PSVs...what's missing? 6

[Zoobie]

I just wanted to start a discussion to solicit some opinions from the experts out there. While clicking through the various forums here I noticed that there are a lot of questions about PSVs (or whatever other acronym you want to use for safety relief valves). In fact, I'm pretty sure I started a couple of them myself. I am by no means an expert and without API 520&521 I would be totally lost.

My question is this: Why do so many engineers struggle with this area of process and equipment design? I'm sure that there are many out there that don't struggle with PSV issues but there are a lot who do. Is there a lack of good training? Should all new process engineers take a course on PSVs? Are we just not reading the reference material? Are we not understanding the reference material? Is the reference material adequate? Is it because of a lot of the 'what if' issues with PSV system design? Or are we just overanalyzing a lot of these systems because they are safety related?

Its great that we can come to Eng-Tips to get advice on PSV related issues but at the same time these should not be exotic problems when it comes to plant and process engineering.

Anways....I was just thinking over lunch...maybe the chicken was bad.

 

[JimCasey]

My big criticism of Engineering curricula is that there is not a dozen hours of coursework in "CODES AND STANDARDS". The only codes I was exposed to were the unified screw thread specs in drafting class. (We used pencils, then) Most PSV applications are heavily governed by either ASME Sections 1 or VIII. But then engineering curriculae don't address equipment application as much as they address theoretical subjects. How do you size a pump? I dunno but look at this cool fourier transform! Since I graduated I have used calculus >ONCE< and that was to design a model airplane wing.

 

[UmeshMathur]

Zoobie:
The real complication is that you are dealing with compressible flow phenomena for (a) vapors or gases, or (b) vapor/liquid mixtures. The velocity at the throat can easily be sonic if the pressure ratio is high enough. There are many textbooks that discuss choked compressible flow for ideal gases (PV=RT) in this situation for idealized venturies, etc. For real gases, there are few books that describe the gory details. Further, in a valve, the flow is reduced relative to a smooth venturi for the same throat area because the geometry is not simple, so you need "discharge coefficients" which are really odd-shaped correction factors.

Besides, you often have to contend with highly non-ideal gases (compressibility factor far from unity). Here, the thermodynamic calculations require solving equations of state (EOS) under adiabatic conditions, a subject area normally in the domain of chemical engineers. Few undergraduates ever go through the computational exercise of solving any but the most trivial EOS models.

For multiphase flow, the hydraulic phenomena are vastly more complex and now you are definitely in the domain of experts with a very high degree of theoretical and computational knowledge. Recent advances have required application of computational fluid dynamics, a highly sophisticated discipline usually requiring graduate level training.

The various API and other standards are based on recommendations from recognized experts, but understanding the full basis for each standard requires a great deal of work - normally far beyond the time available in a typical job environment.

Finally, there is the question of what the actual back-pressure is just downstream of the valve. Since these are normally part of an elaborate flare system network, the back-pressure calculations themselves are highly involved and riddled with trial and error. These are invariably handled using quite sophisticated computer software that predicts multiphase flow regimes reliably for horizontal, vertical, or inclined flow. Recall that the flare system may easily require handling multiphase fluid flow where the gas is highly compressed and subject to severe acceleration (kinetic energy) losses as you go down the pipes. The same thermodynamic issues, requiring solving EOS models, must be addressed here also. This aspect is an order of magnitude more complex than single phase flow, which in itself is bad enough.

So, it is no surprise at least to me that this whole area does not lend itself to facile answers.

 

[Zoobie]

UmeshMathur,

You make some great points. I agree that the nuts and bolts of multiphase flow are not at all trivial. I contend however that PSVs (and the sizing and installation thereof)have been around well before every engineer had a computer on his desk. At that time I'm sure the hand calculations performed were at the undergraduate level (but perhaps not part of the curriculum).

Perhaps though, supporting your points, we are obligated to use the best tools and knowledge that we have at our disposal at the time. However, there are times when engineers can spend vast amounts of time and money to analyze a problem when a 5 minute calculation using rules of thumb can come up with virtually the same answer...been there, done that.

In my experience PSVs have usually been the domain of the plant process engineer (for plants already operating). Obviously he/she typically will not have the expertise and definitely not the time for an in depth analysis. I would be interested in knowing if this is left to 'experts' in most cases. Perhaps it should be.

As an aside, I would be very curious to know what percentage of PSVs are sized and installed correctly...if this has ever been studied I'm sure the results would be an interesting read.

 

[UmeshMathur]

Zoobie:

One cannot argue with the necessity of getting answers in a timely manner. However, my response was directed at the some of the deeper questions you had raised in your original post, to wit:

"My question is this: Why do so many engineers struggle with this area of process and equipment design? I'm sure that there are many out there that don't struggle with PSV issues but there are a lot who do. Is there a lack of good training? Should all new process engineers take a course on PSVs? Are we just not reading the reference material? Are we not understanding the reference material? Is the reference material adequate? Is it because of a lot of the 'what if' issues with PSV system design? Or are we just overanalyzing a lot of these systems because they are safety related?"

I believe that it would be false to assert generally that these systems are over-analyzed, by which I presume you mean that the depth of analysis is excessive and adds no value beyond what a quick and dirty calculation would.

In the past, i.e. before computerized packages became readily available for safety system design and rating, it was common practice to use rules of thumb that invariably led to over-design. Today's environment is much more cost-competitive and hence the need for more elaborate analysis as a way to reduce costs. Also, the regulatory pressures are much more intense and such detailed analyses are becoming far more common in response.

It is hard to answer your last question without conducting a massive survey across many industries. For older plants that have gradually been modified and debottlenecked, it is not unusual at all to find that the original flare systems and PSVs have become inadequate for worst-case scenarios. I have personally calculated relief systems by hand over 35 years ago in a major petroleum refinery and found, to my horror and amazement, that they had become dreadfully undersized within 20 years after the plant was built, thanks to persistent debottlenecking.

Even today, I believe that older plants would probably be well advised to commission an extensive survey and top-to-bottom analysis to ensure they're safe. Also, every plant that undergoes a significant expansion or revamp should mandatorily revisit this area. Even today, many large explosions of pressure and storage vessels are caused by poorly designed relief systems (especially when fire engulfs the surrounding areas).

In my opinion, this discipline has received significantly less attention than it deserves because many decision makers at the local level see only the immediate costs of doing the work, not the longer-term economic benefits in improved plant and worker safety. This unfortunate tendency can be cured only by the aggressive adoption of better policies based on greater diligence on the part of owners.

Again in my opinion, this is partly due to the fact that insurance companies do not recognize the value of such work as a basis to offer reduced premiums to those who are more diligent. If that were the case, much like defensive driving courses to lower automobile insurance rates, I'd bet we'd see some real progress.

 

[Zoobie]

UmeshMathur:

Thanks, once again for your perspective on my queries. You made some very good points.

 

[Samiran]

To my mind, there is a certain amount of subjectivity involved regarding a PRV design. I am not talking of the design formulae but especially the philosophy & the parameters that are used in the calculation. API 520/1 are basically guidelines and many companies choose to deviate conveniently especially on ones that are already installed. You can appreciate what I mean when you start to review the design for two-phase relief. A massive amount of theory has been poured into this subject of divine happiness. Dynamic simulation has drastically changed PRV sizing. But then again question arises - do we go dynamic or static or even nothing. May be do a QRA and ascertain risks and see if we can take a deviation of the code.
PRV design is Engineer's passion- and like any passion it raises a storm - everywhere (not to mention of this Forum).

 

[UmeshMathur]

Samiran is right on the mark: if you perform a dynamic simulation, it is quite remarkable to see how transient phenomena (such as step changes in feed rates, fired heater or reboiler duties, etc.) can cause sharp dynamic transients. Thus, the vessel pressure and / or the relief flow may increase very substantially over the nominal steady-state value before settling back down.

In the past, it was unheard of to perform a dynamic simulation for any but the most rudimentary processes. However, even today, the effort required to do such a simulation is almost an order of magnitude greater than a steady state analysis.

Also, it is only wihin the last 2-3 years that the commercially available general purpose dynamic simulators have reached the proper technical proficiency, in my opinion. Some would say that their cost is still exorbitant relative to most other chemical engineering software.

In the past, the speed of most of the computing platforms (engineering workstations) was inadequate for the sheer volume of number crunching required for problems of realistic size. Finally, the training required for efficient use of dynamic simulators is costly and prolonged.

In summary, despite everything, I think this technology will make a dramatic impact on how we engineer and analyze safety systems in the near future. Note also that the commercial simulators available today provide convenient access to many of the computational requirements referred to in my earlier post of 10 Aug 05 1:33.

As Samiran implies, the number and variety of cases that you might need to run for achieving "technical closure" can be quite high. This might easily be enough to send you off to a philosophical metamorphosis when confronted by tight budgets and time limits - the relationship between missed deadlines and missed promotions or salary increases is known to all.

Therefore, progress in this area will likely be spurred by enlightened plant management, if not insurance incentives.

 

[TrevorP]

Another thing from a purely practical point of view.

Size a pump wrong, and your error is often immediately obvious as soon as you start up - or it is close enough to either modify or to live with it as it is. Usually, there is not a high cost involved, and the only thing damaged is your pride. It's also usually fairly obvious whether you need a pump or not in the first place.

Get a PSV wrong though, and it's a whole different ball game. You won't know it's wrong until it's too late (unless you or someone else manages to pick up on your errors or calculation inadequacies in time). There is no opportunity to test it under real conditions, and the consequences can be extreme. It can also be unclear whether one is needed at all - though I've never had a manager brave enough to attempt to remove one from the scope of a project! Your embarrassment from inadequate sizing could potentially even extend to defending your actions in court, rather than just to your boss.

In practice, most PSV inadequacies I have picked up (of which there have been a few, but not a huge number), are not to do with calculation inadequacies or errors, but rather with a failure to identify the root causes leading to the requirement for pressure relief.

 

[Zoobie]

TrevorP,

Very good points. I started this thread and just in the last week I have been asked to provide some fluid properties (I'm the only one with HYSYS) to someone in our organization who will feed it to an engineering company for PSV sizing. Even though I have nothing to do with the sizing I am really concerned about handing over fluid properties because I am not convinced that the conditions for relieving have been identified.

The vessel is a direct fired amine reboiler. I asked what the maximum expected temperature was to be. The reply was, "Can't we just use the normal operating temperature?" I think I need to wash my hands of this.

 

[UmeshMathur]

We should all thank TrevorP for stating what we have all assumed implicitly: errors in this business can go undetected long after a design or rating calculation was done. We live in litigious times, and such errors may not go unpunished in case of severe loss of life or property many years down the line.

Establishing the worst-case scenario for relief calculations, in itself, requires much experience and good technical judgment. For example, a pool fire around a major vessel may easily generate a huge incremental contribution to the maximum vapor flow from all heat sources combined. Failure to use the proper methods to estimate such additional loads can often be a significant inadequacy. Another sometimes neglected area is the stress analysis of major vessels, piping, and support structures when weakened by high temperatures resulting from pool fires.

That is not all, unfortunately. The computation of fluid physical and transport properties under relief conditions is non-trivial and often requires access to a comprehensive process simulation tool, such as HYSYS, as TrevorP has properly noted. Additionally, as noted in this thread before, relief computations through valves and the flare system are a lot more complex in multiphase flashing flow, relative to single-phase flow. Here, I would say that the same tools as were used for the fluid properties should also be used for the fluid flow analysis. Obviously, the entire network needs to be considered at least at steady-state if not dynamically.

Also, there is a major decision to be made concerning how many of the plant's major process units should be assumed to be engulfed in a fire emergency. This has a dramatic effect on the peak dynamic hydraulic loads in the flare network. In fact, it is this aspect that lies at the heart of good facility layout and design of internal fire-fighting facilities, including automatic cooling sprays.

It is important to avoid cluttered multi-story layouts to improve fire-fighting access and reduce risk of rupturing neighboring pipes and vessels. Unfortunately, in an intense desire to minimize pipe lengths and structural costs, some engineering companies have begun to shrink the plot plan drastically. This is one result of owners favoring an increasing trend toward “fixed fee” bids, WITHOUT prior issuance of an adequate set of engineering specifications and design standards that require consideration of safety issues in a comprehensive way. Having personally been in some very cluttered plants, I can attest to noticing many bad situations where a major leak and fire would likely destroy the entire facility. When I inquired about the relief system computations, I was quite disheartened at the quality and level of analysis that was provided.

Over many years, there have been a number of very costly accidents in the process industries that have subsequently been analyzed in depth. The lessons from these analyses are not taught in the vast majority of chemical engineering design curricula, and professional engineers are somehow expected to learn this “on the job”. When cost considerations trump process safety criteria, it isn’t hard to see why inadequate design practices become entrenched and intrinsically unsafe designs get approved routinely in some quarters.

Not wishing to sound alarmist, I ask again: where are the senior management of the owner companies and the insurance industry in all this?

 

[sshep]

Has anyone ever considered starting a relief system forum in the ChemEng area? This is certainly one of the more popular thread subjects and would recieve far more action than some of the smaller forums like chemical process development, etc.

From my own opinion about why we spend so much effort struggling with PSVs, I think we are tending to add that layer later rather than incorporating the relief considerations into the design up front. Many relief valves could be eliminated by removing isolation valves from the design or other simplification. For example, once we put in isolation and bypasses around a tower feed pre-heat exchanger we have practically insured a relief valve is required at that point. Then if it is liquid or set at high pressure, the whole thing can get ugly, and extend into a region of application not well covered by API520/1. On the otherhand if there were no isolation from the column, the tower relief valve could easily handle this incremental load with minimal additional effort or analysis. These are tough design trade-offs, but designing for simplification of the relief requirements is always something good to at least consider.

best wishes to all,
sshep

 

[Zoobie]

A relief system forum....a great idea. Hmm, I wonder how to go about that.

 

[UmeshMathur]

sshep makes an excellent point, since good design avoids many problems at source. We need to incorporate a thorough understanding of the fundamentals, including familiarity with the codes, into our design procedures.

To sshep's specific point, bypasses around exchangers or other equipment might be provided if it is desired to clean them on-line or take them out of service for maintenance, or if there is a spare which can be placed on line while the other is being worked on, etc. All this tends to be highly application-specific and hard to discuss in general terms. Of course, cloning of bad design practices is the worst culprit of all.

 

[Zoobie]

sshep makes an excellent point about simplicity...but it also raises concerns related to the way PSV sizing is often approached. I have recently experienced a situation where the target was simplicity for correcting a relief valve installation problem....the target appears to have been missed because the relieving case was not clearly understood. One simple hot tap could take care of the problem...except now we have people thinking about two phase flow and other stuff because its 'simple' to just put a new valve where the old one was.

 

[Zoobie]

I guess all you have to do is ask...there is now a relief systems forum in the chemical engineering area.

thanks sshep

 

[sshep]

Zoobie, All thanks goes to you for organizing that forum. Good job! -sshep