Premature foam-flood in an amine absorber: Part 2

Troubleshooting and hydraulic analysis identified vapour channelling in fixed valve trays as the root cause of premature foam-flooding in a high pressure amine absorber.

ROHIT KUMAR Bharat Petroleum Corporation Limited

Viewed : 578

Article Summary

To recap the issues described in Part 1 of this article (PTQ Q3 2021): the lean amine flow rates in the high pressure amine absorber in the hydrocracker unit at Bharat Petroleum’s (BPCL) Kochi refinery were limited by recurring foam-over events. To increase liquid handling capacity, the tower was retrayed by a major tray supplier (Fluor was not involved). The retray was quite standard, using high-capacity valve trays with small fixed valves and push valves and increasing downcomer top area. Surprisingly, higher lean amine flow rates could not be achieved without foam-overs. Even at the same amine and gas rates, the frequency of the foam-over events jumped from twice per month to once per day. Some 54 days after the new trays began operation, the bottom drain line from the pre-knockout drum plugged, causing pre-condensed lighter hydrocarbons entrainment into the amine system, and severely aggravating the foaming issues. This required a reduction in both the amine and gas flow rates. The problem was eventually resolved by replacing the tower by a random packed tower and unplugging the pre-KOD bottom drain line.1

The flood analysis described in Part 1 of this article invalidated any theories regarding jet flood or downcomer back-up flood or insufficient downcomer residence time having any relevance to the observed carry-over events. The flood analysis also explained the absence of events, when operating in diesel mode, by this mode being non-foaming.

However, the flood analysis was unable to explain the events experienced with the revamp trays. A joint investigation revealed a previously unreported phenomenon: vapour channelling inducing premature tray foam-flood.

There were theories of gas breaking into the downcomers due to the zero static seals of the revamp trays. The low weir loads during the 11-24-2019 event and during the gas bypass operation suggested possible operation in the spray regime, at which gas can easily break into the downcomers. Once in the downcomer, this gas can prevent the descent of liquid, causing a downcomer unsealing flood in a manner similar to an experience reported in the literature for another amine absorber.2 There was nothing in the tray supplier’s software to permit this evaluation, but this theory was evaluated during the current rigorous analysis.

Table 1 shows that, despite the low weir loads, operation of the absorber was in the froth, not spray regime based on published criteria.3 The very low C-factors in the absorber render the gas momentum too low to atomise even the small quantity of liquid on the trays. Further, a calculation performed based on the assumption that gas does break the seal into the downcomers showed that its velocity in the downcomers would have been too low to interfere with the liquid descent. This study invalidated the downcomer unsealing theories causing the event at low weir loads.

Pressure drop and weep checks
The middle part of Table 1 evaluates the pressure drops and weep. While the total pressure drop of the new trays is within the range of common practice, the dry pressure drops of the new trays are very low. A common design practice is to keep the dry pressure drop above 13 mm of liquid. The dry pressure drops of the new trays fall short of this criterion and are less than half of the pressure drops of the old trays.

At these low pressure drops, weeping is likely from the new trays that would not have occurred with the old moving valve trays. At low gas velocities, the moving valve floats of the old trays would have seated and covered the tray openings, which would have mitigated weeping and kept up the dry pressure drop.

Two correlations have been recommended for weep calculation:3,4 the Lockett and Banik5 and the Hsieh and McNulty correlations,6 both of which can be found in accessible references.3,7 Based on analysis of FRI tray efficiency data, Colwell and O’Bara4 expressed preference to the Lockett and Banik correlation for low pressure systems and for the Hsieh and McNulty correlation for high pressure systems. For the current high pressure absorber, we preferred the Lockett and Banik correlation, because the physical properties of the liquid and to a lesser degree also of the gas lined up better with the low pressure systems than the high pressure systems that Colwell and O’Bara studied. Therefore our weep calculations are based on the Lockett and Banik correlation, but we also used the Hsieh and McNulty correlation as a check. Table 1 shows that the values calculated from the two correlations were in quite good agreement.

Both correlations showed a significant amount of weeping from the revamp trays even at the higher gas rates, and much more extensive weep, as much as 50-60% of the liquid, after the liquid rates were reduced following the pre-KOD bottom line plugging.

Summers et. al.9 proposed the tray stability factor as a measure of tray stability:

h   = ( DPD /HCL)      0.5                           [1]
where η is the tray stability factor, dimensionless; ΔPD is the dry pressure drop, mm liquid; and HCL is the clear liquid height, which can be calculated from the difference between the total and dry pressure drops or from the Colwell equation.8 Data analysis by Summers et. al. show that when the stability factor exceeds 0.6, weeping is less than 30% and the tray is stable. Weeping rapidly escalates at lower stability factors, with their data analysis showing as much as 50% weep or more by the time the stability factor declines to 0.4. Table 1 shows that, according to the Summers et. al. criterion, the old absorber trays had stable operation (stability factor of 0.7), but the new revamp trays were highly unstable, operating at stability factors of 0.4 or less. 

In summary, the pressure drop evaluation revealed a major issue with the revamp trays: a very low dry pressure drop and a very low stability factor, both well below the range of good operating practice.

Channelling induced foam-flood
Summers et. al.9 also state that in multi-pass trays with low stability factors there is an additional opportunity for the trays to become unstable, totally unrelated to weeping. That instability allows vapour to channel through a part of the tray, travelling upwards without flowing through the other tray passes. One of these authors reported an experience with two-pass trays at low stability factors where he could determine visually that the vapour channelled through a single side of a two-pass tray tower.

Channelling is a common occurrence on trays where the dry pressure drop is low and at the same time a source of vapour channelling exists. This source will induce vapour to preferentially rise through one section of the tray. The dry pressure drop counters it, tending to distribute the vapour evenly. A low dry pressure drop will be ineffective to counter the channelling source, so the vapour will channel.

Channelling leads to premature flooding, excessive weeping, and reduction in tray efficiency.  Common sources of channelling include:
1. High hydraulic gradients on the tray, with vapour preferentially rising near the tray outlet where the liquid head is lower. The channelling occurs when the dry pressure drop is too low compared to the hydraulic gradient. This is termed vapour cross flow channelling (VCFC) and has much written about it.10-12
2. Reflux or feed maldistribution to multi-pass trays. Here the pressure and downcomer head balances can propagate such maldistribution, especially when the dry pressure drop is low.13
3. Vapour maldistribution. With low dry pressure drops such maldistribution propagates from tray to tray and persists through the tower. In one case,14 overflow of liquid on a chimney tray generated vapour maldistribution and, due to the low dry pressure drop, it propagated right through the tower.

Add your rating:

Current Rating: 1

Your rate: