APC

What is APC?

APC (Advanced Process Control) is a family of advanced control techniques that runs on top of a plant’s basic PID control loops and uses process models to steer several process variables at once, so that industrial processes run more stably, more efficiently and closer to their limits. The best-known APC technique is Model Predictive Control (MPC), but feedforward, soft sensors and real-time optimisation also belong to the family. APC does not replace the existing loops: it calculates an optimal setpoint and passes it to the PID controllers in the DCS.

APC helps organisations to save energy, improve quality, reduce variability and make better use of operating limits, particularly in complex, continuous production processes.

🧠 Where does APC sit in the control hierarchy?

Process control is built in layers, each working on its own time scale. At the bottom are the sensors, control valves and PID loops that respond within seconds. Above them sits APC, which typically executes once a minute. Another layer up, optimisation decides which operating point is economically best.

In the Purdue Model, basic control runs on level 1 and operators supervise the process on level 2. The APC server usually sits on level 2 or at the boundary with level 3, close to the DCS and the historian.

🔧 How does an APC controller work?

An MPC-based APC controller works with three kinds of variables:

• Manipulated variables (MVs) - the handles APC is allowed to move, usually setpoints of underlying PID loops (see Manipulated Variable)
• Controlled variables (CVs) - what must stay within limits or on target, such as a product quality, a temperature difference or a valve position
• Disturbance variables (DVs) - measured disturbances that APC cannot move but can anticipate, such as feed rate or ambient temperature

Every cycle, the controller reads the current process values, compares the prediction from the previous cycle with the measurement and corrects its model with that difference (the bias). It then first calculates an optimal steady-state target within all limits, followed by a sequence of future moves to get there. Only the first move is executed; a minute later the calculation starts again. This is called the receding horizon principle.

The benefit comes from squeeze and shift: because APC reduces the variability of a variable (squeeze), its average can safely move closer to a limit (shift). A distillation column that used to run with a wide purity margin can then, for example, meet the same specification with less steam or a higher throughput.

📦 Which APC techniques exist?

Soft sensors are often the quiet force behind APC. A laboratory analysis may only be available after several hours, whereas a well-calibrated soft sensor provides an estimate every minute. Modern soft sensors use machine learning alongside regression, which links APC to Industrial AI and the digital twin.

🕰️ How did APC originate?

APC emerged in the petrochemical industry in the 1970s, when process computers became powerful enough to run models online. Richalet and colleagues published on model predictive heuristic control in 1978, the basis of IDCOM. At Shell Oil, Cutler and Ramaker developed Dynamic Matrix Control (DMC), which they presented in 1979 and which remained a trade secret until 1980. Later algorithms included QDMC (1983) and Shell’s SMOC (1988).

A widely cited survey by Qin and Badgwell counted, based on vendor data from 1999, more than 4,500 linear MPC applications, up from about 2,200 in late 1995. Roughly two thirds were in refining and petrochemicals, followed by chemicals and pulp and paper. The major suppliers since then include AspenTech (DMCplus), Honeywell (RMPCT, later Profit Controller) and Yokogawa (licensing SMOC), alongside APC functions built into DCS platforms from vendors including Siemens, ABB and Emerson.

🏭 Where is APC used?

APC pays off mostly in continuous processes with strong interaction between variables, long dead times and high throughput, where a small percentage improvement is worth a lot of money:

• Refineries - crude distillation, fluid catalytic cracking (FCC) and hydrocracking
• Petrochemicals and chemicals - ethylene furnaces, polymerisation, distillation trains
• Gas processing and LNG - compression and separation
• Energy and utilities - boilers, steam networks and combined heat and power plants
• Cement, minerals and paper - kilns, mills and paper machines
• Food and pharma - evaporation, drying and fermentation; in batch processes mainly soft sensors and trajectory control

In the Netherlands, APC is common in refining and petrochemicals in the Port of Rotterdam and at the large chemical clusters. There, APC often runs inside installations covered by the Seveso-based Brzo regime, which raises the bar for change management and cybersecurity (see Seveso and Cybersecurity).

🧭 How do you approach an APC project?

An APC project usually follows these steps:

1. Benefits study - use historical data from the historian to find where the economic value lies and which limits currently constrain the process
2. Fix the base layer - repair poorly tuned PID loops, sticking valves or faulty meters first; APC cannot compensate for a weak foundation
3. Functional design - choose MVs, CVs and DVs and agree limits together with operators and process engineers
4. Step testing - move MVs systematically (in steps or with PRBS signals) and record how the CVs respond
5. Model identification - derive a dynamic model from the test data, optionally combined with first-principles models
6. Simulation and commissioning - test the controller offline in closed-loop simulation, let it run online in prediction-only mode first, then switch it on step by step
7. Maintenance and monitoring - track performance and update the model when the process changes

The benefit and payback figures come largely from vendor and practitioner publications; a site-specific benefits study is always needed.

⚠️ Why do APC controllers degrade?

APC performance does not last by itself. Catalyst ageing, fouling, new feedstocks or a changed throughput gradually make the model inaccurate. Operators then switch off a controller that makes strange moves, and practitioner studies report that a large share of installations fall back to pre-APC performance or are switched off within a few years. Structured monitoring of the service factor, the number of CVs running at a limit and the model quality, plus a periodic re-test, prevents this.

🔐 What does APC mean for safety and OT security?

• Not a safety function - APC respects operating limits, but safety limits are protected by a separate SIS in line with functional safety. APC must never be credited as a protection layer.
• Write access - the APC server writes setpoints to the DCS. That makes it a valuable target: whoever takes over the server can steer the process within the limits the DCS allows.
• Zoning - place the APC server in its own zone according to the zones and conduits model of IEC 62443 and restrict the conduit to the DCS to the tags that are actually needed.
• Legacy connections - many sites still use OPC Classic with DCOM; migrating to OPC UA with authentication and encryption reduces the attack surface.
• Change management - model updates and new limits are process changes and belong in a formal change process.

❓ Frequently asked questions

What is the difference between APC and PID control?

A PID controller steers one process variable with one output and reacts to a deviation that has already occurred. APC uses a process model to look ahead, controls several variables at once and takes limits into account. APC usually passes its result as a setpoint to the existing PID controllers.

Is APC the same as MPC?

No, Model Predictive Control is the most important technique within APC, but not the only one. APC also includes feedforward, decoupling, soft sensors and real-time optimisation. In refining and petrochemicals, however, the two terms are often used interchangeably in practice.

What is a soft sensor in APC?

A soft sensor is a calculation model that estimates a quantity that is hard to measure, such as product purity, from ordinary measurements like temperature, pressure and flow. Within APC, a soft sensor delivers a value every minute, whereas a lab analysis can take hours. The soft sensor is regularly corrected with lab results.

How much does APC deliver?

The return on APC depends heavily on the process and the starting situation. Published practitioner figures often mention 1 to 5 percent more throughput or lower energy use, with a payback time of a few months to two years. A benefits study on your own process data gives the only reliable estimate.

Can APC protect the safety of a plant?

No, APC is an optimisation layer, not a safety system. Safety limits are protected by a separate safety instrumented system (SIS) that works independently of APC and the DCS. Operators must also always be able to switch APC off, after which basic control keeps the process stable.

Does APC work for batch processes?

APC is mainly used in continuous processes, because the operating point there stays the same for long periods. In batch processes, APC mostly takes the form of soft sensors, trajectory control and end-point prediction, for example in fermentation. Full MPC for batch processes exists, but is more complex because process behaviour changes strongly during a batch.

📌 In summary

Advanced Process Control (APC) uses process models and predictive algorithms on top of existing PID control to steer several variables at once, so that industrial processes run more efficiently, more consistently and closer to their limits. Success depends on a healthy base layer, good models, structured maintenance and a securely designed connection to the DCS.