When advanced control technologies began to gain traction, it was often claimed that they offered a smoother control process and better benefits, as depicted in the graph above. However, the truth is that any control technology, when properly implemented, can yield similar outcomes. Today, our understanding of PID control has evolved significantly. We now recognize that in single-variable control, advanced control and PID can deliver comparable performance. Therefore, advanced control has not become popular simply as a replacement for PID; instead, it offers additional advantages in specific contexts. In many deterministic problems, PID still provides a cost-effective solution.
Despite the scrutiny surrounding advanced control, its gradual adoption is driven by other factors. Although a key characteristic of advanced control is model prediction, using models is a common feature in modern control algorithms. Even PID, which appears to be based on deviation without a model, implicitly incorporates a model when tuning parameters to achieve optimal performance. So, even PID has an implicit model embedded within it.
Advanced control also features feedback correction. It adjusts the model’s predictions using real-time process measurements. If advanced control doesn’t use real-time process measurements for feedback correction, it ceases to be a feedback control system, which makes it difficult to handle uncertainties. Feedback correction is a fundamental property of all feedback control systems. While PID uses deviation for feedback correction, advanced control uses more complex algorithms to perform this correction.
As industrial plants grow in size and complexity, managing hundreds or even thousands of control loops becomes a daunting task. Changes in operational conditions, equipment, or processes make it difficult for PID to maintain safe and efficient operations. Even with structured approaches to decouple, simplify, and layer control problems, solving these issues with PID remains a significant challenge for most engineers. Advanced control, on the other hand, uses linear multivariable control algorithms, which offer a higher cost-effectiveness for solving multi-variable control problems.
In processes like oil blending, where several components are needed to control 20 variables, or in flotation systems, where pump tank levels and flotation foam thickness are interrelated, advanced control is often the more practical solution. Its ability to efficiently handle multi-variable control at a lower cost is a benefit that has not been fully appreciated.
In today’s era of big data analysis, many companies promise customers significant benefits through data-driven insights. However, many of these suggestions are either impractical or difficult for customers to implement. Often, customers are seeking help to actually achieve these benefits, not just theoretical advice. Without an industrial background, big data analysis can quickly become mere tool-driven or formalistic endeavors. As the saying goes, “Customers don’t want to buy a drill, they want to make a hole in the wall.” Yet, we often find ourselves selling the drill instead.
In control optimization projects, we emphasize technology transfer, but it’s essential to recognize that such transfer requires many conditions to be successful, and the results are not always guaranteed. In any project, helping the customer solve their actual problems is more important than simply empowering them. If advanced control software is provided without reducing operational intervention or improving safety and efficiency, the customer is unlikely to invest in it. When customers realize that a flashy tool doesn’t address their pain points, that’s when new technologies begin to fade.
Advanced control, like artificial intelligence, must solve real industrial control problems. It’s crucial that we identify the right issues to address and how to integrate them with process control. This remains an ongoing exploration and one that requires careful consideration.