The journey of modern process control started when the necessity of ensuring process safety and regularity arose in the process industry. Back in 1950s was the time when process control started to be acknowledged as yet another inevitable module of chemical and petrochemical industries. In those days PID (proportional integral derivative) controllers, as perhaps the only control concept, were implemented pneumatically. This did not mean that digital control was not examined. In fact, in late 1950s process control industry had experimented Digital Control and the conclusion they arrived at was: “it is not worth it!”. While this was a true conclusion at the time – due to being expensive, complex, and not being able to generate enough economic profits – we all know how this is changed today.
While in early years control practitioners came to the conclusion that only two layers of control – a lower layer for regularity control and an upper layer for steady-state optimization – were required, with the processes becoming more complex over the years, the control systems had to catch up. By 1970s, Advanced Process Control (APC) concepts were introduced aiming at enhancing process efficiency and reducing variability in product quality. Owing to the growing economic competitions in those years, it became evident, more than ever, that profitability is directly related to the product quality. In addition to the trade offs between process productivity and product quality, energy efficiency and environmental footprint of chemical processes also became of significant importance. These considerations were translated into much more complex processes, which required increasingly complex control systems.
It was in this era that Model Predictive Control (MPC) was born in the chemical industry in order to facilitate more effective control of processes, and was later adopted by academia to further develop its theoretical foundations. MPC controllers rely on a mathematical model of the process to predict the process behavior over a future time horizon. This enables optimization of the process behavior in terms of desired process performance. In simple words, with MPC, we are able to predict the future behavior of the process and act according to anticipated deviations from the target control objectives. MPC has brought a substantial improvement to our process control capabilities in meeting conflicting control objectives in the process industry. It is fair to say that the more accurate the process model is, the more effective MPC would work. For this reason, there has also been an increased emphasis on modeling aspects of chemical processes over the years. Let us not forget that every mathematical model is an abstract and inaccurate representation of reality. While models can be made overly complicated, the computational cost of the complex models may make them less suitable for on-line control applications. In addition to plant-model mismatch, chemical processes exhibit considerable uncertainties and disturbances. This has led to emergence of robust MPC concepts since late 1980s. The idea is to capture as much of the uncertainties and disturbances as possible to robustify the designed control inputs to process perturbations.
In recent years, advances in novel process designs and manufacturing practices have brought about new challenges and opportunities for the process control community. One such is around the importance of batch processes that are commonly used for low volume and high value-added manufacturing. Traditionally, process control was implemented mainly on continuous-flow processes. Today, however, many important processes run in batch mode. Such processes can be found in, e.g., specialty chemical, pharmaceutical, food, and biotechnology industries.
The inherent dynamic nature of batch processes has motivated the process control community to bring in new developments.
Another recent trend is the growing importance of the notion of transient process modes. In intensified processes and miniaturized chemical systems as well as advanced manufacturing systems, it is more important than ever to be able to control the processes during their transient modes in light of economic considerations. Looking into the future, the major domains for the process control practitioners and academics are seen to be big data in the process industries, advanced energy systems, and health-care applications. It is now up to the process control community to embark on new endeavors to best play their role in addressing the societal needs of the current era.
Article provided by
University of California, Berkeley, USA
IFAC Technical Committee 6.1. Chemical Process Control