Even considering the myriad industrial computing hardware and software advances over the past decades, the fact remains that closed-loop proportional-integral-derivative (PID) algorithms endure as the dominant regulatory control technology used for all types of processing and manufacturing applications. And for good reason as PID loop control is flexible, dependable and responsive in service for individual loops. But what if we take a more “meta” viewpoint of process control?
For process engineers and other operations staff who are tasked with maintaining safe, efficient production, the identification and isolation of underperforming PID control loops has been a longstanding challenge. Typical production facilities may employ hundreds or thousands of PID controllers, so manual inspection of each loop is typically not feasible.
Complicating these efforts are changing production conditions, requiring multiple tuning profiles or even a transition to manual control under various circumstances. Indeed, many loops demonstrate distinctly different behavior under varying production phases, and they can interact with each other due to the nature of closely coupled production environments, adding additional challenges.
For these and other reasons, end users need an overarching approach for investigating performance across multiple loops, gaining visibility into root-cause issues and taking action to improve system performance across all scenarios. This article discusses how state-based analytics can help users gain new insights and achieve plant-wide optimization.
Traditional control loop performance monitoring (CLPM) solutions automate the plant-wide inspection of PIDs. By analyzing operational data, these traditional CLPM solutions identify undesirable PID performance characteristics, facilitate the isolation of root-causes and even recommend issue-specific corrective actions. Using information from historized controller tags, most solutions are able to identify a PID’s various operating modes such as:
- Remote cascade
Although these solutions have proven effective at providing a generalized assessment of controller performance based on “in use” data, they have lacked a means for evaluating specific situations defined by conditions such as:
- Various operating phases (startup, normal run time, shutdown).
- Multiple operating strategies defined by increasingly complex process control narratives (PCNs).
- Different products/recipes/batches.
- Assorted run conditions (loading levels, processing pressures).
- Other criteria such as advanced process control (APC) supervisory algorithms in use, or even which operational personnel or teams are on duty.
They have been limited to three or more “phases” to account for when a process is in manual, automatic and off. However, a PID must often perform in a wider array of states based on combinations of conditions. Yet CLPM solutions have not been capable of detecting unique states, or distinguishing the performance attributes of more than one condition, and they have largely been limited because of this.
Introducing state-based analytics
To address these shortcomings, a new generation of CLPM solutions now incorporates state-based analytics capable of distinguishing a control loop’s many and unique operating states. Whether by developing a profile of tuning parameters which can be dynamically applied depending on the operating situation or by implementing an alternate control strategy that accommodates newly discovered process dynamics, users are now better informed and more capable of improving production performance.
States are configurable based on any combination of operating phases, products, run-time conditions or other production-related attributes (Figure 2). These attributes are discoverable by the CLPM solution and accessed from a plant’s data historian.
CLPM software equipped with state-based analytics proactively detects negative performance trends and enables users to more precisely understand and address issues such as:
- The need for variable tuning to optimize advanced control schemas.
- Mechanical issues such as valve stiction or valve saturation, which may lead to bottlenecks or equipment failure.
- Excessive variability due to recurring loop interaction.
These performance issues lead to production inefficiency and even costly downtime, and are common to all types of continuous, batch and hybrid manufacturing environments.
State-based analytics in action
Actual applications in the oil & gas and food & beverage industries help to drive home the benefit of state-based analytics.
In Canada’s Athabasca Oil Sands, most oil producers apply steam-assisted gravity drainage (SAGD) to extract oil-rich bitumen from known deposits. At one location, 16 different extraction well pairs are operated at a single, centralized operating pad. Each of the well pairs cycle through a single test separator every 24 hours. The level controller on the separator must handle a variety of process conditions, and it relies on a single set of tuning parameters. Through the application of state-based analytics, the production company was able to identify major disparities in the performance of individual wells. Equipped with the new insights, manufacturing personnel shifted to an alternative control strategy involving gain scheduling. This enabled consistently optimal performance across all the wells, maximizing the pad’s output.
In another case, a multinational food and beverage manufacturer had implemented APC to improve control of its bleaching process. The company’s primary goal was to optimize raw chemical usage across multiple grades of products. Use of CLPM with state-based analytics permitted the company to assess the true impact of APC on the different product grades as they flowed downstream of the bleaching process (Figure 3). The resulting information allowed the manufacturer to determine on which product grades the use of APC provided an improved outcome versus those on which APC was counterproductive. With some product grades APC was eliminated, reverting the process back to a more basic regulatory control schema and improving utilization of raw chemicals.
When considering a state-based loop analysis solution, users should look for a tool that integrates easily with their facility’s existing data historian and that supports appropriate communications. Most CLPM solutions support OPC-HAD, OLE DB, ODBC and other common industrial communications protocols. Support of the ANSI/ISA-95 standard for comprehensive integration of enterprise control systems should also be considered.
Clearly not all CLPM solutions are the same, as they are designed for different users. Some are limited to performing only the most basic computations for “error.” Indeed, they may lack the ability to evaluate either loop interaction or common mechanical issues that contribute to equipment failure and unplanned downtime. They may not utilize data from a facility’s everyday output changes with which optimal tuning parameters can be calculated. Even so, a more basic set of capabilities may be precisely what’s needed at a facility new to process improvement.
With or without state-based analytics, CLPM solutions are powerful tools and with power can come complexity. When looking for a solution fitting the operation’s needs, consideration should be given to the user interface (UI) and to how information is presented as well as to how the solution is maintained. A cluttered UI can limit adoption or regular use by staff who are needed to capitalize on the CLPM findings.
For years CLPM solutions have equipped process manufacturers with valuable information about the performance of their primary source of control — the PID loop. Equipping users with both a plant-wide view of performance and the ability to drill down into individual loops, traditional CLPM solutions get credit for proactively identifying a range of issues that undermine plant efficiency and throughput.
Real-world manufacturing and production facilities are rarely simple as they are usually characterized by numerous conditions, scenarios and interactions. The combinations of these attributes can be enormous, and the effect of individual combinations on production can be lost within broader trends. The addition of state-based analytics now makes it possible for CLPM users to delve deeper, facilitating the detection, analysis and adjustment of operational conditions that had previously stood in the way of plant-wide process optimization.
Brett Beauregard is the director of product development at Control Station. He holds primary responsibility for the design and development of Control Station’s award-winning portfolio of process diagnostic and optimization technologies. Beauregard is a recognized specialist in the fields of PLC and SCADA programming, process modeling and adaptive process control. Prior to joining Control Station, he held positions in process control for 3M Corporation. Beauregard received a BS in Chemical Engineering, with a concentration in Process Control, from the University of Connecticut.
Robert Rice, PhD is the vice president of engineering at Control Station. He is Control Station’s thought leader, and he has published extensively on topics associated with automatic process control, including multi-variable process control and model predictive control. Dr. Rice has been the recipient of numerous awards for innovation, and for his contributions to the advancement of the process industry. He received his BS in Chemical Engineering from Virginia Polytechnic Institute and State University, and his MS and PhD from the University of Connecticut.