Hydrocarbon Processing Copying and distributing are prohibited without permission of the publisher
Email a friend
  • Please enter a maximum of 5 recipients. Use ; to separate more than one email address.

Lose the pyramid (and find process control success)

07.01.2011  |  Kern, A.G.,  Tesoro Corp., Los Angeles, California


Most people would classify multi-variable control—or advanced process control (APC)—over the past 20 years as a “slam dunk” in terms of process control, automation and optimization success. But it can also be viewed, quite plausibly, as a major diversion that derailed the modern distributed control system (DCS) from achieving its natural place at the center of process control and automation progress. In the process, vast amounts of time, money and talent have been diverted from much more essential process control needs, in favor of APC’s much over-estimated benefits.

As industry comes to terms with the idea that APC may be just another tool in the kit, rather than a game-changing technology, it will help the process control community to find renewed traction and move forward if we adopt a fresh paradigm based on the lessons we have learned, because a world of control and automation opportunities remain, but they are not going to come from APC.

Losing the pyramid.

The traditional pyramid paradigm, with APC prominently at its center, served well to initially focus industry on this promising technology. But the situation we find ourselves in today is an over-emphasis on APC on the one hand, even as it continues to exhibit intractable limitations on the other hand, such as:

• Traditional control-loop design is the product of experience, not the lack of APC. Some “handles” are suitable for closed-loop control, but the majority are not, given the realities of industrial process and equipment operation. Most APCs are built on the opposite assumption that including all handles and interactions is optimal, and as a result, quickly devolve online, in hours or days, not months or years, to utilizing basically the same limited number of handles that were in use before APC ever came along. This is the well known “clamped MV” problem.

• The APC algorithm is a good algorithm, but it is only one algorithm. DCSs have hundreds of algorithm function blocks that are readily combined in myriad ways to deliver smartly tailored solutions to almost any control or automation challenge. And again, the DCS function set is no accident—it derives from a long tradition of process control success.

• APC is the enemy of agility. It is in vogue now, as it should be, to treat process plants as manufacturing plants, where agility is better understood to be essential. But the business and engineering model of APC, with its five-year life cycle, interim inflexibility and high cost of changes, is entirely at odds with the concept of agility.

Finding success.

A fresh paradigm will help the process control community get past APC and get on with the business of process control and automation. Fig. 1 shows one possibility. It’s a “working paradigm” in that it serves not only as a point of departure for process control planning, but also to measure and guide progress at both management and engineering levels. It shows the potential contribution of multiple essential process control competencies and the portion that has actually been captured.

The first bar, something we are all familiar with (a modern DCS), provides a reference value for the others. For example, smart-field devices have the potential to bring as much additional improvement as the original DCS (another 100%). The percentages in Fig. 1 are overall industry estimates, but each corporation and plant site can put in their own numbers and appropriate competencies.

  Fig. 1. Proposed new process control
  paradigm emphasizes multiple essential
  competencies and serves as a working
  guide at both management and
  engineering levels.

Notably, huge opportunity remains untapped in the smart field, smart DCS control, and smart safety system areas. Far from being done, process control and automation are just getting started! HP

The author

Allan Kern has 30 years of process control experience and has authored numerous papers on advanced process control, decision support systems, inferentials, and distillation control, with emphasis on operation and practical process control effectiveness. Mr. Kern is a professional engineer, a senior member of ISA, and a graduate of the University of Wyoming.

Have your say
  • All comments are subject to editorial review.
    All fields are compulsory.


As a senior Process control engineer having more than 35 years past in Control engineering ( Instrumentation / Control / Automation and APC projects and applications ( Shell )!implementing APC / MBPC / ( Smoc 2 Multivariable Algorithm ) I can say positively that MBPC Multivarible Is a real success ( Performance is proved ..with high ROI . All major Petrol Cies have implemented MBPC and all studies have the same conclusions about the real performance of those Algoritms ( some one could be better .than one ..SMOC2 using state variable representation,with state observers,estimators ,kalman filters ,using intermediates state varibles , grey box models ,disturbance models ...etc. Constraints ( hard and soft )!taken in the optimisation criterion all
Are at the point of APC technology .
I told to Allan Kern that i was not agree with him and some friends of mine as Dr Pascal Marquis ( father of Smoc ) were totally against this interpretation
ROI is a good indicator to prove the performance Dr Jacques Richalet a friend of mine ( internationaĺ reknown in Predictive control
Father of iDECOM ,HIECON .,PCR ,PFC has developped hundred of MBPC applications in the world with a great Success !!
That is the proof that MPBC is the best philosophy for multivariable control on complex applications.
I will said again that the success depends of feasabilities sutudies, of the strategy applied ( taking all the constraints...hard ans soft ) and also of the robustess of the algorithm ..and too of the experience and competencies of Process Control engineers ( multiplicinary competence ..
I could wrote a paper on all the pitffalls to avoid for a successful projects in APC
Michel Chandevau Senior Member ISA

R Sens

Mr Kern assertions that APC (MPC) has not been a success and will not be in the near future is contradicted by that fact that almost all of the major oil companies are hiring engineers to do APC work at the current time. It is common knowledge that Exxon and Shell have maintained staff to do APC work for many years. Chevron and Valero have more recently hired to build up APC groups. COP, BP and Marathon are also very engaged with APC work. Suncor and Petrocan have enjoyed success with APC for the past 2 decades. These companies do not engage in APC work without confirming benefits from past and current projects. It is not reasonable to suggest that hundreds of managers and engineers at all of these companies have been deceived.

It should also be noted that most of the articles and conference presentations on successful MPC applications are not vendor sponsored. APC software and service providers long ago realized that papers written solely by those in industry carry more weight with readers.

I encourage Mr Kern to write about what he knows about, which is clearly not APC.

Kalpesh Patel

Mr. Kern says that he has seen a very few APC applications that add value beyond what can be done on a modern DCS. But that does NOT mean that there are only a very few APC applications in the world that add value beyond what can be done on a modern DCS. All it means is that within Mr. Kern experience and exposure with APC applications, he came across only a very few good APC applications.

It is his personal opinion which I do understand and value, but extending ones personal opinion and making public absolute negative statements on APC technology as a whole is not fair. Who knows, part of the reason for the not so good APC applications could be improper design and implementation of the applications or insufficient understanding of the process itself.

Andrew McIntosh

I agree with the statement that "APC is a good tool to solve the right problem, but not an universal one". There are many factors which one must consider in deciding whether it makes sense to apply APC to a process unit. Most of these are technical but some are related to what the operators have grown accustomed to using or can be trained to use.

My issue is with the "absolute" statements that Mr Kern likes to use:
"they are not going to come from APC"
"it continues to exhibit intractable limitations"
"APC is the enemy of agility"

I recently visited a site that I had not been to before. They have over 10 small, medium and large APC controllers running, most of which are performing quite well even though they have not been revamped in many years (and several unit turnarounds). The performance of a few could be improved. It will not take that long to retest and update the models for these. There will be no need to spend "vast amounts of time and money".

Roberto E. Aiassa

I have been a Process Control Application Engineer for the last 17 years (more than 30 years in refining), and I have to agree with the author.

To build an APC controller is easier than to build a traditional control application using DCS' blocks (apparently you don't need any process experience at all...): you put MVs and CVs together in a bag, shake it, and voilà, like magic you have your controller. But after some hours, many MVs will drift to values they will keep forever, and after some days Console Operators will end up clamping MVs they don't want to move (and they are wright most of the time) and turning off CVs they consider shouldn't be honored. So, in the long term you will end up with a control that is no more efficient than a traditional one (in fact an APC can give you some nasty surprises, like when it try to control a CV moving a MV with low gain because the other ones are clamped or out of service), but with an additional maintenance burden (APC maintenance is definitely more complicated). And that is true for most process units that just have first order plus dead time models (like crude units).

Where APC excels, is in highly interactive/non linear units (like fluid cat crackers that have funny models), or when you are trying to control a CV that have models with many MVs with big dead times (like controlling a distillation tower's stream quality that have models with MVs of three/four other distillation towers

So, in brief, APC is not a universal tool: you should use your knowledge of the process to decide what is the best solution for your control problem, otherwise if you try to use APC anywhere, you will end up being like the guy with a hammer, that look to all the problems like nails...

Roberto E. Aiassa

I have been an Advanced Control Application Engineer for the last 17 years (more than 30 years in refining) and I have to adhere to the author's position: APC is a good tool to solve the right problem, but not an universal one.

It is so easy to build an APC (easier than a traditional DCS control, "apparently" you don't need process experience): you put together in a bag MVs and CVs, shake it, and voilà, you have your APC. But, after some time some MVs will drift to a value they will keep forever, and Console Operators will end up clamping MVs they don´t want to move and turning off CVs they don't want to be honored. So, at the end, you will end up with a control that brings you benefits not to much different than the ones you can achieve with traditional DCS control modules, but with an extra maintenance burden to keep the performance. And this is what happen with APCs implemented to control mostly linear process units where there is a simple gain plus dead time (first order) in CVs/MVs models (Eg.: crude units).

Now, just to be fair, APCs excels in no linear process units (Eg.:fluid cat crackers), where CVs' responses are really funny, or when you have to control variables that are affected for many variables with big dead times (Eg.: flooding/quality control of distillation tower stream that is affected by MVs of another three/four distillation towers located upstream).

So, lets be smart enough to use the right tool for the right problem, and not being like the guy with a hammer that looked to all problems like nails...

Andrew McIntosh

It is certainly worthwhile to have a debate on limitations of any control technology, especially those that are in widespread use. However, I know many control engineers who ignore Mr Kern's opinion because he seems to write a steady stream of articles that dismiss APC altogether.

After 30 years in the business, I am fairly confident that Mr Kern has many good ideas which have been used to improve the performance of APC applications. He might consider writing on this topic. It would also be very worthwhile (in my opinion) to write on what should be done to DCS systems to make them more capable. Or write on how smart field devices can be improved or used to more benefit.

It it true that more authors should include sections on how common limitations experienced by some APC applications were avoided. Let's consider MV clamping. I've worked at several sites where this is not an issue. The controllers are designed to be robust, operators are trained to understand that MVs should be clamped only if absolutely necessary, they communicate this information to the next crew (so clamps can be removed when the "condition" has passed) and to the process or control engineer (who investigates, if there is a possibility that it is an issue with the controller itself).

I know of another site where operators cannot adjust the MV limits. If they feel the APC application is not working properly, they turn it off. The service factor of the application is then a pretty good indication of whether it is meeting the objectives.

APC has evolved. There has been recognition by most control engineers that regulatory controls are important to the success of applications and designs need to take into account "where" objectives are best met. There has been a trend to not including every possible CV in an application. But keep in mind the power of a model predictive controller with an LP/QP is the ability to control and optimize a process when the constraint set changes. We do not want to simply "pair up" CVs that are often against limits. We want the controller to respect all important constraints while it is capturing economic benefits.

Yes, it is not trivial to build APC schemes and keep them running. But it is also not rocket science. The benefits are often large so payback is measured in months or even weeks. Control engineers get excited and write articles when this happens!

Allen Kern

I appreciate Mr. McIntosh’s feedback, but I can only concede to one of his points, which is that my view is in the minority. I have always been clear that my view of APC runs counter to the conventional wisdom. Why else write? Aligning conventional wisdom with actual experience will greatly benefit industry, since so much emphasis has been placed on APC in past years. That is my driver.

Aware of my minority position, I actually make an extra effort to be fair and defensible, for example, by painstakingly explaining how “MV clamping”, one of APC’s more obvious and “intractable” limitations, comes about. I'm sure I lose the attention of many readers with this, but such detail is necessary to convey, in this case, why MV clamping is a problem, why it seems unavoidable, and why it shouldn't be blithely ignored, as the APC business has done for 20 years.

I often cite references (my July editorial was not my first one), but mainly I rely on first-hand experience, as it should be, after 30 years in the business, including 15 doing APC deployment and support, 10 of which I was APC leader at one of the world’s largest refineries. During this time, I worked diligently, employing every “best practice” in the book (and inventing a few of my own along the way), in an effort to break through the APC “low performance ceiling”. In this effort, I came to understand the limitations of APC on industrial processes, regardless of how well it works in simulation or how subtle the limitations may be to those not intimately familiar with the operations environment. Although there is surely much I don’t know, compared to most people, I am more studied on this topic, not less.

In addition to employment with two major refiners, I have been actively engaged on this subject for over 10 years, which has brought many opportunities to see APCs and talk with engineers and managers at many sites. I am confident my experience is the rule, not the exception. The only question is, how big is the exception? Logically, there are many successful APCs, but honestly, I have seen very few that add value beyond what could be done by a “reasonably experienced control engineer” on any modern DCS. (However big a matrix, a very limited number of variables remain active in practice, and could be more easily handled by conventional DCS algorithms, without the cost and overhead of APC.)

It should be noted, regarding objectivity of authors, that very few are independent. I do not work for an APC company, nor for an APC competitor. I am an end user, focused on what works in the refinery, and passionate about the success of closed loop control and automation. Those are also drivers. To me, it is the vast pool of vendor-sponsored literature that is “biased” and of “tabloid” caliber.

The efficacy of APC aside, it is difficult to fathom why anyone would take issue with the idea that smart field devices, safety systems, and DCS controls remain a critical challenge in

Andrew McIntosh, AspenTech

Mr. Kern’s opinions do not come even close to reflecting the views of the majority of control engineers working in the hydrocarbon processing industry.

He certainly is entitled to his belief that DCS systems and controls can be improved, but there is no need to make wild assertions about advanced process control (APC) at the same time. There are a large number of successful APC applications which are delivering value year after year on just about every hydrocarbon processing unit in existence operated by many major oil and chemical companies.

Mr. Kern is one of only a very few authors who denies the success of APC. His opinion that “it continues to exhibit intractable limitations” is not supported by a reasonable number of papers written by reputable authors. Nor is the statement that “a world of opportunity remains, but it’s not going to come from APC”.
His article provides no evidence to support the suggestion that smart field instruments can deliver the same benefit as APC, ARC and “best practice” regulatory controls combined. His claim that “a huge opportunity remains untapped in the smart field and smart DCS control areas” is unsupported by solid references.

On the contrary, there is much evidence in literature supporting the fact that the majority of the benefit from process control stems not from tighter control of individual base level regulatory loops but rather from minimizing the variability of constraint variables (including product quality inferentials) which are affected by multiple loops. Further benefits can be captured by pushing the unit closer to not one but the set of constraints representing the most profitable operating point, which is what model predictive controllers do day in and day out, when they are built by reasonably experienced control engineers.

Mr. Kern is also clearly not familiar with modern tools and methods that allow one to revamp model predictive control applications quickly, when he makes the assertion that “APC is the enemy of agility.” I have over 25 years of experience implementing process controls at both the regulatory (DCS) and advanced process control levels. Enough to say that building and updating model predictive controllers can be (usually is) faster than configuring the large number of function blocks required to replicate similar functionality directly in the DCS.

Further, it is unreasonable to compare comprehensive MPC software solutions with simple DCS function blocks, many of which must be interconnected to build an advanced control scheme of any kind.

Related articles


Sign-up for the Free Daily HP Enewsletter!

Boxscore Database

A searchable database of project activity in the global hydrocarbon processing industry


Is 2016 the peak for US gasoline demand?




View previous results

Popular Searches

Please read our Term and Conditions and Privacy Policy before using the site. All material subject to strictly enforced copyright laws.
© 2016 Hydrocarbon Processing. © 2016 Gulf Publishing Company.