Condensation-induced water hammer is a commonly occurring phenomenon in steam and condensate systems. In steam/condensate systems, condensation-induced water hammer occurs when steam bubbles come in contact with, or are encapsulated in, subcooled condensate and then loses heat and condenses rapidly. This creates a low-pressure zone into which condensate moves rapidly. The resultant collision creates a pressure wave that reverberates through the body of condensate. The pressure developed from the collision can be derived from the momentum balance and can be written as follows:
P = ρV2
P = Pressure
ρ = Liquid density
V = Condensate velocity
The velocity can be as high as sonic velocity, i.e., the velocity of sound in the condensate. For steam/condensate systems, this transient pressure can be high enough to create catastrophic damage to piping. The velocity and consequently the pressure developed is a function of the following:
Quantity of noncondensable gases
Directionally, higher steam pressure, lower condensate temperature and higher bubble size favor development of higher-pressure transients. The noncondensable gases reduce the transient pressure due to their presence in the low pressure zone.
For hydrocarbon systems, condensation induced transient pressure waves similar to water hammer in steam systems do not occur often. This is because the fluids are generally multicomponent hydrocarbon mixtures and the sudden collapse of bubbles does not occur. In addition, flow regimes where vapor bubbles can be enclosed in sub-cooled liquid are generally avoided in the design. However, in one plant, there were two cases of condensation-induced transient pressure wave problems in relatively pure component service (~97% propane) that will be described in greater detail.
Case 1: Propane rundown system.
The propane fractionation and rundown system is illustrated in Fig. 1. There are four depropanizers in the plant. The feed to the depropanizers is natural gas liquid (NGL) that contains propane, iso-butane, n-butane and condensate or light naphtha. High-purity propane (~97%) is the overhead product of these depropanizers. Three depropanizers have air-fin coolers and one has a water cooler in the overhead condenser service. The overhead pressure in the columns is controlled by withdrawing the propane product from the overhead. In other words, it is a flooded condenser pressure control where the level in the condenser is varied indirectly to control the pressure. In addition, the columns are equipped with variable pressure-control logic to take advantage of varying ambient temperature during the day. The variable pressure control increases the pressure in the morning and reduces it in the evening with varying ambient temperature. The control logic maintains a minimum subcooling to avoid vapor formation in the reflux drum and for proper functioning of the pressure control system. However, when the ambient temperature increases fast, the control system might fail to maintain the subcooling required and operator action would be necessary to reduce the feed to the columns.
| Fig. 1. Propane rundown system. |
During a capacity-expansion project, booster pumps were added to increase the propane production rate. Downstream of the pumps, there are filter coalescers to remove water carried over from the reflux drum. These are followed by driers and treaters that remove water and H2S from propane, respectively. The driers and treaters are molecular sieve beds that are taken through various cycles of operation such as absorption, draining, depressuring, regeneration, cooling and liquid filling. The switch between the various cycles is automated and air-operated isolation valves are used for this purpose.
The propane rundown is finally routed to storage bullets. The bullets are intermediate storage tanks where propane is routed to underground caverns for long-term storage, then to loading trucks and rail wagons. The bullets also act as intermediate storage for propane that comes from the underground caverns for loading trucks and rail wagons.
Relief valves are located on the filter coalescers and on the line, set at 320 psig. During summer, while the storage bullets and the columns are operated at higher pressures, these relief valves were chattering, causing damage to the valve seats. The relief valves located on treaters and driers that have a similar set pressure did not have any incidents of chattering or valve damage. The maximum pressure recorded by the pressure indicator in the control room was only around 250 psig. Tell-tale gauges (local gauges that record maximum pressure) installed close the relief valves also recorded only a maximum of 250 psig. Even after repeated repairs and calibration of the relief valves, the problem persisted.
Drier/treater controls and relief valves.
Drier/treater controls. Switching the driers and treaters between different cycles of operation is done using automated valves. It was suspected that the premature closing of the valves and/or errors in logic might be causing flow restrictions. The switching logic was reviewed and revised to avoid any potential problems. Also, the closing of the automatic valves were checked by operators to ensure that they were not getting stuck or prematurely closing.
Relief valves. Relief valves were replaced by pilot-operated valves with the following objectives:
Relieving close to the set pressure
The effect of inlet line pressure drop was eliminated by taking pressure signals close to vessels and the main line.
Slower closure of the valves in the event of relieving.
The steps taken described previously did not make much improvement in the situation. Hence, it was decided to conduct a theoretical investigation of other causes of overpressure.
The storage bullets act as intermediate storage where levels and temperatures can fluctuate. Therefore, a pressure transmitter was installed on the bullets to monitor the pressure continuously. The rundown system was modeled using simulation software. With monitoring the bullet pressure, it became evident that the propane in the bullet was not always at equilibrium, and, due to level changes, considerable transient pressure changes were occurring. Some of the lowest pressures recorded in the bullets were used in the simulation for checking two-phase conditions. From the model, the following conclusions were reached:
While the pressure in the bullets was lower than the vapor pressure, considerable vaporization was occurring in the rundown system (~15%)
Some parts of the rundown lines were in the slug flow regime.
Due to a reduction in elevation, the vapor was partly collapsing at a few locations.
From these conclusions, it was surmised that the collapsing of the vapor due to elevation changes might be causing transient shock waves. A phenomenon similar to condensation-induced water hammer in steam systems was suspected to be occurring due to different causes.
Unfortunately, most of the dynamic simulation programs are not capable of simulating this type of transient pressure waves. Hence, only a qualitative analysis from the steady-state simulation results was made.
A backpressure controller was installed close to the storage bullets to isolate the rundown system from the pressure variations in the bullet (Fig.2). Backpressure higher than the expected vapor pressure was maintained and this eliminated the relief valve chattering problem.
| Fig. 2. Modified propane rundown system. |
Case 2: Propane overhead system.
The system in this case is the overheads of the depropanizers, previously described. The overhead system is shown in Fig 3. There are relief valves on the column overhead and the reflux drums with set pressure of 280 psig.
| Fig. 3. Depropanizer overhead system. |
Of the four relief valves on reflux drums, one used to chatter and cause frequent damage and another one received occasional damage. Both of these depropanizers have air coolers in the overhead condenser service. One of the propanizers with air coolers and one with a water cooler did not have this problem.
The operating conditions of all the depropanizers were very similar and we could not identify any deviation of consequence regarding operating conditions. From analyzing historical data the following two interesting observations were made:
Originally, the columns had a hot vapor bypass control for overhead pressure and later the control system was changed to a flooded control system. The relief-valve problems started subsequent to the control system change.
Most of the relieving incidents were reported to occur in the night. Due to variable pressure controls, the columns are operated at lower pressures during the night (180 to 200 psig) against a maximum of 250 psig during the day.
The fact that the relieving incidents occurred during the night, when operating pressures are lower, indicated that the problem was not linked to any rapid fluctuations in operating pressures. Hence, it was decided to look at any difference in installation between these columns. The piping drawings were reviewed and field measurements were made to determine the relative elevation of the relief valves with respect to the bottom flange of the air-fin coolers. The relief valves that used to chatter had a relative elevation of +20 in. and +46 in. above the outlet flange of the air coolers. The relief valve on the system with the water cooler had a relative elevation of +87 in. The system that did not have any chattering problem had a relative elevation of only +1 in. Fig. 4 is a depiction of relative height.
| Fig. 4. Relative location of relief valve. |
Although the control systems were very similar, the pressure in the column with the water cooler never had drastic changes in the overhead pressure due to the reasonably steady water temperature. When the water temperature varied, the variation in pressure was slow and the control system did not make drastic changes. The variations in air temperatures were more drastic and the pressures reached the maximum-allowed pressure of 250 psig in a span of a few hours. Also, in the night, the pressure was reduced up to 180 psig.
During the day, when the ambient temperature is high, the control system will tend to push the liquid level in the condenser down. In the cases where the relief valves are placed above the bottom of the condenser, this will lead to the draining of liquid and developing vapor space in the inlet pipe. In the night, while the ambient temperature reduces the vapor-inlet line might be collapsing and the higher liquid level in the condensers will force the liquid to rush into the transient low-pressure zone, causing chattering of the relief valves.
The preferred solution was to reduce the elevation of the relief valves. However, this could not be done since the flare header was located at a marginally lower elevation with respect to the relief valves. Reducing the relief-valve elevation would have led to pockets in the relief valve outlet line. It was verified that the relief valve on the column had enough capacity to cater to both the column and the reflux-drum relieving cases. Hence, it was decided to remove the relief valve on the reflux drum and to isolate it close to the vessel. This way, the possibility of forming vapor space that may lead to pressure transients was eliminated. All the block valves between the reflux drum and the relief valves were locked open to provide a clear relieving path. This modification was implemented only in the depropanizer that use to have frequent failure. HP
1 Kirsner, W., Steam condensation-induced water hammer, HPAC, January 1998.
|The authors |
George Mani is a process engineering specialist with BP Canada Energy Co. Currently, his responsibility is giving process engineering support to multiple natural gas processing and NGL facilities in operating and implementing projects. Mr. Mani has 30 years of wide-ranging experience in the oil and gas idustry.