Search

Jeff Lewellen (HollyFrontier)

In 2009, we completed a group of projects at El Dorado involving larger coke drums and a deep cut vacuum tower. The coke drum project limited the scope to drum replacement with minimal changes to the balance of the unit. Some of the more significant changes to the unit included:

Fractionation/Wet Gas Systems -

• Higher wet-gas volumes due to higher coke yields.

• Larger swings in gas rate due to drum switching activities. Unit Operators reported this was a significant change in operation.

• Some shift in liquid product yields affected by slightly lower drum pressures, lower drum outlet temperature, and longer drum cycles.

Charge Heater -

• Lower outlet temperature and fired duty requirements to achieve target VCM/coke quality. This is mostly influenced by longer drum cycle times, and less HGO in the feed.

• Increased tube fouling rate leading to shorter times between heater tube decokings.

Drum and Drum Cycle -

the drum volumes approximately doubled with required cycle length extending from 12 to 18 hours.

• Drum superficial velocity decreased with drum diameter, which improved carry-over and entrainment issues (no significant pressure changes).

• The reduced velocity also decreases foam height. However heavier feed and lower drum temperature created more stable foam that can be difficult to collapse.

• The large drums also lengthen drum air freeing, pressure test, and warm-up cycles.

Quench/Blow-down Systems - Our modifications to this system were limited to additional quench pump and air cooler/condenser capacity. The quench tower was also reconfigured with limited success.

• More quench water needed leading to more storage (surge capacity)

• Higher sour water production – proportional with coke yield

• Higher off-gas rates

• More quenched oil/slop from the system

• Marginally sized quench system struggles.

Coke Storage and Handling - We replaced railcars with a feeder-breaker/conveyor system due to logistics of the additional rail and cars. The heavier feed also shifted our coke to much more shot coke production leading to more drum fallouts that are difficult to manage with a railcar system.

• Greater surge volumes are needed in the

o Pit/slab/bin storage area

o Crusher/slurry pump, feeder-breaker/conveyor, or bridge crane/loader systems

o Shipping logistics – railcars, truck, barge

Jet Pump/Cutting system – We operated for several months with the original low pressure/volume jet pump system until the new unit was delivered and installed.

• The smaller jet pump/cutting system led to a significant increase in coke fines to the jet pump system due to extended cutting times. Cutting times more than doubled from previous drums. Extended cycle times or unexpected feed changes led to even greater cutting durations.

o Significant reliability problems developed from erosion issues in the decoking valve, cutting tools, and eventually the jet pump.

o Restrictions in instrumentation and small-bore piping from the additional coke fines.

• After the jet pump/cutting system upgrade was complete, the reliability and the operation of this system improved greatly. However, some of the changes impacting this operation include:

o Decoking rates are limited by bottom unheading device/dump chute plugging limit. Cutting rates are higher in tons per hour, but duration remained longer than the smaller drums.

o the longer cutting times (and potentially changes in feed and coke handling) led to more fine's generation than original drums. However, much improved over the previous 6 months.

o Retrofitted additional settling/fines separation capacity to improved water quality.

Gary Gianzon (Marathon Petroleum Company)

The larger drum can take longer to quench and cut assuming that your blowdown system and cutting water system are not designed for the larger drums.

Eberhard Lucke (Commonwealth E&C)

The extra capacity gained by installing larger diameter coke drums can be used by either increasing the unit throughput or by taking advantage of a longer drum cycle time at the same unit charge rate. Due to the larger diameter, steam stripping and quenching can be more challenging and the probability of hot spots may increase. Before installation of larger diameter coke drums special attention has to be paid to the jet pump pressure. If the jet pump discharge pressure is lower than the recommended pressure for the new coke drum diameter, coke cutting will take a lot longer and will create significantly more coke fines due to the grinding effect of the broader water jet. Jet pump and drill/cutting tool may have to be replaced or revamped to ensure proper coke cutting. Also, the capacity of the coke handling system – pit, pad, attached breaker or feeder-breaker system, slurry system – receiving a larger amount of coke has to be checked for adequate capacity.

Gary Gianzon (Marathon Petroleum Company)

Coke drum velocity limit depends on the capability of the coker main fractionator and downstream equipment to handle entrained coke. One of MPC’s units, processing much lighter feed than design and at a higher charge rate, currently operates with a drum velocity above 0.80 ft/sec. At these velocities, the flash zone gasoil strainer needs to be cleaned twice per week and the backwash interval on the HCGO filters is less than ideal. Raising coke drum pressure will reduce velocity and coke entrainment. Two of our cokers with no coke removal capabilities in the fractionator, operate with a drum velocity of around 0.40 ft/sec. At this low velocity, accumulation of coke in the main fractionator is minimal and only requires cleaning every two years.

Jeff Lewellen (HollyFrontier) 

The El Dorado facility targets coke drum velocities in the 0.5 ft/sec range. In troubleshooting coke carryover issues, a frequently overlooked contributor to high drum velocity can be unaccounted (or under accounted) sweep steam sources. Wet steam caused by steam system upsets or faulty condensate traps can also unknowingly increase drum (and fractionator) velocities.

Eberhard Lucke (Commonwealth E&C)

As a general design guideline, the vapor velocity in the open section above the coke bed should be in the range of 0.5 to 0.6 ft/s while the vapor velocity in the inlet of the vapor line should not exceed about 60 ft/s.

Jeff Lewellen (HollyFrontier)

The ideal coke drum level indication would provide accurate and reliable information to the unit operator for all of the following:

• The foam front level for anti-foam/silicone addition control

• Coke/hydrocarbon level to optimize drum use

• Quench water level to verify (along with pressure/temperature/flow totalizer) adequately quenched drums This is also a demanding environment for instrumentation.

The process conditions are highly fouling, high temperature, changing pressure and composition, and high velocity hydraulic decoking every cycle. Most common technologies use “non-contact” methods with a radiation source and detector system. These include:

• Neutron backscatter

• Gamma point source (density and/or level switch)

• Gamma continuous level

Neutron backscatter

These devices use a common neutron source and sensor housing that directs fast (high energy) neutrons from typically Americium-Beryllium (AmBe) or similar neutron emitter source through the vessel wall into the vessel interior. If hydrogen bearing material is present, the fast (high energy) neutrons are converted into slow (low energy) neutrons which are scattered back to the neutron sensor in direct proportion to the hydrogen density.

Due to significant differences in hydrogen density, the technology is very effective in detecting changes from clear vapor to foam (including light to heavy foam densities) to coke level and detecting water level. The limitations of this technology are:

• Point detection only.

• Measures only the area immediately adjacent to the vessel wall.

• Historically has experienced difficulty with thick wall vessels. Trials at our facility in the 1980s were unsuccessful. However, these detectors are used very successfully throughout the industry.

Gamma point source

This technology utilizes a gamma emitter source (usually Cesium-137) shielded to emitting gamma radiation only across the diameter of the drum with a detector located opposite of the source. The gamma radiation reaching the detector is inversely proportional to the mass of material between the source and sensor. The detector output can be reported as either:

• An analog signal representing density changes in vapor to foam to coke/liquid.

• Or a digital signal between clear vapor and the foam/coke level.

Our previous drums were equipped with this technology utilizing 5 detectors per drum at various levels. The indications were fairly reliable; however, lightening/electrostatic discharge from thunderstorms did cause some problems.

This technology has (and continues to be) used throughout the industry in multiple applications.

However, there are some significant disadvantages:

• Does not detect foam height. Addition of anti-foam may cause the level to drop below the detector, but how far and for how long?

• Predicting drum outages and switch times are difficult with changing rates or feed composition. Extrapolating drum fill rate between level detectors may not provide adequate time to make unit adjustments.

• We have experienced low density foam fronts that did not activate the level indicators resulting in a drum “foam-carryover” event. The foam was confirmed by a contractor scanner, and detector sensitive were adjusted to account for the change.

Gamma Continuous Level

Similar to the point source, this method utilizes a gamma emitter source that is shielded to project a fan shape beam diagonally across the drum. The detector is a lengthened scintillation tube or (more recently) a fiber-optic scintillation bundle. Multiple sources and detectors can be used to expand the range of the level indication to provide a continuous indication for the drum.

Items to note on this technology:

• The level is interpreted from the amount of gamma radiation blocked. It does not directly detect the vapor/foam/coke interface. However, a handheld detector can be used to find this interface if verification is needed.

• Level indication is normally the combined coke and foam level. The foam portion is determined by collapsing the foam front using the silicone anti-foam and monitoring the change in level.

We have installed this technology in our El Dorado facility. In addition to the fiber optic continuous levels, the system utilizes bottom of range point detector to reset zero and a density point detector at top of range for both vapor density correction and the LAHH redundancy to the continuous level.

We have noted a deviation between the level indication at the end of the drum cycle and the level determined by drill stem gauging the coke outage at the beginning of the drum decoking. This deviation at times has been several feet. At these times, we have confirmed coke to be higher on the walls with a depression toward the center of the drums. We have attributed this, in part, to the bed collapsing during the quench/draining steps of the drum cycle.

The advantages of the continuous level technology are best seen when utilized in combination with the vapor density and low-level detectors as a packaged system.

• Foam level is inferred by the change in level with the introduction of antifoam/silicone

• Coke/hydrocarbon level is continuous for the full cylinder length of the drum.

• Redundant high-level indication Probably the most significant disadvantage of this system is the complexity. Although this equipment has been very reliable, it has been in service for less than 3 years.

Gary Gianzon (Marathon Petroleum Company

MPC currently uses both neutron backscatter and gamma continuous level detection and here are the pros and cons based on our operating experience: Gamma continuous level detection.

Pros:

1. Measure the level continuously throughout relevant levels of the drum. We use the continuous level measurement to optimize antifoam usage. The level can also be used to determine the height of the foam front, and this information can be used to adjust furnace outlet temperature.

2. Continuous level detection measures across the drum so it can detect the peak level in the drum.

3. Measures the vapor density across the top of the coke drum, which we use to optimize steam-out time to the fractionators and can indicate foam carryover to the main fractionators. 4. Smaller source than the neutron backscatter which is easier to permit.

Cons:

1. Difficult to calibrate and easily gets out of calibration. This is quite problematic if this is your only source of drum level measurement.

2. Only measures the total level, cannot distinguish between foam, coke, and water. Neutron Backscatter

Pros

1. The neutron backscatter can distinguish between foam, coke, and water.

2. The level is exact and does not drift.

Cons

1. Level can be difficult to detect on very thick drums.

2. Point source only detects about a foot inside the drum, level measurement is not continuous across the drum.

3. Long half-life responsibility

Rajkumar Ghosh (Indian Oil Corporation)

Volatile combustible material (VCM) is an important parameter of Petcoke. VCM is basically unconverted pitch in the coke. The metal and sulphur are controlled by the type of crudes processed, but VCM content of coke is mainly in the control of DCU operators. The obvious benefit of driving green coke VCM to a low level is more distillate yield and less coke yield. Each 1 wt% reduction in VCM could increase refinery GRM by 5 cents/bbl. Moreover, lower VCM implies a relatively harder coke and lesser generation of undesirable fines during the coke cutting. Our objective is to achieve the coke VCM in the lower range of around 9%wt. for fuel grade coke. Coker operating parameters such as coke drum overhead temperature, pressure, COT, recycle, coke drum cycle time and steam quench rate play prominent role in determining VCM in the coke. Some of the rule of thumb for Coker variable on coke VCM are:

• With each 5o C rise in coke drum vapour temperature, VCM in coke decreases by 1 wt%.

• Decreasing cycle time by 6 hrs, increases coke VCM by 1.0 wt%.

• 2-3o C increase in COT reduces coke VCM by 1 wt% Comparison of VCMs in 18hr and 24hr cycle is given below:

As a best practice for reducing coke VCM, we increase the COT by 2o C, an hour prior to drum switch. We continue with the higher temperature through the drum switch and get back to normal COT once the receiving drum top temperature reaches near normal value. Steaming of the offline drum is another factor that impacts VCM.

If the steaming of the offline drum is not done for sufficient time with sufficient amount of steam, this may contribute to an increase in VCM. The steaming amount depends on the drum diameter. For a typical 28 ft drum to a larger diameter drum i.e. 32 ft; steam flow is varied in the range of 15–18 MT/ hr. These figures have been arrived at based on our experience w.r.t. adequacy of Coke Bed cooling & VCM of the Coke. It was also observed that on reducing the amount of steam to around 10 MT/ hr, there was an increase in VCM of approx. 0.5-1 wt%.

Flip side of very low VCM is that the coke becomes very hard and coke cutting operation consumes longer time, resulting in delay in the coke drum cycle. Also, very hard coke (low Hardgrove Grindability Index, HGI) makes it difficult for the customer to crush and use and hence less acceptable to them.

Lowest level of coke VCM achieved at our Cokers is 8.5 to 9 wt%. Typical VCM level that we are achieving on sustained basis is around 9-10 wt%.

Jeff Lewellen (HollyFrontier)

I agree with the primary answer. For fuel grade cokers, optimizing VCM is primarily an economic decision balancing liquid yield with unit operation. Our approach to optimization is:

o First, optimize coke drum stripping steam volume and duration for recovery, and limited by drum and fractionator velocities and available cycle time.

o Second, target coil outlet temperature (COT) to minimize VCM while maintaining required drum cycles and heater run lengths.

Optimum VCM target is somewhat feed dependent with our normal range in the 8 – 10 wt% range.

Eberhard Lucke (Commonwealth E&C)

The drivers to lower the VCM of the green coke can vary depending on what kind of operation you run. Fuel grade coke operations typically don’t monitor VCM on a regular basis. The only advantage in lowering VCM would be to reduce the loss of very heavy gasoil material with the coke product. But the required measures to reduce VCM (longer steam stripping, more stripping steam, or higher coil outlet temperature etc) and associated cost most likely exceed the gains. In anode grade operation, the VCM has a direct relation to the vibrated bulk density (VBD) of the calcined coke and therefore, controlling and minimizing the VCM content of the green coke is crucial to a successful anode coke business. In the past we achieved best results by maintaining a rigorous steam stripping regiment and by implementing a temperature ramp function for the coil outlet temperature that allows to drive more volatile material out of the coke bed before the drum will be steam stripped. With all these measures in place we achieved consistently VCM values around 8wt% and consequently very good VBD results in the calcined coke. The price is a slightly higher energy consumption in the charge heater which can be compensated by further optimization of the control scheme for the coil outlet temperature. Other tests with lowering the drum operating pressure didn’t show the same effect and couldn’t be repeated on a consistent basis.

Tom Lorsbach (UOP)

As introduction, Propylene Recovery Units downstream of FCC units typically consist of a C3/C4 Splitter, a Deethanizer and a C3 Splitter in series. The propylene from the C3 Splitter overhead is treated in regenerable molecular sieve Driers for moisture removal and finally passed through Guard Beds for trace contaminant removal. The principle trace contaminants being removed by the guard beds are carbonyl sulfide, arsine and phosphine. There are a variety of guard bed adsorbents that are used. Metal oxide bound on an alumina support is one type of adsorbent often used for trace contaminant removal. Occasionally hybrid type adsorbents are utilized (downstream of the driers) to remove trace levels of oxygenates, organic sulfur species and nitrogen compounds.

With regard to the molecular sieve driers, there is a small (10-15°F) adsorption exotherm when liquid propylene is first charged to the molecular sieve adsorbent. This small exotherm dissipates quickly as propylene flows through the drier beds. Exothermic runaways are unlikely on molecular sieve type moisture driers operating normally to remove dissolved moisture from liquid phase propylene. Elevated levels of diolefins, oxygenates or metals could potentially cause exotherms in the molecular sieve drier.

During start-up and operation of any adsorption system it is important to stay within the pressure and temperature limits set by the equipment and adsorbent manufacturers. Excessive temperatures can cause equipment failure and result in life-threatening fire or explosion. There are a number of factors that can cause excessive temperatures in adsorbent systems.

1. Introducing a flammable or reactive fluid into a vessel containing air.

2. Introducing a high concentration of a reactive, strongly adsorbed material into fresh or regenerated adsorbent. In applications where there is sufficiently high risk of thermal excursions a low reactivity adsorbent should be considered.

3. Using a highly reactive fluid to heat or cool the bed. Examples of reactive fluids would include ethylene, propylene and other olefins.

As mentioned above, when equipment and adsorbent operating guidelines are followed closely, there is normally a small adsorption exotherm during initial commissioning and this exotherm dissipates quickly as propylene flows through the drier beds.

There have been two incidents of uncontrolled exotherms occurring in FCC PRU adsorption beds.

1. In one case an operator changed the temperature permissive from 122°F to 338°F (50°C to 170°C) in the drier regeneration control system, allowing excessively hot propylene from the freshly regenerated drier vessel to enter the lead COS/Arsine guard bed vessel. The high temperature propylene initiated highly exothermic auto-reduction of the metal oxide to the elemental metal through both guard beds in series. The exotherm caused the paint to blister on the vessels and piping and the guard bed adsorbent was fused, but the system did not lose containment.

2. The second incident was more serious in that the vessel (containing stacked beds of two different adsorbents) lost containment. The vessel operated up flow. The inlet section of the adsorbent bed contained an adsorbent for removal of moisture and trace oxygenates. The exit section of the adsorbent bed contained adsorbent selective for the removal of COS. The investigation of this incident identified vapor phase propylene (instead of liquid phase propylene) being sent to the vessel for 67 minutes, with concomitant bed channeling, high heat of adsorption in the inlet section (where a reactive adsorbent was installed) and poor heat dissipation as the root cause of the failure. High local temperature and stresses exceeding the tensile strength of the vessel shell led to rupture of the shell near the end of the inlet section of adsorbent.

Dwight Agnello-Dean (BP)

Propylene is a reactive fluid and therefore a known exothermic reaction is propylene polymerization which can be catalyzed by the adsorbent material. Typically, these reactions are not significant and are not observed if the adsorbent material is proper for propylene use and the heat of adsorption during startup is adequately dissipated. At higher temperatures these reactions can accelerate rapidly.

Within our network there have been few reports of incidents involving propylene driers and the investigations of these events have not identified common root causes. Learning’s from these events include ensuring the proper adsorbent material is selected and that startup procedures include temperature increase limits with corrected actions if exceeded.

Matthew Meyers (Western Refining)

The panel concluded that 100% of the FCC units in their respective companies used the low slide valve dP over-ride to protect against reversals. The consensus of the panel was that having dual slide valves did not significantly increase the reliability of the system. The IEC 61511, 2003 definition of an Independent Protective Layer is the following:

1. The system must have a risk reduction factor of at least 100 (10-2).

2. It must be designed to protect against a specific event

3. It must operate independently of other protective layers (i.e., No common causes of unavailability)

4. Must have an availability of 0.9 or greater

5. Must be testable

Since the typical slide valve DP override system shares the low signal selector algorithm in the DCS with the regulatory control system, it cannot meet the strict IEC definition of independence. The control algorithm (or changes thereof) itself may be a common cause of unavailability.

The probability of failure on demand (PFD) for the slide valve DP override system depends upon a number of factors:

• Standpipe fluidization and stability

• Redundancy of DP transmitter field-side: all process taps and tubing are independent with reliable purges

• Each transmitter configured to fail low and SIL rated to prevent incorrect configuration changes

• Wiring from each transmitter to the BPCS in separate home run cables with redundant inputs to the BPCS

• BPCS configured to fail low on loss of signal from either DP transmitter

• Constant slide valve modulation and testing with good slide valve maintenance

• Dual slide valve hydraulic pumps with separate motive force sources (e.g., electrical and pneumatic)

• Slide valve skid alarm instrumentation to ensure availability: reserve accumulator in service, reserve accumulator pressure, position deviation alarms, loss of tracking, loss of feedback, etc.

All of these factors should be carefully considered before determining a PFD. Assuming each item is properly maintained and suitably stable over the course of the operating history, a PFD of 10-2 is achievable.

In terms of conducting a LOPA, the slide valve DP override system may be used to reduce the impact event frequency and thereby possibly reduce the risk. This may or may not affect SIL requirements, depending on other layers of protection that may be taken into account and the refinery-specific risk ranking matrix.

Mike Teders (Valero)

Low slide valve DP over-ride is used in every one of the Valero FCC units where a slide valve is used to control the catalyst circulation. The low dP over-ride is DCS based in our FCC units and we consider them to be an effective form of process control. In addition, we are developing a prescriptive design standard for an independent FCC shutdown system for our FCC units. The standard includes protection against a catalyst reversal by using low slide valve differential pressure to trigger a slide valve closure and trip the unit into a shutdown. Our FCC shutdown system is intended to be independent from the DCS based over-ride and would qualify as an Independent Protection Layer (IPL). We would not consider the DCS based over-ride to be an Independent Protection Layer (IPL) against a catalyst reversal if one of the hazard scenarios in the Layer of Protection Analysis (LOPA) included loss of the DCS controls.