DISCHARGE OF NON-REACTIVE FLUIDS FROM VESSELS1

Transcription

1 Brazilian Journal of Cheical Engineering ISSN Printed in Brazil Vol. 34, No. 04, pp , October Deceber, 2017 dx.doi.org/ / s DISCHARGE OF NON-REACTIVE FLUIDS FROM VESSELS1 M. Castier1,*, A. Basha, R. Kanes1 and L.N. Véchot1 1Mary Kay O Connor Process Safety Center. Cheical Engineering Progra, Texas A&M University at Qatar P.O. Box 23874, Doha - Qatar *Corresponding author: E-ail arcelo.castier@qatar.tau.edu; Phone: ; Fax: (Subitted: March 7, 2016; Revised: June 29, 2016; Accepted: August 7, 2016) Abstract This paper presents siulations of discharges fro pressure vessels that consistently account for non-ideal fluid behavior in all the required therodynaic properties and individually considers all the cheical coponents present. The underlying assuption is that phase equilibriu occurs instantaneously inside the vessel and, thus, the dynaics of the fluid in the vessel coprises a sequence of equilibriu states. The forulation leads to a syste of differential-algebraic equations in which the coponent ass balances and the energy balance are ordinary differential equations. The algebraic equations account for the phase equilibriu conditions inside the vessel and at the discharge point, and for sound speed calculations. The siulator allows detailed predictions of the condition inside the vessel and at the discharge point as a function of tie, including the flow rate and coposition of the discharge. The paper presents conceptual applications of the siulator to predict the effect of leaks fro vessels containing ixtures of light gases and/or hydrocarbons and coparisons to experiental data. Keywords: Relief, venting, tank, valve, equation of state. INTRODUCTION Discharges of hazardous cheicals can bring ab serious consequences to populations living close to industrial units, in addition to financial losses. In soe cases, the discharges are accidental, as in the case of leaks. In others, as in operations for pressure relief, they are intentional and ai to prevent further daage, such as the explosion of pressure vessels. In both cases, it is iportant to predict the flow and coposition of the aterial discharged to the environent in order to plan easures of containent and reediation. Due to the growing iportance of this practical proble, several progras for the siulation of leaks and venting fro pressure vessels exist (Woodward and Mudan, 1990; Bendiksen et al., 1991; Haque et al., 1992a, 1992b; Richardson and Saville, 1996; Mahgerefteh and Wong, 1999; Cuber, 2001; Witlox and Bowen, 2002; Renfro et al., 2014; D Alessandro et al., 2015). In order to siplify the forulation and shorten their execution tie, soe of the use approxiations, such as that of ideal gas behavior, that increase the uncertainty of their predictions. Other papers account for deviations fro ideal gas behavior as, for exaple, those of Haque et al. (1992a, 1992b) and Richardson and Saville (1996), who used a proprietary coputer package for physical properties called PREPROP, and of Mahgerefteh and Wong (1999), who used cubic equations of state. Another approxiation occasionally used is that the cheical coponents can be luped into a single pseudo-coponent; however, the coposition of the vented fluid and of the fluid inside the vessel can vary significantly during a siulated period. These references focus on the so-called source ter, which refers to the behavior of the fluid up to the point 1 A preliinary version of this work was presented at the Terceras Jornadas Argentinas de Seguridad de Procesos, Buenos Aires, * To who correspondence should be addressed

2 1150 M. Castier, A. Basha, R. Kanes and L.N. Véchot of its release fro the vessel. The behavior of the fluid beyond the release point can also be rather coplex, as exeplified by the work of Birch et al. (1987), who present correlations to predict how the velocity of the fluid decays as it flows into the atosphere. Other exaples are the work of Winters (2009) and Houf and Winters (2013), who odel the flow of hydrogen into abient air right after leaks fro tanks that store liquid hydrogen at very low teperatures. An interesting aspect of these publications is the rigorous odeling of physical properties using the REFPROP (Leon et al., 2007) therodynaics suite of flash algoriths. Renfro et al. (2014) describe the features of the Blowdown Utility progra for the siulation of discharges fro horizontal or vertical vessels of several geoetries. A non-equilibriu vessel odel is adopted and the siulator allows the choice of several odels for the evaluation of the physical properties of the fluid. Their approach allows the siulation of vessels that contain three fluid phases. However, the article provides no details ab the atheatical forulation and an undisclosed expression is used to calculate the flow rate of the leaking fluid. D Alessandro et al. (2015) describe a siulator called VBsi (Vessel Blowdown Siulator) that uses global phase stability analysis and iniization of the Gibbs energy to find the conditions inside the vessel. VBsi uses the concept of partial phase equilibriu (Speranza and Terenzi, 2005), according to which the liquid and vapor phases have the sae pressure, but not the sae teperature, because of their different heat transfer coefficients, which are calculated by the progra. In this approach, the liquid phase is in equilibriu with a bubble phase while the vapor phase is in equilibriu with a droplet phase. Different expressions are adopted for the flow through the vessel s orifice, depending on whether the discharge is of a liquid or vapor phase. However, their approach seeingly disregards the possibility of phase change at the discharge point. This paper introduces a siulator for discharges fro pressure vessels that accounts for deviations fro ideal gas behavior in all the required therodynaic properties and individually considers all the cheical coponents present. The forulation follows the so-called hoogenous equilibriu assuption, according to which equilibriu is achieved instantaneously and, thus, the dynaics of the fluid in the vessel coprises a sequence of equilibriu states. The underlying atheatical proble is a syste of differential-algebraic equations (DAE) in which the coponent ass balances and the energy balance are ordinary differential equations. The differential energy balance includes the internal energy of the fluid and of the vessel walls. The algebraic equations account for the phase equilibriu conditions inside the vessel and at the discharge point. Due to the size and coplexity of the odel, nuerical solution is necessary. At each tie step during the integration of the ordinary differential equations of the odel, the values of the internal energy and coponent aounts inside the vessel becoe known. Since the vessel s volue is known and fixed, the state of the syste can be found by axiizing the fluid s entropy (Callen, 1985). To copute the discharge conditions, the assuption is that the region inside the vessel close to the discharge point acts as an adiabatic converging nozzle whose operation is isentropic. This leads to an unusual equilibriu proble with specification of fluid entropy and stagnation enthalpy (enthalpy plus kinetic energy). The odel accounts for the possibility of non-choked and choked flow at the discharge point. In the latter, the fluid speed is equal to the local therodynaic sound speed. Therefore, a procedure to copute the therodynaic sound speed in systes with ultiple phases is also part of the forulation. The siulator allows predictions of the conditions inside the vessel and at the discharge point as a function of tie, including the flow rate and coposition of the discharge, which are crucial pieces of inforation when planning for eergencies and studying containent alternatives. The paper shows applications of the siulator to predict the effect of leaks fro vessels containing light gases and hydrocarbon ixtures. ASSUMPTIONS A set of assuptions underlies the developent of the atheatical odel of leak and vent operations fro pressure vessels. A central one is that the fluid inside the vessel is in therodynaic equilibriu at all ties, and that properties such as teperature, pressure, and cheical potential of each coponent are the sae everywhere within the vessel. When the ass and the specific heat capacity of the vessel wall are considered in the siulation, it is assued that the wall and fluid teperatures are equal at any tie. It is also assued that the volue of the vessel s wall reains constant, that is, wall expansions or contractions because of teperature changes are neglected. In the absence of cheical reactions, as is the case considered here, this eans that deterining the state of the fluid within the vessel requires a phase equilibriu coputation. Another key assuption is related to finding the leak or vent flow rate, for which it is necessary to odel the region in the vessel near the orifice. As usual in siilar siulations, the assuption is that this region is a hypothetical converging nozzle that operates adiabatically and reversibly. As such, the flow in the converging nozzle is taken as isentropic. During leaks fro vessels that contain condensed phases, the violent boiling caused by depressurization can swell the liquid phase, i.e., its level ay rise because of bubble foration within the liquid. This effect is neglected, under the assuption that there is instantaneous disengageent of the phases present. Brazilian Journal of Cheical Engineering

3 Discharge of Non-Reactive Fluids fro Vessels 1151 With these assuptions, the atheatical odel consists of a set of differential-algebraic equations (DAE). The ass and energy balances are ordinary differential equations (ODE). The algebraic equations allow the deterination of the state of the fluid inside the vessel and at the exit point at each tie step of the nuerical ODE integration. Furtherore, to deterine the leak or vent flow rate, it is necessary to find several fluid properties at the fluid s exit point fro the vessel, naely, its teperature, pressure, nuber of phases with their aounts and copositions, and therodynaic sound speed. The sound speed is relevant because it is the axiu achievable fluid speed at the exit of a converging nozzle. Flows at sonic speed are referred to as choked and the pressure at the exit point is larger than the backpressure, which is the environental pressure away fro the vessel. Flows below sonic speed are referred to as non-choked and the exit pressure is equal to the backpressure. FORMULATION The forulation follows the recent work of Kanes et al. (2016). We consider storage vessels with no aterial input during the siulated tie. The nuber of put streas is arbitrary and user-defined. The possibility of siulating the siultaneous discharge through ore than one put strea is useful in situations such as the aperture of a relief valve followed by the opening of a bursting disk. We consider vertical cylindrical vessels; if the phase volues are known, it is siple to copute the level of the interfaces. By knowing these levels and the shape and position of the orifices, it is possible to deterine the leaking phase at any given oent. This ay change during the siulation. For exaple, at the beginning of a siulation, the liquid phase ay leak but, as its level decreases, vapor ay leave the vessel through the sae orifice in later stages of the process. In this paper, we assue that the orifice sizes reain constant during the siulated tie. The first step of the solution procedure is to deterine the initial condition of the fluid in the vessel. This is accoplished by solving a flash calculation whose specifications are the fluid teperature, internal volue of the vessel, and aount of each coponent (Espósito et al., 2000). By iniizing the syste s Helholtz energy and using global phase stability analysis (Michelsen, 1982), this calculation deterines the nuber of phases present, their volues, aounts, copositions, and internal energies at the beginning of the siulated process. When the effect of the vessel wall is included in the siulation, deterining the initial condition includes the evaluation of the wall s internal energy, given its ass and specific heat capacity at constant volue. With the initial state deterined, it is possible to start the dynaic siulation of the fluid in the vessel, which requires the solution of a set of differential equations. The differential ass balance of each coponent i, written on a olar basis, is: ns dn i = n dt = 1 i In Eq. (1), t represents tie, n i is the nuber of oles of coponent i in the vessel, and n i is the olar flow ns rate of coponent i in strea, which is one of the streas that exit the vessel. The energy balance takes the for: ( u ) 2 ns du = n h + M + Q dt = 1 2 In Eq. (2), U is the internal energy of the vessel wall (when it is considered) and of the fluid in the vessel, Q is the heat transfer to the fluid in the vessel, and h, M, and u are the olar enthalpy, olar ass, and average fluid speed at exit point, respectively. It is necessary to apply a nuerical ethod to integrate the ODE set coprised by Eqs. (1) and (2). In this work, the Bulirsch-Stoer ethod is used, as ipleented by Press et al. (1992). This is a conventional ODE integrator, not an integrator for differential algebraic equations (DAE). There has been no attept to use a DAE integrator because reliable nuerical convergence of phase equilibriu equations (algebraic) often requires highly specialized algoriths. The adopted strategy has been to solve the algebraic equations every tie the ODE integrator needs to evaluate the right hand side of the differential ass and energy balances, which are represented by Eqs. (1) and (2), respectively. To copute the, it is necessary to deterine the flow rate, coposition, average fluid speed, and enthalpy of each strea that exits the vessel. All these algebraic calculations consider the possibility of single or ultiple phases in the vessel and at the exit point. The first step to accoplish that is to deterine the state of fluid in the vessel. This requires solving a flash proble whose specifications are the current values of internal energy and coponent aounts, as provided by the ODE integrator, and the vessel s internal volue, which is constant. Deterining the equilibriu condition for this set of specifications requires axiizing the syste s entropy (Callen, 1985). A flash algorith developed specifically for this situation (Castier, 2009) is used. This (1) (2) Brazilian Journal of Cheical Engineering Vol. 34, No. 04, pp , October Deceber, 2017

4 1152 M. Castier, A. Basha, R. Kanes and L.N. Véchot algorith is extended in this work to account for cases that include the ass and heat capacity of the vessel wall, which is considered as a solid phase that exchanges heat, but neither ass nor volue, with the fluid phases present. The coes of this calculation are the fluid and wall teperature and fluid pressure, nuber of phases present, their densities and volues, aount of each coponent in each phase, and the olar enthalpy and entropy of each phase, aong other therodynaic properties. With the calculated phase volues (and ass densities, to establish their relative positions), it is possible to find the location of the interfaces for a given vessel geoetry (vertical cylinder, in this work). The position of each exit point is then checked against the interface locations to decide whether each discharge is fro a liquid or vapor phase, or both when an orifice is exposed to both phases siultaneously, as discussed by Kanes et al. (2016). The next step is to odel what happens inside the vessel next to each exit point. The assuptions are that each of these regions behaves as a converging nozzle that operates adiabatically and isentropically and that changes to the nozzle s input conditions have instantaneous effect on the exit conditions, i.e., ass and energy accuulations in the nozzle are neglected. An additional assuption is that the average fluid speed at the nozzle s entrance is negligible copared to its value at the exit point. These assuptions lead to the following two algebraic equations, written on a olar basis: h = h + M dch dch s = s ( u ) 2 where the superscript dch denotes properties of the discharged aterial inside the vessel, at the entrance of the hypothetical nozzle. It should be noted that the fluid ay have one or ore phases at an exit point, but this is not known beforehand. In fact, deterining the state of a fluid with specifications of fluid coposition, entropy, and stagnation enthalpy constitutes an unusual phase equilibriu proble. A new algorith for this situation has been the focus of recent work (Castier et al., 2016). An iportant feature of this algorith is that it begins with the assuption that the pressure at the exit point is equal to the specified value of the backpressure. If it turns that the flow is supersonic under these considerations, the solution is invalid because the sound speed is the upper liit for flow speed at the exit of an adiabatic converging nozzle. When this happens, the proble is solved again relaxing the pressure specification and considering that the flow is sonic. Thus, the sound speed calculations play an iportant 2 (3) (4) role on the overall coputation path. Very iportantly, they are occasionally necessary for systes with ore than one phase, if flashing conditions occur at the exit point. Few algoriths exist to this end. A recent procedure for the calculation of therodynaic sound speed in systes with any nuber of fluid phases (Castier, 2011) is used in this work. For ultiphase systes, this procedure involves the solution of a syste of linear equations forulated with the use of derivatives of several therodynaic properties, phase equilibriu conditions, and conservation equations for volue, entropy, and ole nuber of each coponent. Regardless of being sonic or subsonic, the flow rate at the exit point is obtained using: A u v n = (5) where A and v are the exit area and olar volue at point, respectively. In suary, three flash algoriths are necessary as part of these siulations: A TVN flash establishes the initial state of the fluid in the vessel for given volue, initial teperature, and initial aounts of each coponent. A UVN flash deterines the state of the fluid in the vessel during the dynaic siulation. Its specifications are the current values of U and N obtained fro the nuerical integration of the differential equations and the internal vessel volue, which is constant through the siulated period. A HSN flash deterines the state of the fluid at the leak point, which is assued to be the exit of a hypothetical converging nozzle. Its specifications are the stagnation enthalpy of the fluid that enters the nozzle, its entropy, and coponent aounts. In this work, the Peng-Robinson equation of state (Peng and Robinson, 1976) is used to copute the therodynaic properties, including the fugacity coefficients, residual internal energy, residual enthalpy, residual entropy, and therodynaic sound speed, as well as their derivatives, as required by the phase equilibriu calculations. In this sense, the sae therodynaic odel is used to evaluate all the therodynaic properties of the fluid. The critical properties, acentric factor, and coefficients of a 3 rd degree polynoial correlation in teperature for the olar ideal gas heat capacities are taken fro Reid et al. (1987). Despite the specific odeling choices, it should be noted that the forulation is general and not liited to using the Peng-Robinson equation of state, whose recoended application is for calculations involving hydrocarbon ixtures. Brazilian Journal of Cheical Engineering

5 Discharge of Non-Reactive Fluids fro Vessels 1153 RESULTS This section shows siulation results of cases in which light gases or hydrocarbons are vented fro vertical cylindrical vessels. The possible occurrence of cheical reactions is neglected in all cases. Discharge of nitrogen or ethane This conceptual case copares the discharge of pure nitrogen or pure ethane. The vessel is adiabatic and has a volue of , with height of 2.00 and diaeter of A 25.0x orifice located 1.00 above the botto of the vessel is used (with unity coefficient of discharge). The initial teperature of the fluid in the vessel s interior is 400 K. The vessel load consists of 80.0 oles of nitrogen or ethane and the theral inertia of the vessel walls is neglected. Figure 1 shows how the pressure inside the vessel drops as function of tie until it becoes equal to the backpressure, which is specified to be equal to MPa. It is noticeable that the discharge of nitrogen takes longer, which is an coe associated to the fluid speed at the exit point. Figure 2 displays the corresponding teperature profile during the siulated period. Methane has larger olar heat capacity than nitrogen and its teperature drop is accordingly saller. Figure 3 shows the fluid and sound speeds at the point of discharge. For each copound, these speeds overlap in the beginning of the siulated tie, until 38.7 s for ethane and 40.5 s for nitrogen, eaning that the exit flow is sonic, i.e., choked. The plot also shows that the discharge speed of pure ethane is faster than that of pure nitrogen, justifying the fact that the pressure in the ethane vessel becoes equal to the backpressure before this happens in the nitrogen vessel (Fig. 1). Therefore, these results illustrate how two pure light copounds behave differently during a leak, even though there is only one phase inside the vessel and at the exit point for both of the through the siulated period. Figure 4 shows the effect of increasing the orifice area to 100.0x10-6 2, i.e., quadrupling its area copared to the previous situations. As expected, the ties needed to equalize the vessel pressure and the backpressure are onefourth of the ties reported in Fig. 1. Discharge of nitrogen: coparison with Véchot s (2006) data Véchot (2006) easured teperature and pressure during the discharge of nitrogen fro a vessel initially at MPa and K through an orifice whose area is equal to 1.131x To atch these specifications, the initial aount of nitrogen in the siulation is equal to oles of nitrogen Pure nitrogen Pure ethane Tie (s) Figure 1. Pressure in vessels loaded with either pure nitrogen or pure ethane. Teperature (K) Pure nitrogen Pure ethane Tie (s) Figure 2. Teperature in vessels loaded with either pure nitrogen or pure ethane. Speed (/s) Pure nitrogen 100 Sound speed in nitrogen 50 Pure ethane Sound speed in ethane Tie (s) Figure 3. Fluid and sound speed in vessels loaded with either pure nitrogen or pure ethane. For an adiabatic vessel of negligible heat capacity copared to the heat capacity of the fluid it contains, the Brazilian Journal of Cheical Engineering Vol. 34, No. 04, pp , October Deceber, 2017

6 1154 M. Castier, A. Basha, R. Kanes and L.N. Véchot fluid teperature would decrease fro the beginning until the end of the experient. The experiental data show that there is a iniu in teperature equal to K at 58 s. The nitrogen teperature increases after that, indicating that there is heat transfer to the fluid in the vessel, either fro side the syste or fro the vessel s construction aterial. As Véchot (2006) did not report the ass and aterial of construction of the vessel, this effect was siulated by iposing a constant heat transfer rate to the fluid (equal to 41 J/s) so that the siulated teperature exhibits the sae iniu value, which occurs at 28.9 s of siulated tie. The ratio of the siulated and experiental ties for the iniu in teperature, equal to 0.50, was interpreted as a practical coefficient of discharge for this venting process, which is lower than the typical value of 0.61 for discharges through orifices. Figures 5 and 6 display the pressure and teperature profiles. There is only one phase within the vessel during the siulated period. The experiental data are plotted against the easured tie. The calculated data are plotted against a rescaled tie equal to the siulated tie divided by the coefficient of discharge of The plots end at 58 s, when the iniu in teperature occurs. The experiental and calculated results are in acceptable agreeent during this period. Beyond it, the disagreeent between the two results increases, especially for teperature, because of the adopted approxiation of constant heat transfer rate. Discharge of nitrogen: coparison with Haque et al. s (1992b) data Haque et al. (1992b) easured the gas teperature, wall teperature, and pressure during the discharge of nitrogen, initially at 15.0 MPa and 290 K, fro a vessel of undisclosed construction and insulating (if any) aterials. Haque et al. (1992b) report the following vessel diensions: height of 1.524, internal diaeter of 0.273, wall thickness of 0.025, with a circular orifice whose diaeter is , positioned at the top of the vessel. Using these diensions and assuing that the vessel is of stainless steel 316, whose density was taken as equal to 7950 kg/ 3 (British Stainless Steel Association, 2016), the vessel s ass was estiated to be equal to kg. To atch the proble specifications, the initial aount of nitrogen in the siulation is equal to oles, which Pure nitrogen Pure ethane Tie (s) Figure 4. Pressure in vessels loaded with either pure nitrogen or pure ethane: effect of larger orifice. 0.2 Calculated Experiental Tie, rescaled tie (s) Figure 5. Pressure in a vessel loaded with nitrogen: coparison of experiental (Véchot, 2006) and calculated values. corresponds to kg. Thus, under these assuptions, the initial ass of nitrogen is just 4.7% of the total ass of the syste. The vessel is considered as adiabatic and the internal energy of the solid is considered in the siulation, with the specific heat capacity of steel at constant volue calculated using the following third degree polynoial in teperature (in Kelvin): J cv = T T kg. K T (6) This expression fits experiental data (British Stainless Steel Association, 2016) between K to K with an absolute average relative deviation of 0.30%. Figures 7 displays the experiental pressure data as obtained fro the graphical results of Haque et al. (1992b) and plotted against tie until the end of the experient, at 100 s. It also shows the calculated results plotted against a rescaled tie, which is the siulation tie divided by Brazilian Journal of Cheical Engineering

7 Discharge of Non-Reactive Fluids fro Vessels 1155 a coefficient of discharge of 0.70, in such a way that the final rescaled tie at the end of the run is also equal to 100 seconds. At this condition, the siulated pressure inside the vessel becoes equal to the atospheric pressure. The odel overestiates the pressure at 10 s and 20 s, but there is good agreeent between the experiental and calculated pressures at later ties. Concerning teperature, the experiental data of Haque et al. (1992b) show that the fluid teperature passes through a iniu, of ab 190 K, and then increases, reaching a teperature of ab 220 K at the end of the experient. A possible explanation for this behavior is heat transfer fro the vessel walls to the fluid, counteracting the teperature drop caused by depressurization. The experiental value of the wall teperature drops through the experient, reaching a final value of ab 285 K. As the odel assues a single teperature for the whole syste, it cannot predict different teperatures for the fluid and for the wall. The siulated teperature results show a steady decrease in teperature fro the initial teperature of 290 K to a final teperature of K. This shows that the theral inertia of the vessel wall, whose ass is ab 20 ties the ass of nitrogen, doinates the odel s teperature prediction. Thus, the teperature fro the siulation is between the experiental fluid and wall teperatures, but closer to the latter because of the large heat capacity of the vessel wall. Discharge of nitrogen+ethane ixture fro heated vessel This conceptual case considers the discharge of an equiolar ixture of nitrogen and ethane fro a vessel of the sae size and shape as that of case 4.1, with the binary interaction paraeter of the Peng-Robinson EOS set to 0.1 (Firoozabadi, 1999). The vessel initially contains 40 oles of each copound at the teperature of 400 K. There is heat transfer to the fluid in the vessel at a rate of 1000 J/s. The vessel is initially closed and a relief valve (25.0x10-6 2, located 1.0 above the botto) opens when the pressure reaches 0.35 MPa. A bursting disk (100.0x10-6 2, located 1.5 above the botto) opens when the pressure reaches 0.40 MPa. Unity coefficient of discharges were used for the relief valve and busting disk. Figure 8 displays the pressure as a function of tie. The pressure initially increases because the vessel is closed and there is heat transfer to the fluid. When the pressure reaches 0.35 MPa, the relief valve opens. The effect appears as a change of slope in this pressure-tie diagra. However, because of the sall cross sectional area of the relief valve, the pressure keeps increasing after its opening. When the pressure reaches 0.40 MPa, the bursting disk opens. With a uch larger cross sectional area, its opening causes a sharp pressure drop, bringing the vessel pressure quickly down to the backpressure value. Teperature (K) Calculated Experiental Tie, rescaled tie (s) Figure 6. Teperature in a vessel loaded with nitrogen: coparison of experiental (Véchot, 2006) and calculated values Tie, rescaled tie (s) Calculated Experiental Figure 7. Pressure in a vessel loaded with nitrogen: coparison of experiental (Haque et al., 1992b) and calculated values Tie (s) Figure 8. Pressure in a vessel loaded with equiolar nitrogen+ethane ixture: effect of opening a relief valve and a bursting disk. Brazilian Journal of Cheical Engineering Vol. 34, No. 04, pp , October Deceber, 2017

8 1156 M. Castier, A. Basha, R. Kanes and L.N. Véchot Figure 9 shows the fluid teperature during the siulated period. The opening of the relief valve has little effect on the teperature but, when the bursting disk opens, it drops sharply. Figure 10 is a plot of the total olar aount within the vessel as a function of tie. The aount is initially constant and equal to 80 oles because the vessel is closed; it then undergoes a sall decrease while the valve relief is open but the bursting disk is closed, and finally decreases rapidly when both are open. Figure 11 shows the fluid and sound speeds. The relief valve opens at ab 293 s of siulated tie. Between this tie and the tie the bursting disk opens, at ab 795 s, the flow is sonic and of increasing velocity. Shortly after the opening of the bursting disk, at ab 815 s, the flow becoes subsonic. It should be noted that the flow and sound speeds in the relief valve and in the bursting disk, when both are open, are identical. This happens because there is only one phase within the vessel at all ties and, therefore, the properties of the fluid vented through both devices are equal. Discharge of a two-phase n-hexane+n-octane ixture This conceptual case siulates the discharge of an equiolar ixture of n-hexane+n-octane fro an adiabatic vessel whose volue is , with height of 1.00 and diaeter of The binary interaction paraeter of the Peng-Robinson EOS was set to zero, as usual for hydrocarbon ixtures (Reid et al, 1987). The orifice has an area of 12.57x10-6 2, centered 0.50 above the vessel s botto. The initial teperature of the fluid in its interior is 400 K. The load consists of 100 oles of each coponent and the coefficient of discharge is equal to 1. This section presents results of two cases with different initial fluid teperatures, equal to 450 K and 460 K. In both of the, the fluid has two phases inside the vessel at the beginning of the siulated period. However, the liquid level is low and the leak is fro the vapor phase at all ties. Figures 12 and 13 show the evolution of pressure and teperature in the vessel as a function of the siulated tie. For the initial teperature of 450 K, the fluid in the vessel has two phases through the siulated period. However, for the initial teperature of 460 K, the fluid has two phases up to only ab 670 s, when the liquid phase disappears fro the vessel. In Fig. 13, the liquid phase disappearance coincides with the change in slope that occurs at ab 670 s in the curve that represents the syste evolution fro the initial teperature of 460 K. Figure 14 shows the teperature at the exit point for the two siulated cases. Fro Figs. 13 and 14, it is clear that the teperature at the exit point is saller than in the vessel at any given tie. For exaple, at the beginning of the siulation, the teperature at the exit point is ab 10 K saller than the teperature inside the vessel. The curve Teperature (K) Tie (s) Figure 9. Teperature in a vessel loaded with equiolar nitrogen+ethane ixture: effect of opening a relief valve and a bursting disk. Aount (ol) Tie (s) Figure 10. Aount within a vessel loaded with equiolar nitrogen+ethane ixture: effect of opening a relief valve and a bursting disk. Speed (/s) Fluid speed Sound speed Tie (s) Figure 11. Fluid and sound speed in a vessel loaded with an equiolar nitrogen+ethane ixture: effect of opening a relief valve and a bursting disk. Brazilian Journal of Cheical Engineering

9 Discharge of Non-Reactive Fluids fro Vessels 1157 Teperature (K) Fro 450 K Fro 460 K Tie (s) Figure 12. Pressure in a vessel loaded with an equiolar n-hexane+noctane ixture: effect of initial vessel teperature. Teperature (K) Fro 450 K Fro 460 K Tie (s) Figure 13. Teperature in a vessel loaded with an equiolar n-hexane+n-octane ixture: effect of initial vessel teperature. 390 Fro vessel at 450 K Fro vessel at 460 K Tie (s) Fig. 14. Teperature at exit point in a vessel loaded with an equiolar n-hexane+n-octane ixture: effect of initial vessel teperature. for the initial vessel teperature of 450 K (solid line in Fig. 14) has a change of slope at ab 960 s, which coincides with the transition fro sonic to subsonic flow at the exit point. The curve for the initial vessel teperature of 460 K (dotted line in Fig. 14) exhibits two changes of slope. The first of the occurs at ab 670 s and is a consequence of the liquid phase disappearance fro the vessel. The second one occurs at ab 980 s, when the exit flow becoes subsonic. Figure 15 displays the instantaneous leak flow rates of n-hexane and n-octane as a function of tie for the initial fluid teperature of 450 K. Because the leak is fro the vapor phase, the exit flow rate of n-hexane is larger than that of n-octane at the beginning of the siulation; thus, n-hexane depletes faster. At ab 940 s, the flow rates of the two coponents becoe equal. After that, until the end of the siulated tie, the flow rate of n-octane is larger. This case illustrates the siulator s ability of dealing with the change to the nuber of phases present in the vessel and its effect on the properties of the exiting strea. Discharge of ethane+ethane+propane: coparison with Haque et al. s (1992b) data This case considers experient S9 of Haque et al. (1992b), which used a vertical vessel with torispherical ends and the following diensions: tangent-to-tangent distance of 2.250, internal diaeter of 1.130, wall thickness of 0.059, with a circular orifice whose diaeter is 0.010, positioned at the top of the vessel. Such a vessel has an internal volue of As torispherical geoetry is unavailable in the current version of our siulator, a vertical cylindrical vessel of the sae volue and diaeter is assued, and its corresponding length is It was also assued that the vessel is ade of stainless steel 316 with ass equal to kg. The vessel is initially charged with a ixture of ethane, ethane, and propane, whose ole fractions are equal to 0.855, 0.045, and 0.100, respectively. The initial teperature and total aount in the vessel were taken equal to 303 K and oles, which give an initial pressure of 120 bar and a ass of kg of fluid initially inside the vessel. Therefore, the initial ass of fluid is just 4.4% of the total ass of the syste. As in case 4.3, the vessel is considered as adiabatic and the internal energy of the solid is considered in the siulation, with the heat capacity of steel at constant volue given by Eq. (6). The binary interaction paraeters of the Peng- Robinson EOS are set equal to zero, which is typical for hydrocarbon ixtures (Reid et al, 1987). The existence of a single vapor phase through the siulated period is predicted. Figure 16 copares the experiental (Haque et al., 1992b) pressure with the calculated pressure, the latter plotted against rescaled tie, which is the siulation tie divided by a coefficient of discharge of 0.57, such that the rescaled tie at the end of the run is equal to 1200 seconds. Brazilian Journal of Cheical Engineering Vol. 34, No. 04, pp , October Deceber, 2017

10 1158 M. Castier, A. Basha, R. Kanes and L.N. Véchot Flow rate (ol/s) n-hexane n-octane Tie (s) Fig. 15. Exit flow rate fro a vessel loaded with an equiolar n-hexane+n-octane ixture for initial vessel teperature of 450 K Tie, rescaled tie (s) Despite the siilar trends, the calculated pressure is higher than the experiental pressure. According to Haque et al. (1992b), the final experiental teperature of the wall in contact with the gas is ab 290 K. They also report the foration of a sall volue of liquid phase in the vessel and the final teperature of the solid wall in contact with it is ab 250 K. As the siulator adopts a hoogeneous equilibriu odel, the calculated teperatures of the fluid and of the vessel wall are equal, initially 303 K, and K at the end of the siulation, which is soewhat saller than the final experiental teperature of the wall in contact with the gas (290 K). CONCLUSIONS Calculated Experiental Fig. 16. Pressure in a vessel loaded with ethane (85.5 ol%) + ethane (4.5 ol%) + propane (10.0 ol%): coparison of experiental (Haque et al., 1992b) and calculated values. This work presented the results of a siulator based on the hoogenous equilibriu assuption for discharges fro pressure vessels that contain non-reactive ixtures. The particular feature of this siulator is that it evaluates all the therodynaic properties using the sae equation of state, including phase equilibriu conditions, calorietric properties, and derivative properties, such as the sound speed. The Peng-Robinson equation of state was used, but the forulation is general and it is possible to adopt other therodynaic odels. The conceptual and sensitivity analyses reported here deonstrate that the siulator predicts eaningful trends for the coplex phenoena that take place during discharge operations. The siulated results generally agree with the experiental fluid discharge data when a coefficient of discharge is used. Future work on this siulator will focus on: (a) the effect of liquid swelling caused by bubbles due to non-instantaneous phase disengageent during leaks of condensable fluids; (b) expanding the availability of heat transfer odels available; (c) studying the venting dynaics of vessels with several runaway reactions. ACKNOWLEDGEMENTS A. Basha thanks Itochu Corporation for the support of her M.S. studies at Texas A&M University at Qatar. The authors acknowledge the suggestions of Dr. Valeria Casson-Moreno (University of Bologna, Italy). NOMENCLATURE A orifice area at exit point c v specific heat capacity at constant volue dch h olar enthalpy of the discharged fluid at the entrance of the hypothetical nozzle h olar enthalpy at exit point M olar ass at exit point n i nuber of oles of coponent i in the vessel n i olar flow rate of coponent i in strea ns nuber of streas that exit the vessel Q heat transfer to the fluid in the vessel dch s olar entropy of the discharged fluid at the entrance of the hypothetical nozzle Brazilian Journal of Cheical Engineering

11 Discharge of Non-Reactive Fluids fro Vessels 1159 s olar entropy at exit point t tie u fluid speed at exit point U internal energy of the fluid and of the vessel wall v fluid olar volue at exit point REFERENCES Bendiksen, K.H., Maines D., Moe, R. and Nuland, S., The Dynaic Two-Fluid Model OLGA: Theory and Application, SPE Production Engineering, May issue, 171 (1991). Birch, A.D., Hughes, D.J. and Swaffield, F., Velocity Decay of High Pressure Jets, Cobust. Sci. Technol., 52, 161 (1987). British Stainless Steel Association, topics.php?article=139, accessed on June 22 (2016). Callen, H.B., Therodynaics and an Introduction to Therostatistics, 2 nd ed., Wiley, New York (1985). Castier, M., Solution of the Isochoric-Isoenergetic Flash Proble by Direct Entropy Maxiization, Fluid Phase Equilib., 276,7 (2009). Castier, M., Kanes, R. and Véchot, L.N., Flash Calculations with Specified Entropy and Stagnation Enthalpy, Fluid Phase Equilib., 408, 196 (2016). Castier, M., Therodynaic Speed of Sound in Multiphase Systes, Fluid Phase Equilib., 306, 204 (2011). Cuber, P. Predicting Outflow fro High Pressure Vessels, Trans ICheE, 79, 13 (2001). D Alessandro, V., Giacchetta, G., Leporini, M., Marchetti, B. and Terenzi. A., Modelling Blowdown of Pressure Vessels Containing Two-Phase Hydrocarbons Mixtures with the Partial Phase Equilibriu Approach. Che. Eng. Sci., 126, 719 (2015). Espósito, R.O., Castier, M. and Tavares, F.W., Calculations of Therodynaic Equilibriu in Systes Subject to Gravitational Fields, Che. Eng. Sci., 55, 3495 (2000). Firoozabadi, A., Therodynaics of Hydrocarbon Reservoirs, McGraw-Hill, New York (1999). Haque, M.A., Richardson, S.M. and Saville, G., Blowdown of Pressure Vessels I. Coputer Model, Trans. Inst. Che. Eng. 70, 3 (1992a). Haque, M.A., Richardson, S.M., Saville, G., Chaberlain, G. and Shirvill, L., Blowdown of Pressure Vessels II. Experiental Validation of Coputer Model and Case Studies, Trans. Inst. Che. Eng. 70, 10 (1992b). Houf, W.G. and Winters, W.S., Siulation of high-pressure liquid hydrogen releases, Int. J. Hydrogen Energy 38, 8092 (2013). Kanes, R., Basha, A., Véchot, L.N. and Castier, M., Siulation of Venting and Leaks fro Pressure Vessels, J. Loss Prev. Proc. Ind., 40, 563 (2016). Leon, E.W., Huber, M.L., McLinden, M.O., NIST Reference Fluid Therodynaic and Transport Properties-REFPROP, Version 8 User s Guide, U.S. Departent of Coerce Technology Adinistration, National Institute of Standards and Technology, Standard Reference Data Progra, Gaithersburg (2007). Mahgerefteh, H. and Wong, S.M.A., A Nuerical Blowdown Siulation Incorporating Cubic Equations of State, Coput. Che.Eng., 23, 1309 (1999). Michelsen, M.L., The Isotheral Flash Proble. Part I. Stability, Fluid Phase Equilib., 9, 1 (1982). Peng, D.Y. and Robinson, D.B., A New Two-Constant Equation of State, Ind. Eng. Che. Fundaentals, 15, 59 (1976). Press, W.H. Flannery, B.P., Teukolsky, S.A. and Vetterling, W.T., Nuerical Recipes in Fortran 77: The Art of Scientific Coputing, Cabridge University Press. Cabridge, UK (1992). Reid, R.C., Prausnitz, J.M. and Poling, B.E., The Properties of Gases and Liquids, McGraw-Hill, New York (1987). Renfro, J., Stephenson, G., Marques-Riquele, E. and Vandu, C. Use Dynaic Models When Designing High-Pressure Vessels. Hydrocarbon Process., May, 71 (2014). Richardson, S.M. and Saville, G., Blowdown of LPG Pipelines, Process Safety and Environental Protection, 74, 235 (1996). Speranza, A., Terenzi, A., Blowdown of Hydrocarbons Pressure Vessels with Partial Phase Separation. Applied and Industrial Matheatics in Italy - Series on Advances in Matheatics for Applied Sciences, 69. World Scientific (2005). Véchot, L.N., Securité des Procédés. Eballeent De Réaction. Diensionneent des Évents de Sécurité pour Systèes Gassy ou Hybrides Non Tepérés: Outil, Expériences et Modèle, Ph.D. thesis, École Nationale Supérieure des Mines de Saint-Étienne, France (2006). Winters, W.S., Modeling Leaks fro Liquid Hydrogen Storage Systes, Sandia National Laboratory, Report SAND (2009). Witlox, H.W.M. and Bowen, P., Flashing Liquid Jets and Two- Phase Dispersion - A Review, Contract for HSE, Exxon-Mobil and ICI Eutech, HSE Books, Contract research report 403/2002 (2002). Woodward, J.L. and Mudan, K.S., Liquid and Gas Discharge Rates through Holes in Process Vessels, J. Loss Prev. Proc. Ind., 4, 161 (1990). Brazilian Journal of Cheical Engineering Vol. 34, No. 04, pp , October Deceber, 2017

12