cross-flow interaction and Zone 2, where there is strong evidence of buoyancy for the combined system of jets. The correlation of Brzustowski (1976, 1980) should show good agreement with the lead jet in the flare model since this nozzle fluid is not strongly affected by the downstream jets in the initial jet/cross-flow interaction region. The present results appear to 6 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Reference case: Cone attachments: ii Precessing jet attachments: Ring attachments: < / Fig. 6. Side and front-view images (left and right photographs, respectively) of the flare flames under test conditions A, based on still pictures at 1/30s shutter speed. 7 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Reference case / x, mm x, mm Cone attachments x, mm F ig u r e 7. c o n t 'd x, mm ... 8 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Precessing jet attachments x, mm x, mm Ring attachments $ 1200 900 | 600 400 200 1200 Figure 7. -AC a -600 -400 -200 0 200 400 600 SO Q x, mm x, mm Time averaged (100 s) CCD images of the side and front view of the flames operating at flow condition A (top picture for each flow fixture) and 50% intermittency profiles of the side and front view of the flame (graph images for each flow fixture). The flame trajectory deduced from the intermittency profiles is shown by the solid circle symbols. 9 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Reference case Cone attachments x, m m x, m m Figure 8. Flame intermittency profiles for the side-view of the flare (condition A, Table 1) for the reference case and cone attachments. /\ Zone 2 Zone 1 Figure 9. Side view of the reference case flare at flow condition A (Table 1) showing 50% flame intermittency contour and flare trajectory. The trajectory for single jet flame in a cross flow (correlation of Brzustowski, 1976, 1980) at the conditions used in this work is shown originating from the lead jet in the flare system. Zone 1 depicts the jet and cross-flow interaction region; Zone 2 depicts the buoyant region of the flare flame. 10 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 support this observation with a strong jet/cross flow interaction region in Zone 1 followed by the buoyant region in Zone 2. In addition to the change in flame trajectory due to the impact of buoyancy, there is an expectation that the side-view trajectories should change with firing rate when the cross flow momentum flux is constant. An increase in the angle of declination from the vertical with decreasing firing rate was indeed evident for the reference case and ring attachments as shown by the side-view profiles (50% intermittency) shown in Fig. 10. Estimates of the side-view trajectory for each flare system at different firing rates are also shown in Fig. 11. The reference case and ring attachments showed consistent behaviour with a decrease in slope as the firing rate was reduced. The results for the cone and precessing jet attachments exhibited more complex behaviour with modest changes in slope and a shift in the effective or apparent origin for the flame source. These two cases also exhibited more bulbous flame shape behaviour in the near field, Fig. 7, and this effect may lead to a more complex aerodynamic resistance to the cross­ flow. Gas temperature measurements were obtained along the flame trajectory for most of the conditions noted in the conditions in Table 1. An example of these results for test condition A are shown in Fig. 12 where the gas temperature is plotted as a function of y = £/L, where £ is the flame trajectory measured from the top and centre of the flare and L is the curvilinear flame length estimated from the flame trajectory data. These results show that there is a maximum gas temperature in the region y « 0.2 - 0.4. The solid line in this graph is a second order polynomial fit for all the data at this flow condition. The data for the reference case and cone attachments are randomly scattered about this curve. On the other hand, the results for the ring attachments are typically higher while the precessing jet data fall below this curve - these observations can be related to lower NOx emissions described later for the precessing jets. The maximum centreline temperature of the single, free-jet vertical propane diffusion flame is located at about 60% of the visible flame length (Becker and Yamazaki, 1978). Under the influence of a cross flow, the temperature peak observed in the present work moves closer to the nozzle - Botros and Brzustowski (1978) also observed this behaviour with the turbulent diffusion flame in a cross flow. Temperature profiles at other firing rates followed a similar behaviour with slightly lower temperatures observed at lower firing rates (Pagot, 2002). Gas composition measurements were obtained for O 2 , CO, CH 4 , and NOx at test conditions A-D, Table 1. Data obtained along the flare trajectory at flow condition A is shown in Fig. 13. The O 2 composition typically exhibited minimum values and the CO maximum values (both about 2 - 8 %, by vol.) in the range, y « 0.2 - 0.5, corresponding to the peak temperatures in these flames. The CH4 and CO compositions exhibited wide variability for y < 0.5 and approached the instrument detectability limit near y « 0 .8 . The NOx emissions were in the range, 30 - 60 ppm, for the reference case with peak values near the highest gas temperature region, y « 0.3-0.4. The emissions for the cone and ring attachments were in the range 30-35 ppm, while the precessing jets showed the lowest values of about 22-24 ppm near y = 0.3-0.4. The precessing jets also exhibited the lower gas temperatures at these conditions. The total flame radiation was measured from a single side-view fixed position, 2.92 m from the centre plane of the flare model. These results were fitted to a linear model based on the assumption that the total radiation flux was a linear function of the firing rate as shown in Fig. 14. The reference case and the ring attachments had the lowest radiation flux while the cone attachments and the precessing jets had the highest heat flux. 11 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Reference case Ring attachments x, m m x, m m Figure 10. Side-view 50% flame intermittency contours for the reference case (left graph) and ring attachment (right graph) flares at different firing rates, A - D. 600 E ^ 400 " I------1" o A _ □ B A C V D O E i- i- i- ©■ e 4£V (EV <BV - 1- - * _ a ° zs o i- - 1- i- i- A VA ° B A C * S D A 8 _ . " N 200 i- s a £ -1 -20 0 ■ 0 200 400 600 800 1000 A B A C D £ 600 E □ o □ Q) -20 0 0 F \ 200 o □ o □ o □ A ° □ AO C t^ ° 400 600 800 1000 V Q jf^ v cS7 o 'O 200 Qg O Ct£sX> 0 0 | Precessing jets _l__ I__ I__ I__ -20 0 0 200 400 600 800 1000 x, c m Figure 11. 400 Qjfc 400 N . d 1000 -i------1----- i----- 1----- r O A o _ 800 _ □ B ct: A C Cb a V D CbA 600 " o E 1000 800 . a _i - Reference case ■ i i i - .& . Cone attachments 1 1 1 1 i 0 1 _ 0 9 ❖ _ - i- Ring attachments ....................... -20 0 0 200 400 600 800 1000 x, c m Flame trajectories (side view) for firing rates A --- E with the different nozzle attachments. 12 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 1800 _ ^ v V 'i V y^ » • , + /- 2 std. dev. A gta 1600 1400 O • R eference c a se □ C o n e a tta c h m e n ts a P re c e s s in g jets v R in g a tta c h m e n ts 1200 1000 0 .0 0 .2 0.4 0.6 0.8 1.0 y Figure 12. Gas temperature measurements on the flare trajectory for test condition A. T 20 0s 16 CD O 12 E c\, O 8 T □ Cone attachments A Precessing jets Ring attachments 8 O □ V n □ o' a) o E VV V V o v rx lX & ] V v o ^ & O 0 8 80 Reference □ Cone attachments A Precessing jets Ring attachments 6 60 _ E V ____ S o o < V Ci 4 o 97 V ODO O O r> i__ o X H C 2 40 Reference □ Cone attachments A Precessing jets Ring attachments (X7 ^ V l l _ 2 _____ vO V A 20 I 13. o V n vv 2 o F ig u re O W 4 0 0 0.0 ° V 6 O' C ■Sj y _V 4 0s CD o E X X Reference □ Cone attachments A Precessing jets Ring attachments O Reference n pb5%v vv 0.2 0.4 0.6 0. 1.0 0 0.0 0.2 0.4 0.6 0.8 1.0 y y Gas compositions, O2, CO, CH4, and NOx, as a function of dimensionless downstream position along the flare trajectory for flow condition A. 13 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 F l a r e f ir in g r a t e , k W Figure 14. Total radiative heat flux as a function of firing rate for different flare configurations. Conclusions The hot model studies involved flow visualization, gas temperature, gas composition and radiative heat flux measurements for the different nozzle configurations at five different firing rates. CCD imaging techniques were used to investigate the flame trajectory and structure. The flames were characterized by an initial jet/cross-flow interaction region followed by a transition to a buoyant regime further downstream along the flame trajectory. At the highest firing rate, the gas temperatures reached maximum values at about 30% of the flame length and the lowest gas temperature was observed for the flare model with precessing jets. NOx concentrations in the range 30-60 ppm were observed for the reference case nozzle while the precessing jet model gave the lowest emission rate in the 22-24 ppm range. The radiative heat flux from the flames was linearly dependent on the flare-firing rate and a lower flux was observed for the reference case and ring attachment configurations. References Bandaru, R.V. and Turns, S., "Turbulent Jet Flames in a Crossflow: Effects of Some Jet, Crossflow, and Pilot-Flame Parameters on Emissions", Combustion and Flame, 121, 137-151, 2000. 14 A F R C - J F R C 2 0 0 4 J o i n t I n t e r n a t i o n a l C o m b u s ti o n S y m p o s iu m , M a u i , H a w a ii, O c t. 1 0 - 1 3 , 2 0 0 4 Becker, H.A. and Yamazaki, S., "Entrainment, Momentum Flux and Temperature in Vertical Free Turbulent Diffusion Flames", Combustion and Flame, 33, 123-149 (1978). Botros, P.E. and Brzustowski, T.A., "An Experimental and Theoretical Study of the Turbulent Flame in Cross-Flow", 17th Symposium (International) on Combustion, p. 389-398, University of L e e d s(1978) Brzustowski T. A., "Flaring in the Energy Industry", Prog. Energy Combust. Sci.,, 2, 129-141 (1976). Brzustowski T. A., "Mixing and Chemical Reactions in Industrial Flares and Their Model", PCH: Physicochemical Hydrodynamics, 1, 27-40 (1980). Kalghatgi, G.T., "Blow-Out Stability of Gaseous Jet Diffusion Flames. Part II: Effect of Cross Wind", Combustion Science and Technology, 26, 241-244, 1981. Menon, R., "Characteristics of Multiple Diffusion Flames in Still Air and Cross Wind", M.Sc. thesis, The University of Okalahoma (1984). Menon, R. and Gollahalli, S.R., Combustion Characteristics of Interacting Multiple Jets in Cross Flow", Combustion Science and Technology, 60, 375-389 (1988). Pagot, P.R., "Cold and Hot Model Investigation of Flow and Mixing in a Multi-Jet Flare", Ph.D. thesis, Department of Mechanical Engineering, Queen's University, Kingston, ON Canada K7L 3N6 (2002). Pagot, P.R., A. Sobiesiak and E.W. Grandmaison, "Cold and Hot Model Investigation of Flow and Mixing in a Multi-Jet Flare", Proceedings of Combustion Canada 2003, Sept. 22-24, 2003, Vancouver, BC, Canada (2003). Turnbull, W.N.O., "Experimental Investigation of Sample Probe Effects on Gases Drawn from Flame Zones", M.Sc. thesis, Department of Mechanical Engineering, Queen's University, Kingston, ON Canada K7L 3N6 (1995). Acknowledgement This work was supported by grants from the Natural Sciences and Engineering Research Council of Canada (NSERC). One of the authors (PRP) wishes to acknowledge the support of Petrobras, Petroleo Brasileiro S.A. 15