The goal of automation is to reduce necessary human labor time while improving safety, quality, efficiency, and profitability. Every operational intervention embodies knowledge and represents a problem to be solved, with the ultimate aim of automation being to minimize such interventions. Problems in automation can be categorized into three types:
Deterministic Problems. Many manual operation issues are not inherently complex but are overlooked, with solutions yet to be identified. These problems often have solutions at broader scopes or higher levels, and cross-departmental collaboration can facilitate finding them. For example, standardized operating experience can be implemented with sequential control programs; unreasonable process targets causing frequent interventions and plant fluctuations can be resolved by adjusting targets; raw material-induced fluctuations are best addressed by coordinating with upstream units or procurement; and unstable utilities can be improved by enhancing utility systems.
Uncertain Problems with Deterministic Causal Relationships. These problems require feedback to reduce uncertainty. Without feedback, a feedback control loop must be established based on causal relationships. Once a feedback loop exists, it must meet performance requirements. The primary solution is PID parameter tuning. If performance remains inadequate, advanced structures like cascade or feedforward control can be considered. Some cases may require redesigning the process variable (PV) to strengthen and stabilize causal relationships, such as using a ratio instead of a difference to reduce nonlinearity or implementing PV signal filtering and anomaly rejection.
Uncertain Problems Requiring Causal Relationship Determination. These involve multivariable control problems requiring control scheme design. Implementing multivariable control through single-loop structures is the core approach. Decomposition, simplification, and reconstruction are fundamental strategies. Manipulated variables (MVs) should remain consistent, as override control often deviates from simplicity and optimality. Override control is not about control but safety—a perspective that requires experience to appreciate. Control schemes combining split-range and override control are rarely simple or optimal. The third type of problem is strategic: without reassigning variable pairings, performance-based solutions are often inadequate or overly complex. For example, in an energy-saving control scheme for mutual material supply (as shown in the referenced diagram), manually operating valve V1 may suffice if upstream and downstream unit loads are stable, requiring minimal intervention. However, the presence of a tank indicates potential load imbalances. Automating V1, which affects two liquid levels, is an uncertain problem requiring causal relationship determination. Thoroughly studying this problem enhances control scheme design capabilities. I’ve researched this scheme for two to three years, and recent discussions still yield new insights. Solutions vary depending on whether upstream and downstream units operate in batch or continuous mode, or if pressure control shifts to flow control. Similarly, adding new operational methods requires redesigning the control scheme. For highly complex disturbances, robust design is essential to maximize disturbance rejection.
In many plants, single-loop control dominates—not only due to technical limitations but also because simplicity and optimality are inherently tied to single-loop structures. Cascade control is more common in performance improvements because ideal and real-world conditions differ, and feedforward control is challenging to implement effectively, offering low cost-effectiveness. Split-range and override controls are less common due to limited demand, and their frequent use in control scheme design often indicates misuse. This observation also applies to the control scheme in the referenced diagram. Schemes and tuning can address many process control problems. However, this case study shows that implementing PID can be complex, requiring experience, knowledge, and attention to detail.
What about advanced control? Advanced control simplifies the process: a controller with a fixed model of two controlled variables (CVs) and three MVs can adapt to various operating conditions with different parameters. Advanced control offers a unified paradigm for addressing multivariable control, reducing the complexity of scheme design and enabling flexible handling of uncertainties.