SIMPLE METHODS SOLVE VACUUM COLUMN PROBLEMS USING PLANT DATA

Scott W. Golden, Andrew W. Sloley
Glitsch, Inc.
Dallas

Simple methods can be used to evaluate cornS mon vacuum column problems using actual field measurements.

All that is required is an enthalpy table, a calculator, and an absolute pressure manometer, which can be purchased for about $100.

USING PLAIN DATA

The key to troubleshooting refinery crude or lube vacuum columns is basic plant data.

The process should begin with field measurements of the operating column (pressure and temperature profiles); analysis of the unit heat and material balance; and where appropriate, computer simulations. Such field observations will pinpoint the problem.

Noted distillation troubleshooting specialists Norman P. Lieberman and Henry Z. Kister both warn that the solution to an operating problem comes from the plant and its data, and not from sophisticated computer models.

Computer simulations should be used only as a tool to analyze the data and not as the starting point of a troubleshooting exercise.

Although many techniques may be used to increase cutpoint, many times the largest yield improvements can be achieved on existing units simply by eliminating such problems as leaking collector trays or overflowing liquid distributors.

CASE STUDIES

Three actual troubleshooting cases highlight the use of basic plant data and data analysis to solve problems:

Case No. 1: Poor light/heavy vacuum gas oil (LVGO/HVGO) fractionation, low LVGO product yield
Case No. 2: Poor fractionation in a trayed lube column
Case No. 3: Low HVGO product recovery.

In Case 1, a trayed vacuum column producing hydrocracker-quality LVGO was revamped with structured packing to increase product yield. The LVGO product yield actually decreased after the revamp.

For Case 2, product quality decreased after a unit shutdown. Feedstock changes were thought to be the problem.

In Case 3, a unit revamp was proposed to increase the low HVGO product yield by modifying the furnace and column internals. Greater feed enthalpy was thought to be the solution.

In each case, the problem was identified by field measurements and basic data analysis.

CASE 1

DESCRIPTION: Poor LVGO/HVGO fractionation.

HISTORY: An 85,000 b/d trayed vacuum column was retrofitted with structured packing (Figs. 1 and 2) to improve the recovery of LVGO product.

The column design basis was 85,000 b/d of 12.5 API gravity Arabian Heavy atmospheric residue. The LVGO product was feed to a hydrocracker having a true boiling point (TBP) 98 vol % specification of 870 F.

The main justifications for this project were increased LVGO product yield through improved LVGO/HVGO fractionation bed efficiency, and increased capacity of the fractionation zone internals to allow higher reflux. The revamp design basis was a 5,000 b/d improvement in LVGO product yield.

PROBLEM: Prior to the revamp, the trayed column produced 15,000 b/d LVGO product. The revamped column was designed to yield 20,000 b/d. The column started up and produced between 12,000 and 15,000 b/d, but did not achieve design yields.

The structured packing bed was presumed to have failed to meet the required efficiency and initial theories within the plant centered on possible poor design and/or application of the structured packing.

Many computer simulations were run, but without the benefit of field observation and resultant temperature and pressure profiles, the problem was neither identified nor solved.

SOLUTION: Glitsch suggested field observations and data collection. Field observation showed that the LVGO product quality did not respond (neither improved or got worse) to increased reflux, and LVGO product yield changes did not materially impact the quality over a relatively large range of product yields.

Pressure drop measurements were taken with an absolute pressure manometer and the observed pressure drop was normal (delta P < 0.5 in. liquid/ft packing) for the particular size of structured packing.

The column pressure was extremely stable during the pressure drop measurements. (The manometer used was accurate to about 1 mm mercury.)

Pressure drop in a packed bed is not a reliable indicator of flooding unless it is measured very accurately. The column pressure must be very stable, as it was in this case, and/or two measurements should be taken simultaneously.

In this instance, because column pressure was stable, it was determined that the packed bed was not flooding and that the cause of the problem was elsewhere.

A test run was performed to provide accurate heat and material balance data for the evaluation of the overall vacuum unit operation. The furnace fired duty and efficiency were calculated to check the observed column heat and material balance data.

A computer model was used to evaluate the data. The use of computer models at this stage is appropriate because the input data, obtained in the field, are well-grounded.

Fig. 3 is a sketch of the column showing the general flow scheme. The wash oil is subcooled. This is generally considered poor design practice but because the existing flow scheme in the trayed column was designed this way, subcooling was specified by the refiner.

The overflash flows internally, therefore the material balance shown in Table 1 indicates the wash oil rate and not the true overflash rate. (In this example, to estimate the flash-zone vapor rate, assume the overflash rate is 75% of the wash oil flow rate.)

Table 1 presents the material balance and selected heat balance data for the column test run.

The following observations were made concerning the heat and material balance:

The total flash zone vapor, as calculated by the column material balance (above flash zone), did not match the overall heat and material balance. The absorbed furnace duty, as calculated from the furnace firing and efficiency data, was 15.0 MMBTU/hr higher than the column heat and material balance indicated.
The total flash zone vapor as calculated from the column heat balance (LVGO and HVGO pumparound duties), assuming LVGO and HVGO product rates were correct, showed a much higher liquid rate to the wash zone than the measured wash rates.

Total flash zone oil vapor = Slop oil + LVGO + HVGO + Overflash

The calculated absorbed duty from the model and the calculated absorbed duty from the furnace data matched within 5%. There were 15.0 MMBTU/hr unaccounted for in the column's internal liquid/vapor rates, assuming the measured wash rate was correct.

The above discrepancies indicated that an additional test run should be conducted, with all the pertinent meters zeroed and calibrated.

The results of the second test run confirmed the earlier analysis. The column had an overflash rate 10,000 b/d higher than the wash oil rate indicated. However, this was not the immediate concern of the refiner, because it indicated that the problem was not simply caused by the structured packing revamp.

The very low LVGO product yield indicated that the column was operating as if there were no fractionation zone (almost zero efficiency) between the LVGO and HVGO products. A heat and material balance around the LVGO pumparound was calculated to determine the internal reflux rate (Fig. 4).

Quick calculations on a calculator showed the internal liquid rate to be almost twice the design rate. The liquid distributor was designed for 850 gpm, but was running at 1,358 gpm.

The design and test run data are shown in Table 2.

With an internal reflux rate (calculated using actual field measurements) at 1,358 gpm, it was clear that the liquid distributor was overflowing and not properly distributing liquid to the structured packing.

Glitsch recommended that the top pumparound duty be reduced to 85.0 MMBTU/hr so that the internal reflux rate could be reduced to 750 gpm. The HVGO pumparound duty was then maximized.

RESULTS: Within 8 hr of the recommended changes, the LVGO product yield increased to 20,000 b/d. An online calculation (simplified version of Equation 2, Fig. 4)--using the measured pumparound flow rate, measured pumparound delta temperature, and the measured LVGO product rate--is now used to calculate internal liquid rate. This is displayed to the operator so the distributor does not overflow.

OBSERVATIONS: This problem was identified with a simple heat and material balance around the top pumparound, although a computer simulation could have reached the same conclusion had the entire performance of the column been evaluated using field observation.

The ultimate cause of the poor fractionation was a liquid distributor that was not designed for the full range of crudes that were charged to the unit.

Liquid distributors in large diameter columns should be based on the lightest and heaviest crude slate, and not on an arbitrary turndown value. Liquids distributing in large diameter columns have an efficient turndown to 50% only.

CASE 2

DESCRIPTION: Poor lube column fractionation.

PROBLEM: Following the routine shutdown of a 30,500 b/d lube column, poor fractionation between some of the cuts was observed.

HISTORY: This column had several previous operating upsets, and "explosion resistant" trays had been installed in some sections. As a first step toward solving this most recent fractionation problem, the column had been gamma scanned.

Because conclusions were that the trays were intact and operating properly, the initial assumption was that the feedstock had changed and the trays must be operating outside their acceptable range. Pressure taps were installed between each section of the column.

SOLUTION: A column pressure survey was performed using an absolute pressure manometer to determine if tray damage had occurred during a previous shutdown. The pressure drop measurements refer to Fig. 5 with the measurements taken at the locations shown. The data are shown in Table 3.

A properly designed tray with a weir height of 2 in. will have a pressure drop of 3-5 mm Hg. Therefore, it was clear from the pressure drops measured on the LN/MN, MN/HN, and HN STR sections (Table 3) that there was tray damage.

The poor fractionation between the LN/MN, MN/HN, and HN STR cuts was caused by this tray damage, as was indicated by the low pressure drop. But with gamma scan results indicating no tray damage, the refiner was reluctant to open his tower again without more evidence.

A computer simulation was performed to evaluate the observed fractionation, as it compared to the simulated results from the computer model. Additionally, the theoretical tray pressure drop was calculated, assuming the trays were intact. Table 4 shows the test run product distillation overlaps and the simulated distillation overlaps.

The measured pressure drop and the hydraulic calculations based on the vapor/liquid loadings from the simulation are shown in Table 5.

There was a significant difference between the measured pressure drop and the pressure drop calculated using computer simulation. The most sophisticated tool was the gamma scan and it showed the trays to be active and intact, although the conclusion based on the simplest criteria (pressure drop) was that there was damage to three sections of the column.

RESULT: Glitsch recommended that the column be shut down and the trays repaired. When the column was shut down, these three sections of the column indeed had significant tray damage.

OBSERVATIONS: The more sophisticated techniques used were calculated pressure drop across the tray vs. measured pressure, and observed actual fractionation efficiency as measured by the model. These techniques supported the theory that tray damage was the culprit.

However, the quickest, simplest, and most reliable indication of tray damage was a manometer pressure drop measurement and subsequent pressure drop calculations.

CASE 3

DESCRIPTION: Low HVGO product recovery.

HISTORY: A 48,000 b/d fuels vacuum column was slated for revamp to improve the heavy vacuum gas oil yield. The HVGO was used as a feed to a lube oil hydrotreater with subsequent downstream vacuum fractionation to produce lube oil blendstocks.

The HVGO product was a high profit margin cut compared to the vacuum residue (which was feed to a coker unit). Thus, the refiner wanted to improve HVGO product yield.

PROBLEM: Would the proposed revamp produce the desired HVGO product yield?

SOLUTION: A project was initiated to evaluate the feasibility of increasing HVGO yield by increasing the flash zone temperature and reducing the flash zone oil partial pressure by conversion of the column from a "dry" column to a "wet" column. (Note: A dry column has no stripping section and no coil steam, and a wet column has a precondenser, limiting column overhead pressure to cooling water temperature limitations).

A sketch of the column is shown in Fig. 6, with the operating data. The stream data were:

LVGO product--197,782 lb/hr
HVGO product--209,549 lb/hr
Wash oil rate--139,194 lb/hr
Slop oil recycle--81,010 lb/hr
Vacuum residue--265,960 lb/hr.

The heat balance data calculated from the plant pumparound flow rate and delta T data were:

LVGO pumparound--44.0 MMBTU/hr
HVGO pumparound--82.0 MMBTU/hr.

The collector trays on this fuels vacuum column were designed as total draws. The heat and material balance for the LVGO pumparound zone on a dry vacuum column is shown in Fig. 7.

The pumparound duty is the difference between the vapor enthalpy entering the bed and the liquid enthalpy of the LVGO product, multiplied by the LVGO product rate. For this example, the calculated duty from the pumparound flow and temperature difference equals the duty calculated from Equation 3, Fig. 7:

QL VGO

= flow rate x enthalpy

= 197,782 lb/hr x 230 BTU/lb

= 45.58 MMBTU/hr QL VGO

= M x CP x DELTA T

= 44.0 MMBTU/hr

A test run was performed to evaluate current operation and determine potential yield improvements that could result from the proposed revamp modifications. The data evaluation revealed inconsistencies in the column's measured heat and material balance.

The data shown in Fig. 8 indicated an HVGO pumparound duty of 82.7 MMBTU/hr, as calculated from the pumparound flow and the delta T. If the HVGO pumparound duty is calculated by assuming no liquid flow from the LVGO pumparound draw tray to the HVGO bed, then the duty is 64.0 MMBTU/hr.

Fig. 8 shows the heat and material balance used to determine the consistency of the measured wash oil rate and the calculated rate. The measured wash oil flow rate was 139,194 lb/hr. Using Equation 4 (Fig. 8), the calculated wash rate would be 281,598 lb/hr.

The 142,404 lb/hr difference represents 18 MMBTU/ hr. This corresponds to an overflash rate 63,000 lb/hr higher than the measured overflash. The HVGO product losses were 63,000 lb/hr. These losses were attributed to a leaking HVGO collector tray and an overflowing or leaking slop wax tray.

Glitsch recommended that both trays be repaired before proceeding with the revamp or the projected HVGO yield increases would be seriously compromised.

OBSERVATIONS: Leaking or overflowing collector trays are primary causes of yield losses and should be corrected before a major capital project is considered.

TROUBLESHOOTING

The following are some general comments on vacuum column troubleshooting:

Fuels vacuum columns typically have LVGO, HVGO, and slop wax collector trays designed for total product draw. In practice, this is not always the case and a heat and mass balance around the section often show significant tray leaks--sometimes larger than 20% of LVGO product draw for an LVGO pumparound section.
Many total draw trays (especially the slop wax collector tray) overflow because high liquid level may remain undetected due to faulty level indicators. This is easily checked by placing the product draw flow on manual and making discrete (+10%) changes while monitoring the pump for cavitation.
Keep making discrete changes until the pump begins to cavitate. Any inaccuracies in the level indicator can then be verified, thus confirming that the trays are overflowing.
Calculate the metered overflash on all fuels vacuum columns:

Metered overflash = Wash liquid not revaporized + Entrained liquid + Liquid condensed on underside of HVGO collector tray

Do not consider the measured slop wax rate a measure of the quantity of wash oil not revaporized. It should not be used to control wash oil rate to a minimum.

Entrainment is a function of the flash zone design and is difficult to measure. However, when the slop wax contains greater than 70% residue, entrainment is probably a significant factor and de-entrainment measures should be considered.

The condensation on the bottom of the HVGO collector tray is high on fuels vacuum columns, having an HVGO draw temperature less than 500 F. The vapor from the wash is typically higher than 700 F., therefore the driving force for condensation is large. The HVGO draw tray should be insulated if high cutpoints and low overflash are goals.

GUIDELINES

Some general tips when troubleshooting vacuum columns are:

Obtain an accurate column pressure survey using an absolute pressure manometer. If necessary, two manometers can be used simultaneously to measure low pressure drop across packed beds. In trayed columns, pressure drop measurements are always the most reliable means of evaluating tray problems.
Obtain an accurate unit heat and material balance. This should include:
A.Material balance
B.Heat balance, to include the fired furnace duty and efficiency, and pumparound duties from both sides of the exchangers, where possible.
The collector trays on vacuum columns should never be assumed to be leak-free on an operating column. Heat and material balance data can be used to find significant leaks. After each shutdown, the collectors should be leak-tested with water. Any leaks observed below the tray should be repaired.
Liquid distributors designed for fractionation beds must be able to distribute the liquid without overflowing. In actual operation, it may be necessary to adjust the column heat balance if the distributor is overflowing.