Search

KOEBEL (Grace Catalysts Technologies)

This is a question that comes up relatively frequently and is an area where Grace has done extensive R&D work and publication. I will summarize a longer answer that appears in the Answer Book. So please refer to that response, as well as to some of the publications Grace has in the trade magazines. In order to be able to help customers with the thought of running whole tight oil to the riser, we ran a sample of Bakken crude straight to our Davison circulating riser pilot plant over a moderate Z/M catalyst. Our data for the feed sample is shown on the slide and compared to a published assay of Bakken. We felt it was relatively representative of a typical Bakken crude.

To summarize, the data behaved much as you would expect. There was a fair amount of 650°F minus in the feedstock, so it made significant amount of gasoline. You can see that the cracking yield was on the order of 65% gasoline. A lot of that was the gasoline inherent in the crude initially; it had just come along for the ride. There was not much conversion of those molecules that were present in the feed. As a result, the overall FCC gasoline octane is moderately low. The feed actually did convert very well, although it made very low coke and low bottoms yield.

These are the specifics of the FCC gasoline from those products. You can see that the RON is much lower than 80, even significantly lower than you would typically give an FCC naphtha. It did have the normal response to reactor temperature that you would expect with an increasing RON as a function of reactor temperature. But still, the overall magnitude of the octane was very, very low.

On the other hand, the light cycle oil was a much better quality than you would get from a typical FCC light cycle oil. This, again, is partially due to the material that was inherent in the feed. But overall, the quality of the LCO was not really a function of reactor temperature. It was more a function of conversion level and how much of the material in the LCO you were cracking up the tower while leaving the paraffinic molecules in the LCO. The diesel index was on the order of 35 to 45; whereas with a typical FCC light cycle oil, the diesel index was usually in the single digits. So, the quality of the light cycle oil you get from processing shale oil with the riser is much better than what you normally get with FCC light cycle oil.

The question also asked about some of the other considerations for doing this. One of the considerations is that, with the variability of the tight oil quality, from time to time you may get unusual contaminant metals in the FCC feed. This slide shows one example of a unit that was routinely running Bakken crude to the riser. Over the course of a 10-day period, a spike of sodium hit the unit hard. This spike cost them about 10 points in activity on their circulating inventory. Often, the refinery is not set up to desalt the FCC feed directly. So, if you have a raw crude coming to the riser, you may see this happen, either with sodium or other contaminant metals.

BULL (Valero Energy Corporation)

At Valero, we process desalted pre-flash crude in several of our FCC units. We desalt the crude or remove as many of the contaminants as possible. We also may pre-flash the crude to take off the light gas and naphtha, which is shown in the drawing on the slide. We run it through a desalter to try and take out contaminants. It then goes into the pre-flash tower to take off any light gas and some of the light naphtha, and then take the bottoms of that to the FCC unit. We have done this at several refineries. The pre-flash crude is then sent directly to the FCC feed drawing process in the unit.

In general, the naphtha boiling range material goes straight through, and then we can see a proportional increase in the gasoline FCC naphtha range. The naphtha also has a cat cooling effect without the cat cooler, so you must take that into account when processing crude directly to the FCC. Some of the diesel boiling range material passes through, but we do see some of the heavier components of the diesel crack. The remaining portion of the pre-flash crude is your typical FCC feed at that point. So, the net effects are a reduction in overall liquid yield on the unit and the shift in selectivity to more naphtha and diesel range material.

Two primary considerations for investigating new crude sources for processing directly in the FCC are the contaminant levels that Jeff referred to earlier, as well as the effect of the light material and the impact that will have on your delta coke and heat balance.

SOLLY ISMAIL (BASF Corporation)

All of these tight oils are very low in aromatics components. They typically have shorter length hydrocarbons which, when mixed with vacuum resid, can create salting out or compatibility issues. One way to address the compatibility issues is to extend the analysis and include the peripheral unit; i.e., the furfural unit in a lube plant. We know that in a furfural unit the paraffins are being separated from the naphthenes and aromatics. These paraffins are then sent for dewaxing to make base stock for lube oil while the extract, which is rich in aromatics, and some naphthenes become an orphan stream of the refinery. In this case, I am assuming that refinery does not make grease with the extract. Because this furfural extract stream is rich in aromatics, it is difficult to crack and makes lots of coke, which has always been a problem. Therefore, FCC operators are reluctant to accept furfural extract for processing in their units.

Nevertheless, refineries do process lube extracts (in low concentrations) in the FCC and would generally like to process more or higher levels in their feed to the FCC.

As just mentioned, because of the high tendency for furfural extract to make coke in the regen, refiners have been having a tough time blending it away in the FCC feed. However, when a refiner is processing large amounts of tight oils, the situation is completely different. In the scenario when tight oils are being processed, the regenerators have the problem of running at low dense bed temperatures. I think when processing high concentrations of tight oils in the FCC unit, increasing the levels of furfural extract with tight oils might be a blessing. Not only will the furfural extract provide an increase in the level of coke in the FCC regen, but it should help improve the compatibility issues as this stream contains large hydrocarbon molecules that can keep the vacuum resid in the solution.

JEFFREY BULL (Valero Energy Corporation)

At Valero, we have processed desalted pre-flashed crude in several of our FCC units. We desalt the crude to remove as many contaminants as possible, and we pre-flash the crude to take off the light gas and light naphtha. The pre-flashed crude is then sent directly to the FCC feed drum and processed in the unit. In general, the naphtha boiling range material goes straight through the FCC and, in effect, blends with the traditional FCC-derived naphtha. The naphtha also has a “cat cooling” effect without the cat cooler. Some of the diesel boiling range material cracks and some of the naphtha material passes through the FCC reactor. The remaining portion of the pre-flashed crude is typical FCC feed. The net effect is a reduction in overall liquid volume yield and a shift in selectivity to more naphtha and diesel range material. The two primary considerations for looking at new crude sources for processing directly in the FCC are contaminant levels (metals, sodium, calcium, etc.) and the amount of light material that essentially takes a free ride through the FCC and has a significant impact on the heat balance.

JEFF KOEBEL (Grace Catalysts Technologies)

The introduction of novel drilling technologies has resulted in large amounts of oil from shale becoming available in North America. While fluid catalytic cracking is typically done to reduce the molecular weight of the heavy fractions of crude oil (such as vacuum gas oil and atmospheric tower bottoms), in some cases refiners are charging whole shale oil as a fraction of their FCC feed.5 Also, whole crude oil has been charged to FCC units when gas oil feed is not available due to maintenance on other units in the refinery6 and to produce a low-sulfur synthetic crude7 .

As a model case for understanding the cracking of whole crude oil in the FCC and the effect of process conditions on yields, a straight-run shale oil was processed in the Grace DCR™ pilot plant at three riser outlet temperatures: 970°F, 935°F, and 900°F. The whole crude oil was a light sweet Bakken crude with a degrees API gravity of 42. The properties of the crude were similar to those given in a publically published assay8 . Table 1 presents a comparison of the properties of the whole crude used by Grace and the publically available assay data. Additionally, the straight-run Bakken sample was distilled into a 430°F minus gasoline cut and a 430°F to 650°F LCO cut. The properties of these cuts were measured. Gasoline from the straight-run Bakken was highly paraffinic and had low octane numbers [a G-Con® RON software of 61 and MON (motor octane number) of 58]. The LCO fraction had an aniline point of 156°F and an API gravity of 37.6, resulting in a diesel index of 59.9.

The catalyst used in the experiments was a high matrix FCC catalyst, deactivated metalsfree using a CPS (cyclic propylene steaming)-type protocol. The properties of the deactivated catalyst are given in Table 2. For the three different reactor outlet temperatures, plots of the catalyst-to-oil (C/O) ratio, dry gas, gasoline, LCO, bottoms, and coke yields versus conversion are shown in Figure 1. As expected, lowering reactor temperature increases the amount of LCO produced. As seen in the graphs, cracking straight-run shale oil produces little coke and bottoms. At the same conversion level, lowering reactor temperature results in slightly more gasoline yield (due to increased C/O), which is consistent with prior Grace work.

Plots of gasoline olefins, isoparaffins and RON and MON estimated via G-Con software are shown in Figure 2. Cracking straight-run Bakken shale oil produces a low-quality gasoline with research octane less than 80 and motor octane less than 70.

At constant conversion, increasing reactor temperature results in more gasoline olefins and higher research octane number. Diesel quality is of great interest to refiners. Syncrude produced in the DCR™ runs was distilled to recover the 430°F to 650°F LCO fraction. Aniline point and API gravity of the LCO were then measured to allow calculation of the diesel index, a measure of LCO quality [Diesel Index = (aniline point x API Gravity)/100]. Figure 3 presents data for LCO yield and LCO quality as a function of conversion. As seen in the data, increasing conversion lowers LCO quality as a result of increased cracking of the LCO range paraffins to lighter hydrocarbons. Similar to prior Grace work14, LCO quality follows LCO yield and did not appear to be influenced by reactor temperature at constant conversion. Diesel index values of the LCO produced by cracking whole shale oil were significantly higher than values obtained with typical VGO feeds.

As seen in the results from this study, widely varying ratios of products and product quality can be obtained by changing process conditions. Information from pilot studies such as this one helps refiners to determine the optimum processing setup to maximize yields of desired products. The ability of the DCR to produce sufficient liquid product for properties testing assisted greatly in the measurement of LCO quality.

In addition to yields and operating conditions, contaminants and the impact they have on circulating catalyst inventory should be taken into consideration. A catalyst flushing strategy may be required to ensure that contaminants stay at reasonable levels in circulating inventory. For example, one refiner experienced high levels of sodium on e-cat while processing high amounts of whole crude. The sodium more than doubled and catalyst activity dropped more than 10 numbers, both of which impacted unit performance (Figure 3). The unit utilized purchased e-cat to help flush the sodium from circulating inventory.

PAUL FEARNSIDE (Nalco Champion Energy Services)

The largest concern will be with the main fractionator performance. Issues with direct cracking of undesalted crudes revolve around the increased chloride loading and the resultant increase potential for ammonium chloride salting. Care must be taken to insure the upper section delta P does not increase to the point that daily operations are curtailed. Intermittent slumping of the tower while water-washing and the use of salt dispersant chemistries have worked well.

CHRIS CLAESEN (Nalco Champion Energy Services)

The increased metals content in the feed can lead to increased catalyst deactivation, coke formation, and hydrogen generation which can significantly reduce the FCC profitability. While the metals content should be kept as low as possible by pre-treatment in the tank farm and desalters, the effect of Ni and V can be significantly reduced with the use of a metal passivator program that is injected in the FCC feed.

INKIM (PETROTRIN)

There are several KPIs that are tracked in the monitoring of FCC. Liquidy yields is one parameter that is monitored daily to detect any changes in equipment and catalyst performance. The BS&W (base sediment and water) in the main column bottoms is checked daily as well to monitor the catalyst carryover from the reactor into the main column. The heat and material balance should be done at least weekly, if not daily, to detect changes in the catalyst and equipment performance and to detect any meter errors. In the lean gas stream, the hydrogen-to-methane ratio is monitored weekly as it indicates higher metals content in the feed.

Another KPI that we track at least weekly is the e-cat properties to detect any feed contaminant issues. There are other KPIs that we monitor, and these are mentioned in my Answer Book response.

LARSON (KBC Advanced Technologies, Inc.)

In all cases, it is important to establish the baselines. Then what you are really trying to do is track your standard deviation because you need to know the cause and effect versus normal deviation. The cat cracker conversion might vary as much as 1%. So, looking at absolute values, as opposed to relative deviation, is really important. You get some of the KPIs directly; others are calculated from lab data or rigorous modeling. It is important to differentiate so you know the accuracy of the values you are using and how much they will change. We have typically lumped things into operating things or operating instruction KPIs, and then we also look at planning because planning will give you targets to target. Planning is target to target. Boy! That is oxymoronic.

Planning Targets: You need to hit those. Make sure your lab and operating systems are set up to work in conjunction with operational moves. We will typically look at reactor temperature which will give an indication of octane in conversion. We also look at the debutanized octane material monitoring what the RVP (Reid Vapor Pressure) of the material is; because as RVP changes, you will be changing octane. Refiners do not check their feed quality as often as we would expect. Operators are often chasing product quality when, in reality, there was, a burp in the feed quality, and it was missed until it is three days down the road.

Key Constraints: Map your constraints if you are operating up against them. Know your air blower capacity limit, whether it is the horsepower, blower discharge, or wet gas compressor horsepower. Is it a DP (differential pressure)? Many units are now running a low DP on slide valves encroaching nearer the shutdown trip points, so monitor your unit constraints very well. On a longer term or monthly basis, monitor your catalyst activity for metals and look at your fines.

One of the comments is taken this from the last NPRA (National Petrochemical and Refiners Association) Cat Cracking Session. The EPA (Environmental Protection Agency) came into the meeting and talked about the new regulations that will be coming to the FCC flue gas system. Particulate analysis that crosses your cat cracker will be one of the key criteria you will need to map, if you do not have it now. Many places that did not have upsets are not aware of what the particle distribution is on the slurry bottoms. They do not know the particle distribution of the fines caught. They do not know the particle distribution and are asked to do a cyclone analysis. Track your particle distribution on a regular basis, monthly or at least quarterly, so you have a baseline and know when things change.

We would also look at gasoline selectivity, as well as expansion on a unit, to ensure that catalysts are performing appropriately. We look at the LPG to LCO yield to make sure we are getting the correct cracking distribution and proper fractionating. This is an operator guideline. Are we getting good overlaps or gaps? What is the basis of your operation? How well are your operators performing against that basis?

On the utility side, we look at steam production and/or exchange or fouling to determine if we are getting the right value out of the preheat or if the steam generators are on the slurry circuit. These can be tracked daily or weekly. The bottom line is to monitor them frequently enough so that when you see deviation, you will know it is real as opposed to chasing ghosts. There are enough operator and engineer elements that we have to do now between HAZOP (Hazard and Operability) and environmental regulations, so you must have an efficient way to confirm that you are making as much money on your unit as possible.

MUKESH PATEL (Reliance Industries Limited)

Under KPIs and field properties, are you monitoring total nitrogen or basic nitrogen? As I understand it, basic nitrogen is more important than total nitrogen.

LARSON (KBC Advanced Technologies, Inc.)

I recommend total nitrogen for a couple of reasons: First, as soon as you begin to crack a molecule, you will take that which was in a total system. You will actually create more basic nitrogen just due to the nature of the cracking of the molecule. Second, every refinery I have reviewed in the last 10 years has not really been set up to do a basic analysis. They have a total nitrogen analyzer because of their hydroprocessing units. So, tracking the total nitrogen is just as effective as tracking basic nitrogen and watching the delta impact. It is the relative change away from your normal average feed. I have worked in enough different models, besides the one that KBC is selling, to know that you can use total nitrogen just effectively as basic.

MUKESH PATEL (Reliance Industries Limited)

How do you include the basic nitrogen in your simulation?

LARSON (KBC Advanced Technologies, Inc.)

There are some rules of thumb that apply if you want total to basic. If it is virgin oil, we would apply a basic assumption that one-third of the total nitrogen is basic. If it is cracked, then more than one-third of it is basic. So, it depends upon its feedstock, but there are some rules of thumb that can be applied.

MUKESH PATEL (Reliance Industries Limited)

Many times, we see that the basic nitrogen thumb rule does not apply for various reasons.

LARSON (KBC Advanced Technologies, Inc.)

Nitrogen is a contaminant; so, it always applies, just that what it is in relationship with changes.

KEN BRUNO (Albemarle Corporation)

Please consult the Answer Book as we have provided a very thorough list of recommended Best Practices and KPIs that supplement the suggestions by the panel.

J.W. “BILL” WILSON (BP Products North America Inc.)

Many of these things are quite amenable to being tracked statistically, which helps a lot in actually identifying real deviation or, as Mel said, ‘chasing ghosts’ or chasing random deviations. It is very easy. You can do it with a spreadsheet. We have some specific programs to use, but you can do it about as well with an Excel or other spreadsheet.

WARREN LETZSCH (Technip USA)

I want to make a comment on nitrogen. I think total nitrogen is the best way to go. This is particularly true if you are running residual feeds because small nitrogen compounds in VGO really can affect the catalyst quite a bit. The nitrogen in the larger molecules and the 1050°F plus material really does not seem to be nearly as detrimental. It is typical to run 1500 or 2000 ppm of nitrogen with a residual feedstock. And if you run that with a VGO, you would see a serious deactivation of the catalyst. Total nitrogen is the best way to monitor this impact. I certainly agree with your one-third rule; we have always used that. And for regular VGOs, it is remarkably good.

CATHERINE INKIM (PETROTRIN)

LARSON (KBC Advanced Technologies, Inc.)

Key process indicators, or KPIs, can be broken down into two broad areas of operations (daily) and performance-based indicators and can be further divided into fluid solids systems, operational, and yield for more detailed analysis.

In all cases, it is important to establish base lines and track standard deviation. There will be cause-and-effect change versus the normal deviation or operation of any FCC, given the dynamics of the control systems and typical various of feed and severity.

Some KPIs are calculated or used from daily logs or lab data while others are obtained through the use of rigorous modeling of the reactor regenerator system, as well as the hydrocarbon recovery section. Many refiners will have daily KPIs used to monitor the unit against the operating instructions from the planning group. Those KPIs that require rigorous heat and mass balance will be completed by the pacesetter refinery once a week, using a routine and defined time to collect stream and process data.

The direct laboratory and operating data used in concert with kinetic modeling enhances the performance monitoring of the unit. The use of the kinetic model can differentiate the impact between catalyst, feed quality, and severity. Utilizing TBP (true boiling point) yields versus as yield data can be enlightening for both Operations and Engineering using more rigorous troubleshooting assistance.

It is possible to optimize cost (laboratory) to reduce duplicate samples without jeopardizing the value and accuracy of much needed weight balance data. In general, samples and analysis have three values: actionable by Operations (deviation from the setpoint target), key mass and heat balance reconciliation necessary for LP updates (economic tools must be kept current), and historical trending. The latter is critical for both hydrocarbon and water systems. The frequency and accuracy of the mass and heat balance is an indication of the value a refiner places on the economics of operation.

A list of some of the typical KPIs are shown below:

• Typical Operation KPIs

– Operating Conditions (daily)

• Conversion, debutanized RON, LPG yield

• Steam usage

 Dispersion wt% on feed

 Stripping steam rate – Feed Quality (daily)

– Feed Quality (daily)

• Wt% Conradson carbon, 650°F minus, wppm (weight parts per million) nitrogen, wt% sulfur

– Key Constraints (daily)

• Air blower HP (horsepower), wet gas compressor HP, ΔP on slide valves, ΔP on trays that might indicate flooding, turbine efficiency

– Catalyst Properties (monthly)

• Catalyst losses/opacity

– daily

• Metals/activity

• KPIs to Monitor Optimization

– Yields

• Gasoline selectivity

• C3+ volume expansion

• Dry gas yield wt% [feed rate {scfb (standard cubic feet per barrel)}]

• LPG/LCO ratio

– Key Product Qualities

• RVP of gasoline, key separation (overlaps/gaps/light and heavy key component in bottoms and overheads of columns)

– Utilities

• Steam production

• Exchanger fouling

JACK WILCOX (Albemarle Corporation)

In order to monitor equipment reliability, as well as maintain optimum operation, the following KPIs should be tracked because they define the optimum FCCU operation:

On a continuous (daily) basis:

- Operating conditions, including:

a) Riser outlet temperature

b) Combined feed temperature

c) Catalyst circulation rate (catalyst/oil ratio)

- Feed rate and quality, including:

a) Density (API)

b) Boiling range

c) Key contaminant levels such as sulfur, nitrogen, heavy metals, etc.

- Catalyst properties

a) Fresh catalyst addition and withdrawal rates

- Key equipment constraints, including:

a) Main air blower limit

b) Wet gas compressor limit

c) Hydraulic constraints

d) Equipment temperature limitations

e) Key equipment operation, such as cyclone inlet vapor velocities, horsepower recovered (if unit has a PRT)

- Product yields and qualities

a) Product recovery limitations (fractionation, treating)

On a weekly basis:

- Test runs, including:

a) Complete heat and weight balance

b) Feed and product quality properties

c) Circulating catalyst, including both physical and chemical properties

d) Establishment of current limiting constraints

Once per year, and preferably a short time before and after a unit turnaround, a complete equipment and operational evaluation should be performed, including:

- A hydraulic survey, including:

a) A single-pressure gauge pressure survey of the entire reactor/regenerator section from the main column overhead to the flue gas recovery section; based on the single-gauge survey, a complete pressure balance of the reactor/regenerator is developed.

b) A single-gauge pressure survey of the main column and vapor recovery unit

- Thermography survey of the reactor/regenerator vessels

- Utility consumption, including all steam and air supply sources

- Critical equipment performance and limitations are established, including:

a) Cyclone solids and vapor loadings

b) Distributor pressure drops and nozzle exit velocities

c) Major rotating equipment, including the air blower, wet gas compressor, flue gas expander operation

- Establish flowing catalyst fluidization characteristics

GIM (Technip Stone & Webster)

Proper design of cyclones and cyclone support systems will extend the life of cyclones with proper maintenance. But like the tires on your car, it will need to be replaced towards its end-of-life. Just like checking for remaining treads on your tire, one common way to check the remaining life of your cyclone is to measure and log the thickness of your cyclones for each turnaround from their first installations to last turnaround dates. We can use that thickness in various locations on the cyclones and trend them to predict, before the next turnaround, whether a cyclone will need to possibly be replaced or even how much repair may be required on the next shutdown.

In general, cyclone systems are designed for at least 20 years of life. There are two components that determine the life of cyclones: the condition of the metal and the condition of the refractory. Obviously, steady-state operations with limited upsets and excursions will extend cyclone life. Any excursions beyond the original design temperature will obviously decrease the life of the cyclones.

The adoption of the so-called Life Fraction Approach will determine the allowable stress, which considers the normal operating range of temperatures over which cyclones are operated during the course of a year. Excursions beyond the design temperature reduce the allowable stress used for the design of cyclones. The use of this Life Fraction Approach for determining the allowable stress for design of cyclones will allow operators to analyze the life of the cyclone’s metallurgy by comparing its assumed annual temperature profile versus actual operations at temperatures below the designed temperature, as to life of cyclones. Obviously, operation at temperature above design temperature will reduce the life of cyclones.

INKIM (PETROTRIN)

The inspection methods we use to determine the condition or remaining life of cyclones are visual, hammer, dye-penetrant, and ultrasonic testing. The results of the inspection and the thicknesses of the cyclones are compared to the original thicknesses and past values. Based on the findings, repairs or renewals are done. If there is significant erosion or thinning of about 40 to 50%, we will recommend replacement.

In the past, we have determined remaining expected life based on creep calculations, as Steve had said. There was an incident in the past where, due to a change in feedstock, we had to operate at abnormally high temperatures above the design temperatures. We consulted the cyclone vendor to get a revised life expectancy. They considered the past operating temperature profile, the original design, and the abnormally high conditions we wanted to run in the creep calculations. Thus, we were able to determine if we needed to shut down before the plant turnaround or whether we could use the cyclones up to the scheduled turnaround.

KEVIN KUNZ [Shell Global Solutions (US) Inc.]

Shell uses much of what the panelists mentioned; and like them, our methods have evolved for evaluating and checking cyclone remaining life. Years ago, Shell used techniques similar to those described by Catherine Inkim of PETROTRIN – visual inspection and hammering for integrity – but not a lot more. Today, though, Shell does more inspection at midlife using liquid penetration, particularly around high-stressed welds. Shell will selectively collect samples for carburization, sulfidation embrittlement, creep, and corrosion behind the refractory. The discovery of cracks often leads to weldability checks and testing for sigma phase in the regenerator and/or carburization and/or sulfidation, and creep in the reactor side. Shell also checks for cyclone erosion but rarely replaces the cyclones for erosion problems unless the cyclones are at the end-of-life. Strict adherence to velocity constraints and the advent of vortex stabilizer technology has improved cyclone system life. Vortex stabilization has dramatically extended cyclone life and minimized its repair. The norm is 15 years with little refractory repair in the cyclone cone.

Unlined portions of diplegs are checked with a mixture of remotely operated cameras, as well as a combination of manual and automated ultrasonic thickness measurement.

Shell selectively uses FEA (finite element analysis) in areas where metal degradation (creep and/or corrosion) is limiting cyclone system life. This guides local repairs and reinforcements rather than when complete cyclone system replacement may not be warranted. In some cases, we assess potential creep damage using actual historical temperature data pulled from the Process Engineering department, including monthly averages and detailed data around upsets.

We have sample plates with combinations of erosion-resistant refractory and metallic anchor materials placed in a number of regenerators. Data from these sample plates will be used to predict, rather than react to, corrosion behind the refractory.

KRISH KRAHEN (Marathon Petroleum Corporation)

Can you comment on techniques used to assess a decrease in weldability of cyclones as they age, and it becomes harder to do crack repairs?

KEVIN KUNZ [Shell Global Solutions (US) Inc.]

We check primarily for sigma phase. Now, the exact test that our mechanical department uses is not a specific ASTM (American Society for Testing and Materials) method. And if sigma phase is present, there is a limit to the amount of sigma phase that can be seen in the cyclones before you begin to run into trouble with weldability repairs. Shell checks for sigma phase primarily by doing in situ liquid penetrant inspection. The amount of cracking is trended TA turnaround) to TA. Based on the trending of number of cracks and the difficulty of doing local repair welds, we remove samples to represent the mix of plate and weld processes used to make the cyclones. These are polished and etched to reveal the extent of sigma and any remaining unconverted ferrite.

LARSON (KBC Advanced Technologies, Inc.)

This situation is not just in the U.S., although this is primarily a U.S. group here. I have seen locations where they cycled the cat cracker a lot more than we might do in the U.S., sometimes bringing down a unit every two years as opposed to every four years. Those up-and-down heating cycles degrade the run length of the unit even though you might think that cyclone should operate well. Cycles of startup and shutdown have caused a real degradation in the performance of those units. You will begin to actually see the barrels start to change and not be quite so round. As you are going in, you should monitor not just the metallurgy itself, but also the barrel to make sure it maintains its shape; because that will become a problem in your ability to maintain catalyst performance, and you will start to see erosion. We know of a company in Japan right now that is working on its 45th cycle on its cyclone. They cyclones were replaced last over 20 years ago.

STEVE GIM (Technip Stone & Webster)

Proper design of cyclones system, including the support system, will extend the life of the cyclones as will proper maintenance; but like tires on your car, they do wear out and need to be replaced at the end of their life. One refinery maintenance manager used to measure and log the thickness of his reactor and regenerator cyclones on each turnaround, from their first installation through the latest turnaround. He also used to measure the refractory thicknesses at various locations on the cyclones. He then could trend refractory loss from turnaround to turnaround and predict, before the next turnaround, whether the cyclones needed to be replaced or even how much repair would be needed. As his FCC tended to run the same rate and essentially the same feeds all of the time, the cyclone degradation over time could be trended.

I would also like to thank our friends at Emtrol for providing some empirical data from their knowledgebase.

Average Life of Cyclones: The average life of regenerator cyclones is about 20 years, when eliminating those systems that have been destroyed in the first operating campaign due to upset conditions and those operated well below design conditions extending life well beyond the norm. In general, cyclone systems are designed for a 20-year life. There are two components that determine the life of the cyclones: the condition of the metal and the condition of the refractory, both of which begin to deteriorate after the initial startup. Steady-state operation, limiting upsets, excursions, and afterburn extend the life of the cyclones. Utilizing a standard approach for determining allowable stress from the ASME code for a given design temperature does not directly address upset conditions that can greatly reduce the life of a cyclone system. For example, a one-hour upset at 1800°F will reduce the life of the cyclones by many years, depending on the actual design temperature used for design (1400°F, 1450°F or 1500°F).

Life Fraction Approach: Adoption of a “Life Fraction” approach for determining the allowable stress, which considers the normal operating range of temperatures that the cyclones are operated over the course of a year, including excursion temperatures (1500°F, 1600°F and even 1800°F), normally reduces the allowable stress used for the design of the cyclones but provides for operation at upset temperatures for a predetermined period of time. Use of a “life fraction” approach for determining the allowable stress for design of the cyclones will allow the refiner to analyze the life of the cyclone metallurgy by comparing the assumed annual temperature profile to the actual operating temperatures. Operation of the cyclones at temperatures below the design temperature adds life to the cyclone system, while operation at temperatures above the design temperature reduces the life.

CATHERINE INKIM (PETROTRIN)

During turnaround, the cyclones are inspected for cracks, erosion and corrosion, and thinning. The inspection methods used are visual, hammer, dye penetrant, and ultrasonic testing. Welds are examined for cracks. The cyclone supports and body are checked internally and externally for breakages, erosion, gouging, holes and thicknesses, lining and hex mesh for erosion/corrosion, and valves for movement.

The results of the inspection of the thicknesses of the cyclones are compared to the past values, as well as the original thickness of the cyclone material. Depending on the past history or severity of the findings at the current inspection, certain parts of the cyclones (for example, diplegs) may be renewed, as well as repairs done to the Resco lining and hex mesh and weld buildups. Some degree of repairs is always anticipated. However, if there is significant erosion/thinning of the order of 40 to 50%, then recommendations are made for replacement.

In the past, due to changing feedstock quality and mechanical issues with the air grid, we were forced to operate the regenerator at higher temperatures than originally designed. At that time, the cyclone vendor was consulted to provide a revised life expectancy for the cyclones if we continued to operate at the new abnormal conditions. Based on the historical operating temperature and the higher new temperatures, they provided the remaining expected life based on creep calculations and thus determined if a shutdown was warranted before the planned turnaround. Similar analyses can be used to determine if a particular operating mode would be economical, even though it may mean replacement of cyclones sooner than original design life.

INKIM (PETROTRIN)

In the history of our unit, there have been no outright failures of cyclones or cyclone supports, just partial failures. In our experience, the major cause of failure in cyclones and cyclone supports has been erosion leading to thinning, cracks, and breakages. We have observed varying severities of erosion in the reactor and regenerator cyclones throughout the history of the unit. In the very early years, failures due to erosion were more frequent, and cyclone repair and replacement were required within five years or less. However, subsequent to the introduction of Resco AA-22 lining, replacement was required less frequently while some repairs were usually done at every turnaround.

Another failure type we have observed in the past is overstressing due to improperly designed cyclone supports. That was as a result of insufficient hanger rod clearance. We have also observed thermal fatigue associated with the use of quench steam to the cyclone outlets. For cyclone and hanger systems, excessive thermal cycling and temperature excursions above the design temperatures can lead to reduced life and subsequent failure.

Our design standard specifies that the minimum total design life must be at least 100,000 hours, which is about 11.4 years, accounting for short-term temperature excursions. We have been able to operate for lengths longer than 11.4 years based on our inspection. For our reactor, we have managed to operate with the cyclones for 29 years without replacement, but we did have to do minor repairs to them. Those cyclones were changed out as part of an upgrade because of the remaining life of the reactor at that time. The current cyclones in the reactor are now 17 years, and we have done some partial replacements on those. The regenerator cyclones also had longer run-lengths in the more recent past, with the last total replacements coming at 14- and 17- year intervals. The current cyclones in the regenerator are now over 12 years.

GIM (Technip Stone & Webster)

We have seen cyclones that have been torn apart during a campaign due to excessive high vapor velocities and catalyst velocity. We have also observed cyclones which operated for only a few years due to poor cyclone design. Additionally, failures in support systems are usually correlated with degradation of cyclones. Feedstock change from sweet to high sulfur content can affect some of the older carbon steel straps and hangers. Depending on the operation and cyclone design, the time to failure could be short; meaning, one campaign, part of another, or extended over a number of campaigns.

The biggest effect of cyclone life appears to be the amount of time the cyclone operates within the original design parameters. Multiple shutdowns with temperature excursions will obviously shorten the life of cyclones and their supports. Also, the high velocity will wear on the refractor protection and eventually on the cyclone metal itself. Cyclones running within design parameters can expect to operate between 16 and 20 years. Those which often run above the design parameters and have frequent shutdowns will obviously have a shorter number of campaigns.

ROBERT “BOB” LUDOLPH [Shell Global Solutions (US) Inc.]

We found that in regenerator service, cracking in the welds is by far more common for systems that are properly designed for thermal expansion. Older systems have shown creep. The key to design is allowance for upset conditions that you anticipate during the campaigns. In the reactor service, distress, distortion, and even cracking has been observed if coke builds up and prevents free thermal expansion of the system. Key to success is to design for minimum or no sliding parts. In designs where relative movement of adjacent pieces cannot be eliminated, the design accommodates a “breaking force” for the coke. In some cases, materials are used which have shown that coke adhesion is not as strong.

Where galling is a concern, stainless steel with Stellite® surfacing has been used. For carbon steel cyclones that operate with moderate sulfur in feed, carburization and sulfidation take place. In a number of cases, the hanger itself is not necessarily the weak link; but rather, the attachment to the cyclone body. For maximum life, Shell uses a mix of carbon steel and chrome-molybdenum alloys. Carbon steel is economical and easy to repair, but the chrome-molybdenum will extend the service life, especially in high stress applications.

J.W. “BILL” WILSON (BP Products North America Inc.)

You have a few options. From my experience, unfortunately they are not particularly universal. From the standpoint of time to failure, if the failing problem is related to erosion, you can usually develop a tracking mechanism using cumulative velocities over time. We actually use velocity to the cyclone. But as I said, the answer of how far that curve can go seems to be unique to each unit; so, you have to use unit’s maintenance history to figure out your actual limits.

Bob talked about coke buildup. What can also happen is that you can get a coke ratcheting effect where the coke can build up causing you to shut down the unit. Coke and the steel have different expansion coefficients, so there is usually a little space for coke to build up again. So gradually over time, this coke deposition just pulls part the system. That is another reason why repeated shutdowns and startups are not good.

I would like to discourage the idea of there being a defined maximum life to the cyclone. For years, I have been hearing people say it is 20 years. There are currently lots of cyclones older than 20 years that are working fine, so it is really matter of the condition of the cyclone and the metallurgy. For regenerator cyclones, sigma formation that usually shows up is from an inability to actually repair them. You can have thermal stresses that cause distortion to a cyclone. These stresses are usually related to improperly designed hanger systems or other support systems. Other reasons are repeated operation well above where it should operate or repeated thermal cyclings of the cyclone: getting hotter and colder, hotter and colder.

CATHERINE INKIM (PETROTRIN)

In the history of our unit, there has been no outright failure of all the cyclones, or all the cyclone supports, but there have been several partial failures. In our experience, the major cause of failure in cyclones and cyclone supports has been erosion leading to thinning, gouging, deformation, holes, cracks, and breakages. The cyclones are usually inspected at four-year intervals, and the cyclones are assessed by visual, hammer, dye penetrant, and ultrasonic testing.

We have observed varying severity of erosion in reactor and regenerator cyclones throughout the history of the unit. In the very early years, failures due to erosion were more frequent and cyclone repair/replacement was required within five years or less. Subsequent to the introduction of Resco AA-22 lining, replacement was required less frequently, while some repairs are usually done at every turnaround for both regenerator and reactor cyclones. The majority of repairs include repairs to Resco liner and hex mesh, removal and renewal of cracked welds, weld buildup on areas of eroded metal, and partial replacement of dustbowls and/or diplegs as necessary.

Other types of failures observed in the past include:

• Overstressing due to improperly designed cyclone supports: On one occasion, there were several cracks on the gas outlet tube weld attachment to the plenum and the hanger rods and stiffening bars showed signs of distortion and cracking. Reappraisal of the design by the fabricators indicated that there was insufficient hanger rod clearance.

• Thermal fatigue associated with the use of quench steam/water (both intentional/unintentional) to cyclone outlets and to the plenum resulted in several cracks in the gas outlet tube weld and severe distortion and failure of the plenum.

For cyclone and hanger systems, excessive thermal cycling and temperature excursions above design temperatures can also lead to reduced life and subsequent failure. In one instance, we had to replace cyclone supports prematurely owing to the expiration date of the creep life being before the next scheduled turnaround.

For the reactor, we have managed to operate with cyclones for 29 years without replacement and with only minor repairs at each turnaround before the entire reactor was replaced as part of an upgrade, in addition to the vessel reaching its end of life. The current reactor cyclones, which were installed at that upgrade, have been in service for approximately 17 years with partial (dustbowl) replacements done at the last turnaround and full cyclone replacement recommended for the next turnaround.

The regenerator cyclones have also had longer run-lengths in the more recent past, the last two total replacements coming at 14-year and 17-year intervals. The current regenerator cyclones have been installed for 12 years thus far.

Generally, we have found that some degree of weld buildup is anticipated at every turnaround (four-year cycle). Some breakages in cyclone supports and renewals in diplegs dust bowls are anticipated every eight years in the first instance after replacement and then every four years. While our design standard specifies that the minimum total design life must be 100,000 hours with accountability for short-term temperature excursions accounted for, we have managed to get longer run lengths based on inspection of our cyclones.

STEVE GIM (Technip Stone & Webster)

Observations: We have seen cyclones that have been torn apart during a campaign due to operation at very high vapor and catalyst velocity, and we have seen cyclones operating for only two years due to a poor cyclone design. Failures in support systems are usually correlated with the degradation of the cyclone. High internal reactor vapor and catalyst velocity can erode the cyclone supports over time. Feedstock change from sweet to higher sulfur content can affect older carbon steel straps and hangers.

Time to Failure

Depending on the operation and the cyclone design, the “Time to Failure” could be short – one campaign or a portion of it) or extend over a number of campaigns. The biggest effect on cyclone life appears to be the amount of time it operates within its design parameters. Multiple shutdowns or temperature excursions can shorten the life of both the cyclones and their cyclone supports. High and very high velocities wear on the refractory protection, and eventually on the cyclone metal itself, shortens lifespan. Cyclones that run within their design parameters can be expected to operate for 16 to 20 years. Cyclones that run often above their design parameters or have frequent shutdown of temperature excursions probably last for two campaigns and maybe a third with risk of failure during the campaign.

Support System is a Key

The support system is a key component of the cyclone system and needs to be considered when purchasing a cyclone system. A properly designed support system will adequately accommodate the differential expansion between the hot internals and the relatively cool vessel, which is further exasperated in larger diameter vessels. Furthermore, maximizing the load supported directly by the vessel minimizes the loads imposed into the plenum, which further extends the life of the cyclone system.

A poorly designed support system imposes additional stresses into the cyclones and/or plenum, which will not only greatly reduce the live of the cyclone system but also result in significantly greater maintenance costs and turnaround time.

A few tips for proper support systems:

1. With no bracing, the cyclones will dance and be fatigued.

2. Lateral bracings that are not free to expand from the shell.

3. Add pivot points with slots if bracings are mounted to shell.

4. Set in midpoint at the startup to move freely

CHRIS STEVES (Norton Engineering)

Failures due to erosion are very common in FCC cyclone systems. Cyclone age, flue gas velocities, and repair techniques all play into the erosion failure time. A new, properly designed regenerator cyclone will last a five year run without failure. As long as good inspections are done, repairs are identified, and the correct repairs are made, the cyclones can last 20 to 30 years without failure. Pay close attention to the catalyst losses during the run, especially if the cyclone velocities are not within the design guidelines of the OEM (original equipment manufacturer).

Case Study: A 50,000 bpd FCC unit has air rate was increased by adding a smaller air blower. The flue gas rate increased 14%, with a corresponding increase in regenerator cyclone velocity. Within two and half years, the cyclone barrels and transition ducts started to wear away. Catalyst loses increased from 1. to 20 7 tons per day (tpd) within a month. One month later, the losses were over 40 tpd. When the regenerator was first opened, a trickle valve was seen on the regenerator floor, and this was quickly assumed to be the problem. Only after repeated discussions was the decision made to build the scaffolding needed to properly inspect the cyclone barrels and transition ducts. Many holes, ranging in size from pinholes to holes large enough to fit an arm through, were found. These cyclones were 20 years old and had never had this extensive amount of damage in prior runs. Subsequent review of inspection records revealed that the repairs completed during the prior outage may have been less than adequate. A combination of age, prior substandard repairs, and increased cyclone velocities all played a role in the eventual failure mechanism.

Take into account the cyclone age, prior repairs, and current and anticipated cyclone velocity when evaluating the cyclone damage. All three aspects are part of the failure time. Start the paperwork for new cyclones when the refractory hex mesh telltale signs start showing, and pay closer attention to the catalyst losses when the cyclone velocities go above the OEM’s recommended max.