Currently most controllers used in the chemical industry are based on conventional PID algorithms. Priorly recognized for their simplicity and robust structure it is difficult to find good controller settings even though PID controllers only have a few adjustable parameters . In this work, we present a new approach which enables online tuning of conventional controllers. The advantage it has over conventional methods lies in the online model identification by closed-loop experiments. These enable us to avoid undesirable state variable changes. As a result, we are able to identify the model parameters more quickly, which in turn reduces the number of necessary experiments and the amount of measured data. Moreover, this leads to a faster calculation of controller settings and reduced commissioning time. Thus, in this contribution, we describe in detail how the proposed approach can also be automated by eliminating the need to manually tune and retune PID controllers, when they are deployed or subject to system behavior changes.
In most cases a good process model is required for controller tuning. Controller settings are first determined by identifying the process behavior, which are usually calculated by time-consuming and expensive experiments. Therefore, it is preferable to design experiments in such a way that the resulting data yields maximum information with respect to model parameters. For this purpose, we use the approach of optimal experimental design . By calculating optimal conditions, it maximizes information yield produced by experiments in order to increase the accuracy of estimated model parameters. The challenge of our approach is the need for online model identification by closed-loop experiments. For this purpose, we use a technique proposed by Körkel and Arellano-Garcia . According to this concept, parameter estimation is applied in order to match the model to a real process. Applied to our domain, the outcome is a sequential approach to calculate model parameters in three steps. In the first step, the experiment is planned. Here, the optimal value of the control variable throughout the experiment is determined. Next, the planned experiment is executed and the measured data is analyzed. In the last step, the quality of the model is validated and new model parameters are calculated by applying the least square approach . After the process model is identified, we calculate the controller settings and implement them in the online mode.
This work introduces the proposed solution strategy for the online controller tuning and closed-loop model parameter identification algorithm. In order to validate the approach and supply a comparison to common tuning methods, experimental results will also be presented. Furthermore, we will discuss the automated method's performance and influence to commissioning time.
 Shamsuzzoha M. and Skogestad S., The setpoint overshoot method: A simple and fast closed-loop approach for PID tuning, J. Process Control, 2010, 20, 1220-1234.
 Rivera D.E., Morari M., Skogestad S., Internal model control. 4. PID controller design, Ind. Eng. Chem. Res. 1986, 25(1), 252-265.
 Skogestad S., Simple analytic rules for model reduction and PID controller tuning, J. Process Control, 2003, 13, 291-309.
 Ziegler J.G., Nichols N.B., Optimum settings for automatic controllers, Trans. ASME, 1942, 64, 759-768.
 Barz T., Arellano-Garcia H., Wozny G., Handling uncertainty in model-based optimal experimental design, Ind. Eng. Chem. Res., 2010, 49, 5702-5713
 Körkel S., Arellano-Garcia H., Online Experimental Design for Model Validation, 10th International Symposium on Process Systems Engineering – PSE2009, 2009