Oct-2016

Troubleshooting vacuum systems

Case histories of refinery vacuum tower problems solved are gleaned from five decades of experience.

NORMAN LIEBERMAN
Process Improvement Engineering

Article Summary

The objectives of operating a refinery vacuum tower are typically to produce asphalt of proper viscosity or to minimise the heavy virgin gas oil content of the vacuum residue going to the delayed coker feed or to industrial fuel oil. In many cases, the ability to achieve these objectives is the largest factor in determining refinery profitability. Principally, the main objective of the refinery operator is to restore the degraded vacuum in the vacuum tower flash zone (see Figure 1). The following case histories illustrate the problem that occurs in many, if not most refineries at the start of every summer. The problem is that both ambient air and plant cooling water temperatures increase. This results in the first stage ejector surging, a step change degradation in vacuum, and vacuum tower pressure instability.

Increase in vacuum tower top temperature
In one 40 000 b/d crude unit at an East coast US refinery, producing delayed coker vacuum resid, during most afternoons starting in July the first stage ejector began to surge and continued to do so until the late evening when ambient temperatures declined. An analysis of the off-gas from the third stage ejector (see Figure 1) indicated a high concentration of both iso- and n-pentanes. As the C5 olefin content remained low, we concluded that the first stage ejector was being overloaded with light virgin naphtha components originating in the crude – that is, in the vacuum tower feed from the crude atmospheric tower bottoms. Apparently, the recovery of light naphtha in the crude tower from the atmospheric residue was poor. In cooler weather, when the LVGO pumparound return was also cool, the tower top temperature (T1 in Figure 1) was cold enough to absorb most of the lighter virgin naphtha components.

However, at higher ambient temperatures, the LVGO pumparound trim air cooler was unable to suppress the vacuum tower top temperature sufficiently to condense out enough of the virgin pentanes and hexanes to avoid overloading the first stage ejector.

To correct the problem on-stream, we improved the operation of the LVGO cooler by: tightening slipping fan belts; increasing the fan blade pitch; adding water mist nozzles to humidify and cool the air to the tube bundle (we did not spray water on the tubes themselves); brushing off the underside of the tubes with a broom; and adding vane tip seals to reduce air recirculation.

Adding water mist nozzles should be minimised since actually wetting the tubes will corrode the fins. This measure was only employed as needed for several hours on hotter days.

Note that the fundamental problem, which has yet to be corrected by the refiner, is poor stripping tray efficiency on the bottom six trays of the crude atmospheric tower. Likely, the trays are upset as the observed pressure drop across the trays was extremely low.

High first stage ejector exhaust pressure
On another 35000 b/d crude unit, the viscosity specification for the paving grade asphalt product of about 10000 b/d could not be achieved. The problem again was a step change loss in vacuum during periods of increased cooling water temperature. The cooling water was drawn from a local river. Hence the problem persisted, even at night when ambient temperatures dropped.

A field pressure survey indicated that the pressure at P1 (see Figure 1) was 20 mm Hg above the design critical discharge pressure. An elevated pressure at the discharge of the first stage jet will cause the jet to lose its ‘sonic boost’.1 Alternatively, the jet loses that portion of its compression ratio due to being forced out of its critical mode of operation by excessive back pressure at P1.

To correct this problem, through a trial and error procedure, we found that reducing the motive steam pressure to the second stage jet, from 180 psig to 140 psig, lowered the first stage jet discharge pressure at P1. This restored the sonic boost to the first stage jet and hence reduced the vacuum tower top pressure. This permitted the asphalt product viscosity to be restored to meet paving grade asphalt specifications.

Note that the design motive steam pressure to all three ejectors was 180 psig. The reason why the performance of the second stage ejector improved, as the steam pressure was reduced to 140 psig, was due to:
• The steam nozzle (see Figure 2) inside the second ejector had become loose
•  The nozzle itself was constructed from 316 stainless steel
• The body of the ejector was carbon steel
• The motive steam was wet (which in itself is a problem)
• Galvanic corrosion then caused the threads in the ejector body (but not the nozzle itself) to erode
• Motive steam then partially bypassed the steam nozzle.

After the summer, when this failure was corrected in the autumn turnaround, the design 180 psig motive steam pressure to the second stage ejector was found to be optimum. To partly avoid this problem in the future, Teflon tape was wrapped around the threads of the steam nozzle, which may retard the rate of galvanic corrosion.

There are many other examples of vacuum system problems, in both vacuum towers and condensing steam turbine surface condensers. We have assembled a summary of malfunctions which represent the most common problems we have seen over a period of five decades in troubleshooting and retrofitting refinery and petrochemical plant steam ejector vacuum systems.

The general lessons that we have learned are that conventional descriptions as to how converging-diverging steam ejectors function do not always correspond with the actual operation of vacuum tower ejectors. This is based on observed external temperatures along the jet. Most vacuum system malfunctions originate with problems external to the ejectors and with the ejector steam nozzles.

Seal drum sludge
This problem at a plant in Wyoming involved sludge accumulation when the seal drum covers the bottom of the seal legs. The condensers then could not drain properly and filled with condensate. The standard design for the seal drum is to have the bottom of the legs 4in (10 cm) above the bottom of the seal drum. The correction here was to cut off the bottom 12in (30 cm) of the seal legs. A temporary fix was to blow the seal legs out with steam.

Loose steam nozzles
The steam nozzles at a plant in Louisiana became loose where they screw into the body of the steam ejector. Sometimes this was due to corrosion of the carbon steel ejector (the nozzle itself is stainless) and sometimes the nozzle just came loose without any apparent corrosion.
The correction was to wrap the steam nozzle threads with Teflon tape, use larger washers to maintain proper spacing between the steam nozzle and diffuser inlet, and tighten the nozzle forcefully up against the spacers. The back end extension of the ejector body should be stainless and not carbon steel. A temporary fix was to reduce motive steam pressure below the design motive steam pressure.

Plugged steam nozzle
Moisture carry-over from local waste heat kettle boilers at a plant in South Africa contained silicates that then deposited in the throat of the steam nozzles, restricting steam flow through the nozzles. The correction was to employ a lower level in the waste heat boilers. There is a clean-out plug in the back of the ejector. A temporary fix is to unscrew the plug and wire brush out the steam nozzle. This cannot be done with the ejector in operation.

Reverse flow of motive steam
Reverse flow only occurred when discharging dual parallel ejectors to a common condenser at a plant in Texas. Steam flow from the discharge of the stronger ejector flowed backwards through the weaker ejector into the suction of the stronger ejector, which was then overloaded.

The problem usually involved hardness deposits in the throat of the weaker ejector steam nozzle. A temporary fix is to block in the process side of the weaker steam nozzle.

Leaking condenser channel head pass partition baffle
Closure between the channel head tube sheet and pass partition baffle in the first stage condenser was eroded out by sand in the cooling water at this plant in Alabama. Water then by-passed the condenser tubes.
Correction involved using silicon sealer along the edge of the channel head pass partition baffle.
And a temporary fix was to inject shredded fibreglass into the cooling water to the condensers.