Ethylene Decomposition Relief Sizing

High Pressure Polyethylene Industry
David I Shannon
E. I. DuPont de Nemours & Co., Inc.
2695 Old Bloomington Hwy (P.O. Box 526 – 77902)
Victoria, Texas 77905
Phone: 361-572-1813 E-Mail:
This paper discusses revisions to a method previously used for sizing relief devices
(rupture disks) on vessels in high pressure Polyethylene processes. The Re-evaluation
was forced by an unexpectedly high peak pressure in a product separator (High
Pressure Separator - HPS) when an ethylene decomposition occurred. A method, based
on revised equations for calculating the rate of pressure rise, is presented for
determining relief requirements for an ethylene decomposition in a confined space. The
method applies to both HPSs and Polyethylene Autoclave Reactors. Causes of
decompositions and effect of comonomers are also discussed.
As a result of an incident at a DuPont facility, ethylene decomposition relief device
sizing was revisited. In the process of doing this, other incidents were identified at
licensee facilities that also had unexpected results. After of the reanalysis of ethylene
decomposition data, a revised method of sizing was developed. This information has
been published in the past at the June 1992 ASME Pressure Vessel and Piping
Conference in New Orleans, La., and at a Society of Polymer Industry safety Conference
in February 1995. In recent years it has been observed that many High Pressure
Polyethylene Facilities have not made changes needed to address the risk identified by
these previous presentations. Additional discussion has been added about causes and
associated changes in decomposition severity, and effect of comonomers (vinyl
In January 1987, a DuPont facility experienced an incident where the HPS pressure
exceeded 5000 psig (extrapolated to >6000 psig, instrument maximum scale was
5000). The relief sizing for this vessel indicated that the rupture disks should have
been large enough to keep the event pressure from rising after the disk burst
(3300psig). Technically, even this event would not be a problem for this vessel if it had
a high enough design pressure, but the design pressure of 4500 psig only allowed a
peak pressure of about 5735 psig based on 85% of the hydrostatic pressure. The


The temperature increase is substantial due to poor mixing in the line. pressure was increased and decreased 3 times in a short period of time. These resulted from a repair of the steam condensate drain of the HPS bottom exit (tails) line and a poor but accepted operating procedure. Div. but also increasing the temperature in an already hot line (250oC). Reactor relief design has resulted in no events on a DuPont Reactor where all of the disks ruptured. This piping opened at the joints during this incident. When flow out the bottom of the jacket was restored. To break through the perceived plug. It was assumed that decompositions only result from propagation from Reaction or from external heat input. SIZING BASIS: The original work on ethylene decomposition sizing was done. Each of these incidents resulted from one pressurization of this line adiabatically compressing 250oC ethylene from less than 10 psig to 2400 psig. against an automatic valve.hydro-static test pressure was 6750 psig (designed based on ASME Section VIII. The event itself actually started in the line exit the bottom of the HPS. not low temperature conditions of start-up or shutdown. essentially acting as a piston compressing the hot gas at the end of the line which happened to be a broken guard valve. the poor procedure was exposed by the 2 decomposition events. not only increasing the pressure in the line. indicating that the available relief area was high enough to handle the pressure increase generated by the decomposition. It did not consider start-up or shutdown conditions that might resulted in different relief protection. the line (25 ft long) was pressured up. This design was based strictly on the operating conditions expect while making polymer. Two similar events occurred in 1994 at another facility. 2 and 3 for this temperature(s)) due to adiabatic compression. The operator involved was unable to establish flow through this 550 psig steam jacketed line. The poor operating procedure was masked by a condensate drain that had required flow out the top of the jacket. 439 279 . Based on that work the Reactor and High Pressure Separator (HPS) relief protection was designed. The piping in the off-gas system of this vessel was only designed for 2700 psig. with steam on the jacket. Relief protection for the piping was sized for process upsets only. This resulting temperature exceeded the ethylene decomposition temperature (see reference 1. Based on experience this assumption was invalid. the HPS disks are there to protect from a decomposition. within DuPont. 1). in 1952 through 1954 to assist in the design of a new High Pressure LDPE Autoclave process. The original HPS rupture design was for a decomposition that started at normal operating conditions.

that change with operating conditions and compositions. Equipment purging or plug removal can adiabatically compress gas in heated downstream piping. but also many different oligomers.. 440 280 . Or there can be flow regime problems in the HPS bottoms line that can adiabatically compress gas to very high temperatures. oxygen and peroxide initiators. and do. peroxides and other catalysts if used). Another is heat input from equipment or pipe jackets. an empirical calculation is necessary to get calculable results. the downstream one being closed. This was developed from tests run by initiating decomposition in a number of different sized static bombs and stirred autoclaves. squibs. particular copper. The decompositions were initiated by hot wire. e. and possibly some polymer as reported in reference 3. In the High Pressure Separator (HPS) in a Low Density Polyethylene plant.g. In a Polyethylene Reactor decompositions can be initiated by too much reaction initiator (oxygen. They can be initiated by not removing all of the oxygen from the process during equipment purging. hydrogen. with indefinite complex mixtures. external heat. flow is in the slug flow regime due to the wrong line pipe diameter for the length of line required. In reality the result of an ethylene decomposition event is methane. They are approximations intended strictly for sizing rupture disks for ethylene decompositions. propagate back into the HPS Original Rupture Disk Sizing Basis: Work in some unpublished laboratory work done in the 1950’s showed development of equation (1) for the rate of pressure rise during a decomposition. 4. The equation was developed by fitting the pressure rise versus time data. It should be recognized that the equations which follow do not represent a rigorous kinetic model. using HPS high pressure gas to quickly pressure up a heated bottoms line that can be plugged with polymer or may have multiple valves.DECOMPOSITION CAUSES: Ethylene decompositions result from ethylene breaking down. So. And they can be catalyzed by impurities such as metal compounds from packing. Decompositions that start in heated lines can. The first one expected is propagation of a decomposition from reaction. 5 and 6. decompositions have multiple potential causes. The chemical reactions that are normally proposed for this reaction are below: C2H4 Æ C + CH4 and C2H4 Æ 2C + 2H2.

the vessel volume.3 * K’ * P.0. in2 V = vessel volume.= 2. (2) dP ---. were estimated the same way. equation (3). psia.147 * P1. Based on the three incidents mentioned above and a number of incidents on other facilities known to DuPont. ft3 Note that the available relief area includes the effective capacity of any relief devices attached to the process lines and the flow through any line that leads to another system with significant capacity. This equation can only be used once the vessel pressure actually opens the rupture disks. This was done by fitting the rate of pressure decrease after the decomposition reaction was complete. (3) dP ---. So. 441 281 . the value of dP/dt in equation (3) is negative. The violence coefficient K’ for the HPS involved in the 1987 incident was estimated based on the laboratory data at similar pressures. and the total rupture disk flow area.= . when a rupture disk bursts.= 2. The design case for the vessel involved in the 1987 incident used a K’ of 0. Process vessels are not closed bombs.(1) dP ---. where dt S = A/V A = rupture disk area.2.0. sec. until then only equation (1) can be used. T = time. the pressure immediately starts back down. Equation (2) was also developed for disk venting rate as a function of the pressure.147 * P1. this is not normally the case.5 * S. such as off-gas line with relief valves.3 * K’ * P . where dt K’ = Violence coefficient P = Pressure. they have rupture disks so equations (1) and (2) can be combined to yield a single relationship.5 * S dt Ideally. such as the reactor. Violence coefficients for other vessels.

7 in2 and a burst pressure of 33 psig. instead of psia. includes 13 plant incidents within DuPont.807 Table 1 Closed Bomb Data Init T. as reported in the 1992 ASME Paper (8). 442 282 . Since the dP/dt was greater than zero the data was revisited resulting in the changes discussed below. rupture disk area of 31.With a vessel volume of 355 ft3.3423 * Ln(Den0) – 0.77 30.82 11.2 The data for the regression that produced equation (5) can be found in Table 2 below. Effect of Initial Density The actual peak pressure experienced in the 1987 incident indicated that something was not being considered. C Init P.2273 * Ln(Den0) + 6.4097 (5) Ln(K’) = 2. the value of equation (3) at burst is negative (-550 psi/sec).36 16. lb/ft3). (4r) Ln(Pmax) = 1.82 6. psia 150 22000 150 22000 250 4500 250 4500 250 2250 250 2250 150 4500 150 4500 Pmax 43000 68300 20100 18700 7500 8570 32900 32300 Init Den. or it’s Licensees over a 40 year period. All DuPont laboratory data and plant incident data were reviewed to look for variables not accounted for in these equations. (4) Ln(Pmax) = 1. it’s joint ventures. psia) and Violence Coefficient (K’) are functions of the mass of ethylene in a vessel at the start of the decomposition as measured by the initial density (Den0.896 The data for the regression that produced equation (4) can be found in Table 1 below.2 16. it was later found that the units for the result were in Kpsia. However. lb/ft3 30.77 11. Equation (4) shows the closed bomb pressure relationship and equation (5) the K’ relationship. so the statistics were revisited for this paper and equation (4r) generated as a replacement.663 * Ln(Den0) – 6.36 6. This revealed that the closed bomb pressure (Pmax.

lb/ft3 ref 7 lb/ft3 Lab Lab Lab Lab Lab Lab Plant Plant Plant Plant Plant Plant Plant Plant Plant Plant Plant Plant Plant Lab Lab Lab Lab Lab Lab Lab 23.55 10.2 9.2 27.7 19.352 -----24.1 18.8 10 7 5 0.6 6.4 27.7 11.9 22.2 3.37 25.4 28.81 10. Initial Density.47 27.1 28.7 8 10 7 4.98 15.63 0.2 29.458 10.692 2.11 30. So.92 10.7 16.51 0.Table 2 Violence Coefficient Data Initial P.2 8.35 26.084 6.022 10.176 0. Plant barg K' Data 9000 12700 14000 22000 22000 22000 27000 26000 13000 24000 11000 3770 3770 4002 3886 3712 3654 3770 2600 9000 9000 2250 2250 12700 14000 22000 621 876 966 1517 1517 1517 1862 1793 897 1655 759 260 260 276 268 256 252 260 179 621 621 155 155 876 966 1517 170 150 150 242 299 283 243 249 215 253 222 142 260 288 280 260 264 254 140 360 276 250 250 350 375 370 2. Initial Initial Violence Lab or psig P.698 ---- Additional regression work has been done on this data also to determine the effect of using different equations of state to determine the initial gas density of the system.053 11.1 24.82 30.3 10.57 0.5 15.73 28.450 * * * * Ln(Den0) Ln(Den0) Ln(Den0) Ln(Den0) 443 283 – – – – 6.61 24. lb/ft3 Density.896 7.47 0.352 16.78 20.944 26.2 10.67 10.060 7.54 0.871 -18.4 22. or two equations of state (Relich-Kwong and Benedict-Webb-Ruben--Lee-Kessler).1 27.48 0.785 6.663 2.62 10.421 11.41 10. The results of the regression are shown below: Original RK BWR-LK Actual Density7 Ln(K’) Ln(K’) Ln(K’) Ln(K’) = = = = 2.5 29.143 10.94 21.9 10.BWR-LK Initial Kwong Initial Density.22 29.57 1.99 11.18 0.1 10.698 6.78 16.514 .367 10. four equations were generated by regressing the data in Table 1.2 12 Relich. T.703 10.4 10.7 6.603 27. C Coefficient.59 6.2 10.73 0.8 20.738 -25. using the different sets of density data from either actual data tables (from Ref 7).8 26.78 6.99 16.735 2. the reason for which is unknown.1 26.15 27.3 3. In doing this update it was noticed that regression of the same data could not duplicate the original equation.6 26.91 0.11 22.7 24.87 10.

The standard deviation for these results was about 5% indicating that the four equations provide about the same result.5 * S ) * td. the results will not change enough to provide an unsafe rupture disk sizing result. where td = calculation time increment. (5r) Ln(K’) = 2. 2600 psia.0. the equations above provide a good estimate of ethylene decomposition peak pressure given the initial conditions. Therefore. (equations (1) – (5)) a computer simulation was written to solve equation (3) for small time increments. If the original equation is used.To determine if the difference in these equations is significant. 355 ft3 vessel.450 * Ln(Den0) – 6. where the rupture disk bursts. 3300 psia burst pressure). it is recommended that the equation regressed on actual density data be used (equation (5r) below. with 31.7 in2. For each case the same rupture disk area of 31. At t = tmax. The plot shows pressure constant until t = 0.7 in2 and volume of 355 ft3 were used. So. the pressure rise portion of equation (6) drops out of the equation since there is no longer any ethylene to decompose. The pressure at time tmax is the peak pressure. not shown. 5 different densities were input to each equation in the “Peak Pressure Estimation” method discussed below. The graph in Figure 1 indicates a good fit to the actual event. Simulation of other events at different operating conditions. sec The results are shown in Figure 1 (1987 incident – 140oC. The times tr and tmax are calculated by equations (7) and (8) below.514 To test this peak pressure algorithm.3* K’ * Pt – 0.147* Pt1. then rising based on equation (6) without any venting until t = tr. 444 284 . indicate a good match for peak pressure. or an equation of state is used to estimate ethylene density. since it is based on actual data. changing the differentials to differences leading to equation (6) below: (6) Pt+1 = Pt + (2.

a more rigorous calculation has to be made. sec 445 285 . or by integrating equation (3). where 2. This can be done by unsteady state calculations as discussed above (equation (6)).. To do this.1987 7000 6000 Pressure. psia tmax = Decomposition duration. (7) tmax Pmax Ln(------) Po = -----------------. equation (1) is integrated and solved for tmax to get equation (7). or pressure at which reaction is completed without a relief device. psia 5000 P HPS Press. psia 4000 3000 2000 1000 0 -5 -3 -1 1 3 5 7 Time. Po = Initial pressure. the rupture disk burst time and the duration of the event need to be calculated. sec Peak Pressure Estimation If the value of equation (3) is greater than zero.3 * K’ Pmax = Closed Bomb Pressure. In order to estimate the peak pressure during a decomposition event. psia.Figure 1: Ethylene Decomposition Incident .

5 * S * e0. where Y2 X = 2.63066 (Wt.3 * K’ The amount of time the rupture disks are venting until peak pressure is reached is calculated by equation (9).6. The integration is done between the rupture disk burst pressure (Pr) and the vented pressure (Pv). (10) X2 Pv = ----.147 * Pr0.3*tv For the incident conditions above (1987) equation (10) yields a peak pressure of 6497 psia (K’=0.896 . the time to rupture disk burst (tr) is calculated by equation (8). tr is reference). which agrees with the unsteady state method results shown in Figure 1.By substituting the rupture disk burst pressure (Pr) for the closed Bomb Pressure (Pmax).147 * Pr0.5 * S + Z Z = 0. (8) Pr Ln(------) Po tr = -----------------2..636). This peak pressure is the maximum pressure expected in a vessel during an ethylene decomposition. EFFECT OF COMONOMERS A small amount of data is available for Vinyl Acetate decompositions from a DuPont internal report.3 * K’ – 0. from tr to tmax (or zero to tv. A fit to this data combined with the ethylene K' data gives the following: Ln(K') = 2. where tv = venting time to peak pressure. (9) tv = tmax – tr .663 Ln(Eth-Deno) .Fr.3*tv Y = 2. As shown in Table 4 the violence coefficient for the 80% VA system is more than 1 order of magnitude lower.3 * Pr0.5 * K’ * e0.2.5*K’*2. sec The times calculated above are then used in an integrated (rearranged) version of equation (3): equation (10) which is used to calculate the peak vented pressure (Pv).5*K’*2.VA) 446 286 .

Jens and Gerhard Luft (October 1999). “Runaway phenomena in the ethylene/vinylacetate copolymerization under high pressure”.15 7.0315 .49 6.0047 . 3. Neumann. 447 287 . with the significant reduction in K’ with increasing vinyl acetate. Jens and Gerhard Luft (1998). Chemical Engineering and Processing 37 (1998) pp.58 .51 12. So. 2.0046 .11 . for example.06 7.0063 .48 .225 14. 55-59. and no closed bomb pressure data. is approximately 1/2 of the ethylene K'. and a modified DuPont method. which assumes a steady state mass flow rate that is sufficient to move energy from the vessel at a rate equal to the maximum rate at which it is input into or released within the vessel”. The major limit here is that the effect of VA on the closed bomb pressure is not known. 45. 2214-2222. “Studies of the Thermal Decomposition of Ethylene under High Pressure”.Density Coefficient Predicted lb/ft3 lb/ft3 K' K' 19. See reference 9 for details.253 ALTERNATE METHODS: There are at least two other methods of calculating decomposition required relief area. discussed above (based on ref. pp. R. 10. Chemical Engineering Science 28. Albert. No.94 . The effect that this has on an ethylene decomposition is that K' for 18% VA. Albert.Ethylene Decomposition Violence Coefficient Data Initial Initial Press. 1505-1514. in them. sizing for ethylene only is recommended.127 15. AICHE Journal Vol.02 . One recent paper presented a comparison of the DuPont method. o Psig C 3758 272 2420 330 2880 302 2865 329 3085 361 Initial Initial Violence Ethylene Density Eth. More data is needed to accurately do these calculations for systems with Vinyl Acetate. 8 equations from 1992) and what was designated as “conventional method. Temp. Luft (Aug. or other comonomers.19 14. “Thermal Decomposition of EthyleneComonomer Mixtures under High Pressure”.036 . pp. Table 4 80% Vinyl Acetate .This is the ethylene K' as a function of initial density with the addition of a term that corrects for VA concentration. 1973).00 10. and G. REFERENCES: 1.22 7.

M. 25. 862-864. Pachec ’72. pp. 9.. 448 288 . No. 2). Kurihara and Masaaki Takehisa (1972). Vol. 130-134. Process Safety Progress (Vol. Bonsel. No. pp. 605-612 (Department of Research. ASME. “Safety Studies on the Explosive degradation of Compressed Ethene”. “DECOMPOSITION VENTING IN A POLYETHYLENE PRODUCT SEPARATOR”. Japan Atomic Energy Institute). 12. 6. International Union of Pure and Applied Chemistry: ETHYLENE – International Tables of the Fluid State. “Relief Vent Sizing for Deflagrations”.4. Hiromasa and Masaaki Takehisa (1982). Loss Prevention and Safety. Takasaki Institute. and D. Sullivan. Hiromasa. Codes and Standards and Applications for High Pressure Equipment. pp. PVP. Chemical Engineering Technology 67. Harald and Gerhard Luft (1995).I. pp.238. “Studies on the Explosive Decomposition of Ethylene—Mechanism of Ethylene Decomposition under High Pressure”.F. 7. 7. Watanabe. Watanabe. K. 8. 225-232. No. 5. Shannon (1992). pp. “The Explosive Decomposition of Ethylene”. Koatsu Gasu 19. 1992. H. Darby. (Reference Book) DeReuck. Ron (2006). J. 191-195.