,//V --, --: _j NASA Technical Memorandum 104492 AIAA-91-1379 Improved Thermodynamic Modeling of the No-Vent Fill Process and Correlation With Experimental Data (NAS A- T'_- I C)4¢,92 ) [MPRBV,_ -) THe i_*,:_r'Y NA -_I C N92-II129 M:-IjEL [',, ; _F THe..NQ-VC,':T _- [LL PI'?:LLSS *_;',_ C_,RRLLATIO_q _[TH _w.P_:Rl'a!"aT-_,L -:ATA (NASA) Unclas l,a r) CSCL 21L+ _3/2o 0047542 William J. Taylor and David J. Chato Lewis Research Center Cleveland, Ohio Prepared for the 26th Thermophysics Conference sponsored by the American Institute of Aeronautics and Astronautics Honolulu, Hawaii, June 24-26, 1991 IMPROVED THERMODYNAMIC MODELING OF THE NO-VENT FILL PROCESS AND CORRELATION WITH EXPERIMENTAL DATA W.J. Taylor* and D.J. Chato* National Aeronautics and Space Administration Lewis Research Center Cleveland, Ohio 44135 Abstract Convection heat transfer coefficient The United States" plans to establish a permanent manned presence M mass in space and to explore the Solar System have created the need to effi- m mass flow rate ciently handle large quantities of subcritical cryogenic fluids, partic- P pressure ularly propellants such as liquid hydrogen and liquid oxygen, in low- Q heat input to zero-gravity environments. One of the key technologies to be developed q heat flux for fluid handling is the ability to transfer the cryogens between storage T temperature and spacecraft tanks. The no-vent fill method has been identified as t time one way to perform this transfer. In order to understand how to apply this U total internal energy method, a model of the no-vent fill process is being developed and corre- u specific internal energy lated with experimental data. The verified models then can be used to W work design and analyze configurations for tankage and subcritical fluid depots. W power This paper discusses the development of an improved macroscopic thermody- Greek Symbols: namic model of the no-vent fill pro- cess and correlates the analytical p density results from the computer program implementation of the model with spray cooling efficiency experimental results for two differ- ent test tanks at NASA Lewis Research Subscripts: Center. cond condensate Nomenclature gas ullage gas A area in inlet C specific heat constant volume inf interface V h specific enthalpy ig liquid/gas *Aerospace Engineer, Member AIAA. fig liquid liquid spray and the ac=umulating liquid. If the condensation of the par parasitic ullage gas occurs at a sufficiently high rate, no ullage compression sat saturation occurs; the pressure will, there- fore, decrease after an initial sgas saturated gas pressure rise. Plots of the tank pressure versus time for two typical tank tank hydrogen no-vent fills from Ref. 6 are shown in Figs. I and 2. Figure 2 wall tank wall depicts a fill in which the tank pressure decreases after the initial pressure rise. No-Vent Fill Process The no-vent fill process was Macroscopic Thermodynamic Model identified early on I as a key tech- nology for the handling of cryogenic Chato 4"6 has previously reported liquids in a low- to zero-gravity on the development of a macroscopic environment. The present procedure thermodynamic model for the no-vent for the no-vent fill which incorpo- fill process. This paper describes rates tank chilldown via charge, the latest revisions made in this hold, vent cycles was defined in model as it i8 implemented in the Refs. 2 and 3. The no-vent fill pro- NVFILL program at NASA Lewis Research cess allows a propellant or storage Center. Chato'8 original model tank to be filled or replenished divided the no-vent fill process into without venting or requiring the tank two steps. In the first step, the contents to be positioned via a wall is chilled from its initial tar- settling acceleration. Briefly the get temperature to the temperature of no-vent fill process proceeds as out- the incoming the liquid. In the sec- lined below. The tank wall tempera- ond step, the bulk liquid accumulates ture, and thus the tank wall energy and the ullage vapor i8 compressed content, is reduced to an initial and condensed. The key assumption target value via a series of charge, made by Chato in analyzing the wall hold, and vent cycles. The target chilldown step of the no-vent fill temperature is selected to allow the process was that all of the incoming transfer process to the tank to liquid was vaporized until the wall achieve 95 percent liquid fill by temperature matches the temperature volume without exceeding a specified of the incoming liquid. Based on maximum pressure. Once the tank wall this assumption and the first law of temperature has been reduced to the thermodynamics, the governing equa- tions for the wall chilldown were as target temperature, the fill phase of 4 the process can begin. During the follows: fill phase, liquld is continuously injected into the tank until the desired fill level is achieved. Ini- -Mwa Ii dt tially, vapor is generated as the incoming liquid cools the walls fur- ther. Liquid also begins to accumu- dM_as (2 ) = mln late in the tank during this time. dt The accumulated bulk liquid com- presses the vapor in the ullage space. Simultaneously, the vapor in (3) dUga" = _±n(hsga- ugu) the ullage is condensing due to heat Mgas dt and mass transfer to the incoming 2 The basic equations presented by the wall is completely cooled and for Chato4 for the liquid accumulation, the parasitic heating of the accumu- vapor compression and condensation lated bulk liquid_ as initially these step are presented below: were thought to be largest sources of error in the model. The parasitic dMg., . (4) heating of the bulk liquid is defined m -mcond as the heat leaks to the liquid in dt the tank due to either the experimen- tal setup or the ambient environment. In the CCL-7 test rig, am described d(MgalUgll ) (5) in Ref. 9, these heat leaks are dt " I_c°ndhgas + _lg exemplified by the axlal conduction of heat along the tank walls from the dMliq . . (6 ) top lid mounting flange and the heat w Sin ÷ mcond radiated from the outer wall of the dt vacuum jacket to the inner wall where the heat is conducted into the bulk duli q dM11q fluid. These modifications to the Mllq dt + uliq dt + qlnf (7) model required Eq. (1} to be changed as follows: ÷ mtnhin + mcondhltq "Wlg C_in f -Mwall dt (8) -- 1 0 dt where A is the spray cooling effi- ciency. The spray cooling efficiency qlnf I mcond(hga, - hllq) (9) is defined as the fraction of the incoming liquid mass that is vapor- ized through contact with the tank walls during any time increment. mcond 1 Thus, the value of the spray cooling hgas - hll q) efficiency is between 0 and i. The partial vaporization of the incoming P gas/. fluid necessitated the development of Wlg. _--_in + mcond) (11) a new equation for the liquid mass Pnq accumulation to replace Eq. (6). The convection heat transfer coefficient in Eq. (i0) governs the (13) dMllq - (I - _in + " heat transfer between the liquid d---t-- mc°nd spray droplets and the ullage vapor and is calculated from a correlation Additionally, Eq. (14) replaced 7 presented by Brown. The complete Eqs. (2) and (4) in the original methodology employed in using this model, with _ going to 0.0 when the correlation is presented in Ref. 5. wall has been chilled to the incoming liquid temperature. Comparisons of the results from this model with test data, presented dMgas (14 ) in Ref. 6, indicated the need to dt 1 Xml. -- mco.d change the model for modeling the top spray fill configurations tested at Equation (7) was modified to NASA Lewis in the Cryogenic Compo- account for the parasitic heating of nents Laboratory Site 7 (CCL-7) test the accumui_zed bulk liquid resulted rig. The basic modifications allowed as shown below: for the accumulation of liquid before 3 step, the remainlngequatlons, duli q dMliq Eqs. (4) to (11), are solved for each M_iq dt 4. Ull q -- dt ÷ qinf (15) time increment during the second step of the process. + q_r + mtnhin + mcondhliq • _lg The logic flow of the program had to be modified to account for the Equation 5, which performs the energy balance for the ullage vapor, partial vaporization of the incoming was modified to the form shown in liquid. Rather than having a two step procedure with the sequential Zq. (16). performance of the two steps, now the dMgaaUga, dMg.s wall chilldown and the liquid accumu- " %.-_- - _co.eh... lation calculations, Eqs. (8) to (10) (16) and Eqs. (12) to (17), are performed in each time step untll the wall is + Wig + )_l_i.(h gn - ug,.) chilled down. Once the wall chill- down is complete, only Eqs. (8) to Equation (11) was also revised (10) and (13) to (17), with A equal to account for the partial vapori- to 0, are evaluated in each time zation of the incoming fluid. step. The thermodynamic properties of the fluid are evaluated as ) (17) required in each time step via calls to the GASP program. B The latest version of the NVFILL program (5.4) Once this basic model encom- is written in Fortran and runs on IBM passing Eqs. (8) to (i0) and (12) to PC compatible microcomputers. (17) was developed, it was imple- mented in the NVFILL computer program. Empirical Data Receiver Tank Configuration NVFILL Computer Program In the CCL-7 test rig, which is The NVFILL program was written described in detail in Ref. 9, the by Chato to implement the macroscopic liquid is thermally conditioned in thermodynamic model. It uses an the supply tank prior to performing explicit time-marching solution of the transfer. Two different size the basic thermodynamic equations receiver tanks are used. The tank volumes are 1.24 ft 3 for the small presented in the previous section. Both the original and the revised receiver tank and 5.0 ft 3 for the implementations of the model in the large receiver tank. Both tanks have NVFILL program assume that all of the cylindrical body sections with an heat and mass transfer at the liquid elliptical bottom dome and a lid. vapor interface occurs between the Figure 3 is a schematic of the tank ullage vapor and the incoming liquid geometries. Both receiver tanks have spray droplets. The heat and mass a conical spray nozzle mounted in the transfer that occurs at the free sur- top dome which sprays downward. The face of the accumulated bulk liquid spray half angle is 60 °. The tanks is not considered. The original are both commercial vacuum jacketed version of the code followed the two- dewars constructed of 304 stainless step procedure delineated in the pre- steel. vious section. During the first step of the fill process, Eqs. (I) to (3) are solved for each time increment. Upon the completion of the first 4 CCL-7 Large Receiver Tank Analytical Results The results from the initial CCL-7 Large Receiver Tank series of tests, during which 9 no-vent fill tests were performed, The results of the nine large are presented in Ref. 6. This test- receiver tank tests were analyzed ing demonstrated the feasibility of using the revised version of the the no-vent fill process with both NVFILL program. The results for 4 of liquid hydrogen and liquid nitrogen. these analyses (2 nitrogen cases and The graphs of the tank pressure his- 2 hydrogen cases} ere presented in tories for 4 of these tests along Figs. 4 to 7. Only 4 cases were with the original and revised analy- selected for presentation in this tical modeling results are presented paper, as they represent the most in Figs. 4 to 7. All of the test significant results obtained with the parameters and conditions are pre- revised model. Both of the nitrogen sented in detail in Ref. 6. test cases, 6 N2 and N3, (Figs. 4 and 5) exhibit a large initial pres- This initial testing was con- sure rise followed by a pressure ducted to verify the operation of the decay over the remaining duration of facility and the data collection sys- the test. The two hydrogen cases tem. More rigorous experimentation presented, H2 and H4, (Figs. 6 and 7) with this tank is currently being were the only successful tests with conducted. hydrogen in the initial testing at CCL-7. The remaining test cases were of such short duration, they were not CCL-7 Small Receiver Tank included here. The model results of Ref. 6 are included in Figs. 4 to 7 A series of 18 liquid hydrogen for comparison purposes. These new no-vent fill tests with a top spray analyses also provided the oppor- fill configuration were performed tunity to reexamine the experimental using the 1.24 ft 3 dewar. The results and, based on this examina- primary variables in this test series tion, revise two of the analysis were the liquid inlet mass flow rate, input parameters. which ranged from 0.3 to 3.8 ib /min, and the initial tank wall temper- The initial average tank wall ature, which ranged from 55.3 to temperature had to be changed in most 167 OR. The liquid inlet tempera- of the analyses. The original model tures ranged from 32.8 to 35.9 OR. inputs for the tank wall temperature While not all of the tests were suc- were based on a simple volume cessful in achieving the target fill weighted arithmetic average of the level of 95 percent, half of the tank wall temperatures. These aver- tests were completed with final fill age wall temperatures were calculated levels at or above 94 percent. from a discrete nodalization of the However, neither of the two tests tank walls which was based on the conducted at the low inlet mass flow location of temperature sensors rates (approximately 0.3 Ib /min) mounted to the wall. The revised were successful, and only three of temperature calculations use the same the eight tests with inlet mass flow nodalization but also account for the rates of approximately 0.7 ib /min temperature dependency of the spe- were successful. Table 1 presemnts a cific heat of the wall material. summary of the test parameters for This approach was necessitated by the each run. A more complete discussion presence of large temperature gradi- of all the tests and the results ents axially along the tank wall at obtained can be found in Ref. i0. the start of the tests and the large 5 variation in the specific heat of The revised version of the NVFILL stainless steel over the observed program requires two additional temperatures. The resulting initial inputs, one for the parasitic heat average wall temperatures are higher leak and the second for the spray than those calculated assuming a con- cooling efficiency. All of the ana- stant specific heat for the tank wall lytical results presented in thl8 and provide a better estimate of the paper used a value of 0.00 for the inltlal tank wall energy content. parasitic heating due to the small magnitude of the heat leak and the The liquid inlet temperature was short duration of the tests. Boiloff the other parameter to be revised. tests conducted at CCL-7 with the two Based on the temperature difference receiver tanks, filled to various measured between the supply tank levels with H2, indicated the para- fluid and the liquid temperature sitic heat leaks to the tank contents measured at the flow meter, a dis- are on the order of 18 Btu/hr and tance of approximately 20 ft, the 30 Btu/hr for the small and large liquid inlet temperatures for some of receiver tanks respectively. Wlth the hydrogen tests were increased test durations ranging from approxi- approximately 1 to 2 "R. These tem- mately 1.5 to 7.0 min for the small perature increases account for the receiver tank, the total parasitic heat leaks to the incoming liquid heat leak to the tank has a maximum hydrogen due to the distance (approx- value of 2.1 Btu. Similarly the imately 15 ft) from the liquid tem- total parasitic heat leak for the perature sensor to the tank inlet. large receiver tank tests is approxi- The temperature difference between mately 7.5 Btu. These small heat the supply tank and the flow meter leaks are negliglble in comparison was inversely proportional to the with the energy of the incoming liq- liquid mass flow rate, thus for the uid, and thus do not impact the final high flow rate cases the temperature pressure in the receiver tank. difference became negliglble. The nitrogen test cases were not affected Analysis of the tank and spray as the liquid mass flow rates are 3 nozzle geometry determined that the to i0 times higher than those used in maximum percentage of the incoming the hydrogen test cases. The input spray that could strike the tank parameters for the two sets of analy- sidewalls was 63 percent after liquld ses are compared in Table 2. accumulated in the elliptical bottom dome section of the receiver tanks. One other parameter to be con- Assuming a constant Inlet mass flow sidered in these analyses is the rate, the percentage of the incoming tank-mass-to-volume ratio. The flow that strikes the sidewalls results presented in Ref. 6, used a decreases to 0 percent over the dura- reduced tank-mass-to-volume ratio of tion of the fill. Thus the time- 2.1 ib /ft. This effectively reduced averaged value is 32 percent. Based the energy content of the tank wall on the information in Ref. 11, the at the start of the fill process. As spray cooling efficiency for droplet discussed in Ref. 6, this is an sprays varies between 3 and 20 per- attempt to compensate for the fact cent. Multiplying these efficiencies that the tank lid assembly does not by the percentage of the incoming cool down significantly during the liquid spray mass that strikes the fill tests conducted with high inlet side walls, yields an overall average mass flow rates, thus the energy in spray cooling efficiency that ranges the lid is not transferred to the between 0.1 and 6.4 percent for our fluid in the tank. The revised anal- test configurations. A value (5 per- yses also used this reduced value. cent) near the top of this range was selected for the analyses presented CCL-7 Small Receiver Tank in this paper, as the wall chilldown was accomplished fairly rapidly All 18 of the hydrogen top spray during the tests. The model imple- configuration no-vent fill tests con- mentation assumes the spray cooling ducted at CCL-7 using the small efficiency remains constant until the receiver tank were analyzed with the wall chilldown is complete. In real- revised NVFILL program. As was the ity, the spray cooling efficiency case for the large receiver tank will increase as the wall is chilled tests, the liquid inlet temperatures and the temperature difference for the analyses were estimated based between the wall and the liquid spray on the temperature difference of the droplet is reduced. fluid measured between the supply tank and the flowmeter. The inlet The plots of the data and the mass flow rates for these tests had analysis results for the two nitrogen to be calculated from the tank fill cases are presented in Figs. 4 and 5. level versus time data due to a fail- The partial evaporation of the incom- ure in the instrumentation associated ing spray provides a much more accu- with the flowmeter. Table 3 sum- rate representation of the initial marizes the variable input parameters pressure spike for the hot wall cases for each case. when compared with the analytical results presented in Ref. 6; however, The tank-mass-to-volume ratio analysis of the bulk liquid accumula- used in all these analyses was tion, vapor compression and condensa- 6.96 ibm/ft 3. The analytical results tion step still shows a discrepancy for the 15 of the 18 test cases along between the experimental and the with the empirical data are presented analytical results. This difference in Figs. 8 to 22. The results for can be attributed to the model the tests identified as 9088A, 9088E implementation, which forces the and 9094H were not included due to tank's contents to thermodynamic anomalies present in the data. Each equilibrium instantaneously when the of the figures depicts the tank pres- conditions in the tank are such that sure versus time for the respective bulk boiling occurs in the accumu- test run. The results for the low lated liquid. flow rate (0.3 lbm/min ) cases (9072A and 9075A) are shown in Figs. 8 The five hydrogen test cases and 9. The empirical data clearly discussed in Ref. 6 were also ana- indicate that there is some error in lyzed. The empirical data and the the test instrumentation particularly results from the modified NVFILL with regard to the liquid level in program for two of these tests are the tank at the initiation of the presented in Figs. 6 and 7. The ana- test and the liquid inflow rate. lytical results for these cases were Despite these errors, the program affected most by the changes in the results track the pressure history of input parameters for the initial wall the tank with a maximum difference temperatures and the liquid inlet from the experimental data of 5 psia. temperature. These changes produced It also appears that the program better correlation between the ana- slightly over predicts the tank pres- lytical and the test results, espe- sure at a given fill level. cially with regard to the magnitude of the initial pressure rise for the The six moderate flow rate (0.6 H4 case and the final tank pressure to 0.8 lb /min) cases (9072B, 9075B, for the H2 case. The correlation 9075D, 9080A, 9081A, and 9081E) were between the test data and the analy- also analyzed. As shown in Figs. 10 sis results for the different tests to 15, excellent results were was improved in all cases. obtained for all of the cases, the predicted pressures differing from the lid assembly was cooled down at the empirical data by less than the beginning of the test and thus had little influence on the tank wall 3 peia. The results show the impor- tance of the liquid inlet temperature energy content and thereby the energy to be transferred to the incoming in the analysis as the incoming liq- uid energy dominates the final condi- liquid. The major differences tion of the fluid in the tank. between the test data and the analyt- Therefore, in order to obtain the ical results for the high flow rate analytical results shown in Figs. I0 cases are in the predicted initial to 15, the inlet temperatures were pressure rlme rates; indicating that either the initial wall temperatures adjusted upward by approximately 1 °R, again accounting for heat leaks and/or the specified spray cooling to the incoming liquid. With regard efficiency used in the analytical to the initial pressure rise, the model were too high. initial temperature of the tank wall and the spray cooling efficiency strongly influence the analytical Summary results, particularly for low inlet mass flow rates. A simple macroscopic thermody- namic model of the no-vent fill pro- In the 8 high inlet flow rate cess developed by Chato was revised cases (9093A, 9093B, 9093C, 9093D, to account for the partial vaporiza- 9094A, 9094B, 9094C, and 9094H), the tion of the incoming liquid spray and liquid inlet mass flow rates ranged for the parasitic heat leak to the from 2.0 to 3.8 ib /min. Initial accumulated bulk liquid. This analyses, not presented herein, of revised model was implemented in a all but two of these test cases did new version of the NVFILL computer not correlate well with the experi- program and the results were compared mental data. The analytical results with empirical data for two receiver exhibited an initial pressure spike tanks tes£ed at NASA Lewis. The mod- of shorter duration and larger magni- ifications improved the correlation tude than the test data. Examination between analytical and experimental of the test data showed that the lid results for both hydrogen and nitro- assembly was not being cooled during gen. However, based on the results the fill process. This phenomenon of multiple runs of the model, the was also observed in the large largest improvements in the analyti- receiver tank 6 tests. Instead of cal results were due to the more exact calculation of the initial wall reducing the tank mass-to-volume ratio, as was done previously, new temperatures and increasing the estimates of the tank wall tempera- incoming liquid temperature to ture were made by accounting for the account for the parasitic heat leaks temperature change undergone by the to the liquid in the lines between lid and thereby the energy removed. the flowmeter and the receiver tank. As can be seen in Figs. 16 and 18 to The analytical results obtained with 21, these reduced initial wall tem- the revised inputs, in conjunction with the revised model correlate with peratures enabled the model to repli- cate the experimental data quite both the process time line and the well. For the two cases, 9093B and tank pressure versus the volumetric 9094C which are plotted in Figs. 17 liquid fill percentage. With suffi- and 22, where the model did predict cient attention to the process the behavior of the receiver tank inputs, results were obtained that without having to reducing the aver- differ by less that 5 psia from the age initial wall temperature, an experimental data. examination of the test data revealed 8