On evaluating robustness of fuzzy logic controllers through Taguchi methodology

This paper completes the set of pre-established settings for PID-like fuzzy logic controllers obtained at the Laboratoire d´Electrotechnique et d´Electronique Industrielle. These settings are based on just one open-loop experimental response such as Ziegler-Nichols or Broida methods for conventional controllers and lead to very powerful industrial tuning of fuzzy logic controllers (FLC) for nonevolutive and open-loop stable systems. Until now, these settings have not taken into account the robustness of the fuzzy control law. In this paper, an original method for evaluating robustness of a fuzzy logic controller is described. This method is based on Taguchi quality methodology and experimental product-plans. The robustness is evaluated in the case of white noise, open-loop process misidentification or a high order process. The Taguchi quality methodology is here applied on a very common industrial benchmark and considering an IAE-criterion to obtain a new set of pre-established settings, that are robust ones and come in complement of the existing set. The validation of these settings is done on two experimental systems: a DC/DC buck converter and a permanent magnet synchronous machine