BOKSUWAN SUNGWAN

Department & title：

Graduate Student, Dept. of Mechanical Engineering and Intelligent Systems

Destination：

Department of Chemical Engineering, Norwegian University of Science and Technology(NTNU), Trondheim, Norway.

Period：

from 28 AUG. 2012 to 23 SEP. 2012

Research theme:The implementation of Model Predictive Control for Optimal steering of Micro Magnetic Manipulators.

Summary of research activities during your overseas visit

The main research activity is mainly focused on the implementation of the optimal control structure which is called global and local control. The implementation of the model predictive control(MPC) is divided into two layers called Global and Local ones. In the local control, the command is supplied by the global control. The control structure in the local control is very important in that what variables should be kept constant. Therefore, the research conducted at the NTNU University is about choosing the best "magic" variables when those variables are kept constant the system or process is running in the optimal way or at least near the optimal policy. One of the most important technique is the null space method which is widely applicable. One example addressed the idea of self-optimizing control is long running.
We select self-optimizing controlled variables as linear combinations c=Hy of a subset of the available measurements y. With no implementation error, it is locally optimal to select H such that HF=0, where F is the optimal sensitivity, with respect to is disturbance d. However, ignoring the implementation error is a serious shortcoming for some applications. To compensate for this partially, it is important to use measurements y that are independent and not sensitive to measurement error. Another shortcoming is that a new set of controlled variables(for the unconstrained degrees of freedom) must be found for each possible set of the active constraints. The global properties of the proposed variable combination c)Hy must be evaluated by computing the loss for expected disturbances and implementation errors using the nonlinear model.
In the local control, the implementation is based on the PID controller which proposed by a host professor himself. This table is based on SIMC-rule with the optimal for a given robustness (Ms value) shows that the SIMC-rule give settings close to the Pareto-optimal. The exception is a pure time delay processes where the SIMC-rule gives a pure integral controller with somewhat sluggish response. In addition, the approximation of the system is considered because the PID turning table requires the systems to be described in terms of a first-order plus time delay or a second-order plus time delay.

Research outcome obtained

After the research completed, I got the implementation of the model predictive control(MPC) which is divided into two layers called Global and Local control. In the global control, I can implementation by the model predictive control or we call supervisor control. The link between global and local control is the reference signal which is the variable kept constant. The main research is focused on how to select this magic variable. The technique is called "Self-optimizing control". After I came back I can applied this technique in my current research. The main idea of this technique is called Null space method which offers the optimal operation of the system. Actually, Self-optimizing variable is widely applicable for example, companies business, consumers, chemical process plants, biological systems, and so on. Therefore, the results of this research are very useful and widely applicable.
For the local control, the research was focused on the PID controller which the host professor proposed the parameter turning table himself. Nowadays, this table is considerably used in the chemical process plants. This technique is based on the SIMC method. The basis for the SIMC method is a first-order plus time delay model. I learned a new effective method to obtain the model from a simple closed-loop experiment. An important advantage of the SIMC rule is that there is a single tuning parameter(tc) that gives a good balance between the PID parameters(Kc;tI;tD;), and which can be adjusted to get a desired trade-off between performance(tight control) and robustness(smooth control).
After almost one month oversea research, I can make the implementation of the control structure called global and local control which is implemented by the model predictive control and PID controller, respectively. The link between the global and local control is the magic variable. This variable is selected by the Self-optimizing variable.

Reflection and suggestion on the program, particularly from the viewpoint of internationalization of researchers and students

This program is very useful and powerful that offers the best chance for researcher, especially young researcher because I am able to exchange the experience with the worldwide host professor who has his own original idea.

Created: November 13, 2012 / Last modified:November 16, 2012