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.

Use adaptive control to improve process operations

04.01.2014  |  Golightly, R.,  AspenTech, Houston, Texas

Refiners and chemical/petrochemical producers can now deploy APC and enjoy sustained benefits by implementing a continuous process for controller maintenance.

Keywords: [advanced process control] [adaptive control] [process safety] [controller maintenance] [economics] [plant]

Sometimes a problem seems to linger forever. For the many years I have worked with advanced process control (APC) practitioners, APC controller maintenance has been a topic near, or at the top of, their problem list. Given that APC practitioners are not generally demure personality types and are extremely passionate about their business, the conversations about APC maintenance can be “dynamic” at times. Fortunately, with the recent developments in APC, the tenor of the conversation has shifted dramatically, and there is great optimism.


When considering APC’s history, there is no doubt that the angst surrounding controller maintenance is understandable. Traditional maintenance approaches have been expensive, disruptive and labor intensive. Arguably, the most prominent and visible issue has been the need to take controllers offline to collect open-loop process data suitable for model identification, which resulted in lost stability and benefits. Similarly, step testing was disruptive and resulted in further costs due to off-spec product and lost throughput.

Maintenance issues

In attempts to address the maintenance problem, many APC vendors have developed tools to assist in revamping APC models and, to one degree or another, they did deliver some benefits. Nonetheless, these solutions never completely addressed the core problems: Maintenance was still done infrequently. Controllers were still taken offline to collect current open-loop process data. The scope of effort was still high due to a lack of precision in identifying problematic aspects of the models. Step testing was still done in an aggressive manner to shorten the duration of the testing period.

These issues have affected not only the core base of users of the technology (like refiners and olefins/bulk chemical producers), but they have also created barriers to APC adoption in adjacent industry segments. The initial and ongoing cost of ownership relegated the technology to very large-scale processes where the benefits were large enough to justify the effort and expense of developing APC solutions. The deep skill set required to be an effective practitioner further hampered growth of the installed base.

It needs to be pointed out that these advances are the result of over 10 years of concentrated effort. The solution required innovation in just about every area of the technology, from the model identification algorithms to the optimization engine. It is driving considerable change in practitioner workflows and methodologies. And as we observe the early adopters of the technology, we are seeing a wholesale change in the economics of APC, with lower initial costs and less erosion of benefits over time.

Wholesale change in economics

The game-changing innovation is called adaptive process control. To appreciate the significance of this achievement for APC technology, Table 1 points out some of the key innovations developed during the 10-year journey.


How it works

Rather than build another tool for sustained value, the latest adaptive process control software develops a controller that is more self-sufficient and requires less maintenance. APC maintenance becomes a continuous built-in process rather than a project. It automates the process of assessing model quality, collecting current data and generating new models as needed.

Adaptive vs. sustained value

In the traditional model of controller maintenance (sustained value), revamping the controller was a lengthy and costly project. Under adaptive process control, however, the controller is modified over time in more of a continuous process (Fig. 1). The model update occurs without the need to take the controller offline and enables a company to reap the benefits of both control and optimization while the model is under maintenance. Model quality analysis (Fig. 2), which continually runs and assesses the accuracy of the model, can detect when degradation of performance occurs. It can pinpoint a specific part of a controller, thereby helping engineers to determine the underlying cause of the degradation in performance.

  Fig. 1.  In adaptive process control, the controller is modified
  over time in a continuous process.

  Fig. 2. Model quality analysis runs continually and
  assesses the accuracy of the model.

The innovations in the model identification algorithms increase efficiency by enabling the use of data with lower signal-to-noise ratios. This improvement, in turn, enables the use of smaller test step sizes, resulting in less disruptive testing. By performing very small perturbation tests, adaptive process control is able to maintain process stability and, in parallel, generate data that is sufficient to create a new and accurate controller model. The control engineer has the ability to define the degree of trade-off between the length of the testing period vs. the degree of optimizing control.

Adaptive modeling creates candidate models for the engineer for review. Importantly, the engineer always has the final decision about which models to deploy online. In addition, adaptive process control runs an automated test agent to continuously monitor the process and alert the engineer in real time of any problems that occur within the workflow. As a result, adaptive process control shifts maintenance from arduous projects to a pragmatic online approach for continuous maintenance.

Now, engineers can do everything required to update control models without the need to turn off the controller. Of equal importance, the software design ensures that the process control engineer remains central to the decision-making process in deciding which new models to deploy.

More than maintenance

One happy consequence of the automation in adaptive process control is that it reduces the cost and effort to apply APC to new controllers, not just to the existing ones. This area is where process manufacturers will see maximum benefit. That is, applying APC to multiple controllers in parallel. With adaptive process control doing more of the work, the engineer can split time between controllers and complete multiple modeling efforts in the same time span. With this approach to controller modeling, users have reported a minimum 25% reduction in effort for single projects and twice the benefit for multiple parallel projects. In addition, companies that are well along the maturity curve with APC are uncovering new rollout opportunities in secondary units and utility systems. Due to these new economics, companies are re-thinking the traditional wisdom of where it makes sense to apply APC.


Consider the following scenario: A typical refining unit with APC generates a minimum annual benefit of $2.5 million (MM) but gives away 35% across a five-year cycle due to maintenance-related issues. That’s a loss of $875,000 per APC application. Multiplied across an average of nine major units under APC control per refinery, this totals almost $8 MM in potential lost benefits over a five-year period that can be traced back to downtime or sub-optimal operating periods of the controller. Of this time, 60% is usually spent in revamp, 30% is underutilization due to the controller not operating at peak performance and the final 10% is attributed to plant turnaround. With adaptive process control, refiners can get online faster to begin accruing benefits sooner after turnaround and reduce sub-optimal operation over the life of the controller.

Adaptive process control introduces a new economic rationale for APC and easily justifies its pursuit. With this latest advance, it’s easier than ever to unlock APC’s tremendous operational benefits. Refiners and chemical/petrochemical producers can now deploy APC and enjoy sustained benefits by implementing a continuous process for controller maintenance. Adaptive process control is here, and it realizes the decades-long goal of more sustainable APC solutions. HP

The author

Robert Golightly is senior manager for manufacturing product marketing at Aspen Technology. He is responsible for solution marketing for advanced process control on a global basis, primarily dealing with bulk chemicals, refining, specialty chemicals and polymers. 

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


Good One.

allan kern

The determination of AspenTech to improve model-based matrix control (MPC) is commendable. As Mr. Golightly points out, the history of MPC performance has been rocky, to the great dismay of industry, who enthusiastically embraced MPC.

Like AspenTech, I have been studying the problems of MPC performance for nearly 15 years, but have come to different conclusions about the root causes of well-known "degraded" MPC performance. I believe it is a matter of process models changing daily, and even in real time, as feedstocks and feed rates change, as they do daily in modern refineries. It also has to do with the wholesale manner in which MPC utiilizes models for predictive control, especially when coupled with general model inaccuracy.

Although I salute efforts in industry to improve MPC performance, the solution proposed in this case could be simply the latest in a series of "root causes" and solutions that have emerged over the years, all of which have eventually gone by the wayside, while the problems have remained.

Based on my analysis, even an ongoing model quality analyzer will not resolve the problem -- what good is a one week cycle, when processes change on a daily and hourly cycle? Moreover, the models need to be ready for the new changes, not ready to learn from them. And finally, as I mentioned, the wholesale application of predictive control is itself inappropriate (I believe) in an operations environment, where very conservative control action to preserve process stability is the order of the day.

To adopt the proposed solution is not to take MPC performance to the next level, but to take support to the next level. Now users will need to be expert in MPC, performance monitoring, and at using the new online tools. This adds yet more costs and complexity to a technology that has already grown too expensive and complicated.

My work leads me to believe that multivariable constraint control and optimization can be accomplished without models -- indeed, many "degraded" MPC applications in industry are basically ignoring their models today, and functioning primarily based on the direction of the gain (positive or negative) only. Model-less multivariable control (getting rid of dependence on models altogether) is the most promising approach to multivariable control performance going forward.

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.