Revamps 2026 Issue

Pressure (PSIG) Temperature (˚F)

Butane

7000

DIB #1

6900

6800

Horizontal 300 ft

6700

DIB #2

Vertical 75 ft

Reux (bpd) Stripping DP (psid)

6600

Note: Data 1 minute averages

6500

1:29

1:37

1:46

1:55

2:03

Horizontal 100 ft

Horizontal 75 ft

Alkylate

Figure 2 Cyclic pressure drop in stripping zone

Figure 3 Debutaniser feed schematic

A combination of the high-capacity PFMD and ECMD trays was installed in the tower to meet the revamp expec- tations. Tray installation was successful and met all require- ments. The revamp success story is documented in the AIChE paper.1 Troubleshooting the revamp After a very successful start-up, the team became very interested in tower performance testing. The author con- siders performance testing to be a critical part of the pro- ject. The performance test not only validates the design but also provides a good source of data for tower benchmark- ing and enables operations to get a good feel for the tower as it approaches limitations. In this case, operations got a ‘three lessons in one’ per- formance test, but the author will only cover the most significant learning from the test. The full results of the per - formance test were presented in the AIChE presentation on which this article is based. The goal of the performance test was to push the tower trays and peripherals to their limit. Throughout the day, the internal recycle rate (tower distillate-to-feed rate) was increased to load up the reactor section. This would allow higher feed rates and increase tower loading (requiring more reflux and reboiler duty for constant separation). After the recycle rate was increased, the tower distillate was placed in flow control while the reboiler steam rate was increased to intentionally load up the tower. Small changes were made to the reboiler duty to artificially load up the tower. At 70% of design rates, the tower entered a distinct pressure-drop cycling mode that intensified as throughput increased to 90% of the design rate. Above 90% of the design rates, the amplitude of the cycling seemed to be constant. Every 540 seconds (nine minutes), the pressure drop would go through a defined cycle. The pressure drop would cycle 110% from high to low off an average. Figure 1 shows the standard responses in dP (ΔP) for a typical downcomer and jet flood situation. Unlike typical jet and downcomer flood, the debutaniser pressure drop would never escalate into a liquid build-up or exhibit the typical characteristics of flood (loss of temperature differ - ential, loss of liquid levels, or massive liquid carryover).

The observed ‘flood’ was very unusual based on the author’s experience. The author had only seen this type of ‘stitching’ pressure drop when a previous tower was on the verge of dumping. However, it operated this way only for a short (<4 hr) time frame. The debutaniser would operate in this ‘stable’ mode for days, weeks, and even months on end. There was never any evidence of uncontrolled prop- agation resulting in an unstable column that affected downstream column operation or product qualities. After operations got accustomed it, it became a non-issue to maintain. However, discovery work was required to address this unique issue, even though there was no real ‘urgent problem’ to solve. During the cycling, the tower products were sampled at tight intervals. Even though this was not the typical method to flood, the author was interested in the effects of flooding. The samples were taken at tight intervals to ‘catch and identify’ operating deviation. The samples did not reveal any loss in efficiency that would show up as heavy key or light key breakthrough into the respective products. The sample results did not indicate that the tower was losing efficiency or separation performance during cycling compared to normal flatline pressure drop operation. In fact, even the control point temperatures were flatlining during the evaluation period, which suggested normal fractionation. The magnitude of the pressure drop cycling seemed to indicate that a single tray, or at most two, were flooding. This would show up as a stacking of a couple of feet of light hydrocarbon liquid. The pressure drop cycling was about 0.1-0.25 PSID (depending on feed rate). Figure 2 shows a subset of the tower pressure drop data and the corre- sponding flow diagram (see Figure 3 ). By process of elimination, it was quickly ruled out that the lowest tray (highest loaded) was flooding. Typically, three events accompany a stripping section flooding event. First, the reboiler duty and return temperature will fluctu - ate/cycle, and this was not occurring. Second, when the bottom tray floods, the flood frequently propagates up the tower, resulting in an accompanying response of liquid build-up, high pressure drop, and bottoms liquid inventory

Revamps 2026