November 2025
Special Focus: Process Controls, Instrumentation and Automation
Choosing the right type of APC
Process plants face increasing demands for efficiency, profitability and operational reliability. Advanced process control (APC) solutions are integral to meeting these demands by reducing variability and optimizing performance. Two widely adopted APC strategies are advanced regulatory control (ARC) and model predictive control (MPC). Selecting the right approach depends on the process characteristics, objectives and economic considerations. This article will discuss the pros and cons of each route, and how to choose between them.
While ARC is built inside a distributed control system (DCS) with no extra tools, licenses or hardware, MPC is licensed software that must be installed in dedicated servers and connected to a DCS. MPC also requires extensive modeling, monitoring and ongoing maintenance. The costs associated with MPC are higher than ARC in both capital and operational expenditures, as it requires specialized skills and/or maintenance contracts. While enthusiastic engineers would like to embark on an MPC journey, many challenges can be handled efficiently via ARC. Therefore, it is worth understanding the specific applications of ARC vs. MPC well before embarking on an APC project. This article will illustrate a simple methodology that can help determine an owner-operator’s course of action, as well as a specific case study example.
ARC. ARC refers to a combination of a proportional-integral-derivative (PID) controller, signal selectors and split ranges, among others, to create a control strategy that is beyond single PID controllers. The following are some examples—the actual control could include one, a combination of most, or all to build a larger scheme:
- Cascade control: A primary controller sets the setpoint for a secondary controller, improving disturbance rejection (e.g., tank level to bottom flow cascade).
- Feed forward control: Proactively adjusts control actions based on measured disturbances (e.g., feedrate to reboiler steam flow cascade in a column).
- Ratio control: Maintains a fixed proportion between two variables (e.g., fuel-to-air ratios in furnaces).
- Override control: Selects the most critical control variable to maintain constraints (e.g., a compressor’s anti-surge valve control opens the recycle valve to prevent a surge, overriding the pressure controller).
- Split-range control: Distributes control effort across multiple actuators (e.g., steam header pressure control—steam letdown vs. venting).
The principle of ARC is to convert real-world process dynamics, as measured by three parameters—dead time, time constant and process gain—into PID tuning parameters, and then use a combination of PID controllers, such as feed forwards and constraint override selectors, to build advanced control strategies within a DCS or programmable logic controller (PLC) that executes at every second interval.
The benefits of ARC include:
- Zero or minimal cost: The DCS resident-type advanced control schemes require no investment in software licensing or hardware, as they are typically designed and implemented in-house. However, if needed, an ARC expert can be hired to initiate design, which can then be implemented by in-house control engineers.
- Execution time: ARC executes every second, leading to faster responses to disturbances and deviations from the setpoint. The execution time can be reduced or increased, but it is recommended to stick to 1 sec as a default.
- Little maintenance: ARC is very robust in terms of maintenance requirements, as there is no model matrix that requires regular correction. Generally, no maintenance is needed until there is a change in instrumentation or process design, either due to a change in field instruments or actuators with ones with different specifications, or a major feed quality change. Usually, they continue to run without incident for years.
- Handles non-linear processes: Implementing ARC strategies, such as gain scheduling, can be built to handle non-linear processes very well and execute per second without any artificial intelligence (AI), machine-learning (ML) or neural network. They can provide superior or equivalent performance to MPC in non-linear environments. For example, polymer reactor control is non-linear, and ARC has been successfully demonstrated for this application in various plants worldwide when coupled with soft sensors.
The limitations of ARC include:
- Applicable for non-multivariable processes only. This accounts for nearly 50%–60% of the process industry (i.e., 50%–60% of processes worldwide can live without MPC for operating closer to constraints). If the process is truly multivariable (as distinguished by a diagonal control matrix), this will require an MPC-type APC since ARC will require too many complex control loops, rendering it impractical.
MPC. MPC uses mathematical models to predict future process behavior and optimize control actions dynamically. It is particularly effective in managing multivariable systems with significant interactions. FIG. 1 shows an example of an MPC control matrix, where an array of control variables (CVs) is on the top row, manipulated variable (MVs) are in the columns, and the dynamic behavior between each MV and CV pair is recorded via step testing. MPC is an effective technology for multivariable linear or partially nonlinear processes and maintains a large set of parameters within predefined targets, including production rate and product qualities. It can also be given an open target to push the process to constraint limits while reducing variability due to interactions.
FIG. 1. An example of a dynamic model matrix of an MPC model.
The benefits of MPC include:
- MPC excels in handling processes where multiple variables influence each other, ensuring stability and optimal performance. MPC handles multiple CVs and MVs simultaneously, accounting for their interactions and correlations. For example, a crude distiller can even move them toward the optimum targets while managing constraints and rejecting disturbances in multiple variables simultaneously.
- Predictive capability: MPC predicts future process behavior based on mathematical models, allowing preemptive control actions to avoid disturbances and maintain stability. Therefore, it can reduce variability by anticipating disturbances and compensating for them proactively.
The limitations of MPC include:
- Cost: MPC requires dedicated investments in software licensing, hardware and specialized expertise. Maintenance often necessitates dedicated APC teams or external contracts.
- Non-linear limitations: When applied in non-linear multivariable processes, MPC leads to limited success or, at times, operator dissatisfaction leading to poor utilization and switching off the MPC.
- Execution time: MPC only executes every few minutes. This time can be reduced to half a minute, although it is not advisable for faster execution, as too many variables with different process dynamics are required to be moved simultaneously, making it difficult to monitor or make sense of the data.
Key considerations for choosing ARC or MPC. The choice between ARC and MPC is driven by process characteristics, control objectives and resource availability. The following considerations can guide decision-making.
To answer the question about which type of advanced control is suitable for a certain plant or process, the answer lies in the interactions between the number of process variables involved. If there are many variables that all affect each other when any one of them is perturbed, then the process is both multivariable and interactive. In such a case, it is better to purchase an MPC to avoid building too many control logics via ARC. For example, contrast a crude distillation column vs. a binary distillation column. In a crude distillation column, changing one variable (e.g., kerosene draw) requires an adjustment to the naphtha and diesel draws, pump-arounds and the residue draw, among others, requiring an MPC. Conversely, changing bottoms or upper draw in binary distillation is usually automatic against cascade control and does not require many variables to be changed—an ARC will suffice.
To simplify the decision, a control matrix can be developed by listing CV in a row and MV in a column; start marking if an interaction exists between the CV and corresponding MV, as shown in FIG. 2. If the control matrix developed in this way is dense, then the process is truly multivariable, and an MPC is necessary. If the control matrix is diagonal, as marked in FIG. 2, then the process is not truly multivariable. A binary distillation or furnace firing controls and a DCS-based advanced control scheme will suffice.
FIG. 2. Control matrix: ARC or MPC.
CASE STUDY
A refiner produced hydrogen (H2) from both a steam methane reformer and naphtha reformers. However, due to a demand-supply imbalance, production was usually kept higher than consumption, leading to header overpressure and continuous flaring. Due to consumption of extra imported feed into the methane reformers, the loss was estimated to be $13 MM/yr.
It was decided to utilize an APC for zero H2 loss to compensate for the supply-demand dynamics, such that the H2 is made available in a minimal amount of time by increasing feed to reformers should there be a sudden increase in demand. As the control matrix was not very dense and the owner-operator did not want to buy a full MPC for just one application, an ARC was chosen. FIG. 3 shows a simple yet effective demand-supply balance control scheme for the H2 header. The idea was to identify overpressure and send a signal to adjust the reformer’s feed rate gradually until zero flaring of the header was achieved, and then ramp up the reformer’s feed rate gradually if more H2 was needed.
FIG. 3. Simplified ARC design for H2 header pressure balance control.
The schematic was implemented in the DCSa and commissioned, as shown in FIG. 4. Once started, this immediately led to zero H2 flaring (FIG. 5), reducing unnecessary production and saving $13 MM/yr.
FIG. 4. Implementation of the ARC scheme in a DCSa.
FIG. 5. DCS trend of flared H2.
NOTE
a The DCS was manufactured by Yokogawa
Comments