Project (and later Master) for Sigurd Skogestad 2023-24. --- 1. Tuning of cascade control systems Project SiS-1. Cascade control is widely used in industry, yet the design methods are not well developed. In particular, it is not clear what the time scale separation between the loops should be. What happens if the time scale separation gets too small, for example, less than 4? In the projects, examples will be used from the Perstorp Company (Professor Krister Forsman), Some details: Tuning of the two controllers should be done sequentially. The inner (secondary) controller C2 (fast) is tuned first based on the process G2, and with this loop closed, the outer (primary) controller C1 (slow) is tuned. It is strongly recommended to use a design method (e.g. SIMC PID-tuning) where the closed-loop time constants tc1 and tc2 are used as design parameters. A rule of thumb is that the time scale separation between the loops should be between 4 and 10. A larger time separation helps to protect against process gain variations in both the inner and outer loops, Note here that a process gain decrease inner loop is “bad” as it translates into a larger (”slower”) value of the actual tc2. This reduces the time scale separation tc1/tc2 and in addition tc2 appears as an effective delay as seen from the outer loop. On the other hand, for outer loop, a process gain increase is “bad” as it translates into a smaller (”faster”) value of tc1 which reduces the time scale separation. For this reason, it is often recommended to have a time scale separation of 10 (or larger). However, the disadvantage with a too large time scale separation is that it ”eats up” more of the available time window, which may be a problem with many layers of cascade control. To avoid eating up the time window, the solution is to tune the inner loops more tightly. Parameters to be considered is the gain, delay and time constant in the two process parts (G1, G2).To evaluate performance one may consider IAE and robustness margins. --- 2. Tuning of anti-windup schemes Project SiS-2. There exists many approaches to anti-windup, and the first task is to compare the three most common schemes on some case studies 1. Limiting the bias 2. Anti-windup based on positive feedback implementation of I-action 3. "Back-calculation" scheme with extra tracking time parameter. The next step is to make recommendation of how to choose the tracking time, for example, when we need anti-windup for CV-Cv switching, decoupling or cascade control. --- 3. Structure and simulation of selector logic Project SiS-3. Selectors are used for steady-state CV-CV switching and in a few cases for dynamic reasons (e.g., cross-limiting control). In the project, the aim is to simulate some typical process examples (see the paper). Anti-windup using back-calculation should also be included. --- 4. Simple schemes for nonlinear decoupling and feedforward control. Project SiS-4. The objective is to propose and simulate some simple schemes for nonlinear decoupling and feedforward control om industrially relevant case studies. --- 5. Default tuning of PID controllers (including scaling of variables) based on limited information. Project SiS-5. It is important to have simple rules for default tuning in order to avoid spending too much time on PID tuning during startup. The first part of the project is to search the literature for such rules. Next to propose possible new rules. Finally, to simulate the rules. --- 6. Comparison of selector on input or setpoint (cascade) for CV-CV switching Project SiS-6. The objective is to compare the two options by simulating some industrial cases. --- 7. Comparison of alternative schemes for MV-MV switching Project SiS-7. There are three alternative schemes. These should be compared on typical industrial case studies. --- 8. Floating pressure or temperature control. Project SiS-8. It may be possible (and desirable) to have the same variable being controlled twice in the same cascade hierarchy. For example, one may have two pressure controllers on top of each other (one fast for stabilization and one slow for optimization with sets the setpoint to the fast controller), or there may be a VPC in between so that pressure is “floating" (uncontrolled) on an intermediate time scale. Another case may be stabilizing control of a reactor, with PD-control in the inner loop, with setpoint set by a slow PI-controller. --- 9. Linear decoupling with reverse (feedback) implementation Project SiS-9 The task is to compare the performance of the reverse implementation of Shinskey with the more common implementation. In particular, compare the two for cases with input saturation and switch between manual and auto. Also consider the case where the RGA is close to 0.9 and there may be internal stability problems with the reverse implementation. 10. Reactor temperature control for silver catalyzed formalin plants (Dynea) With sumnmer job (Nils Arne Susort) X. Also other topics are possible. Please contact Sigurd Skogestad [skoge@ntnu.no] --- General text (for all project proposals): Simulations may be with Matlab, Simulink or Hysys. For updated to the project proposal and other projects from Sigurd Skogestad see here: https://folk.ntnu.no/skoge/diplom/prosjekt23/ This project is part of an effort to improve the theoretical basis for "advanced regulatory control" solutions. The topics will also be covered in the module "Advanced pocess control". The project may also be extended to a Master thesis on this or a related topic. For more details about the problem and motivation behind the project, please see the following paper: S. Skogestad "Advanced control using simple elements - An important and challenging research area". Annual Reviews in Control (2023). https://folk.ntnu.no/skoge/publications/2023/skogestad-advanced-regulatory-control_arc/ This paper ends with the following conclusion: In summary, it is proposed that a lot more academic research is focused on developing theory for the advanced regulatory control solutions described in this paper. Indeed, the problems are very challenging. For example, the mathematical problems related to the optimal decomposed and decentralized control solutions are in general non-convex, and the stability analysis of switched systems (for example, with selectors, anti-windup and split range control) is very difficult and may result in limit cycles and chaotic behavior. This, in addition to an unclear problem definition, may scare academic researchers away, but hopefully the importance of the problem and the prospect of seeing the solutions being used in practice and thus benefiting humanity, may provide motivation to consider these challenging problems.