Self-Learning Fuzzy Logic Control for Industrial Processes [2015-2019]
My Role: Principal Investigator (Post-Doctoral Researcher)
Host institutions: Institute of Systems and Robotics (ISR-UC)
Financing: Foundation of Science and Technology (FCT)
Reference number: SFRH/BPD/99708/2014.
Abstract: The main objective is to research and contribute for the automatic learning of a Fuzzy Logic Controller (FLC) from data obtained from a given process while it is being manually or automatically controlled, in order to control nonlinear industrial processes. Additionally, the methodologies may also be used to understand a process for which there is little or no information available, since the FLCs are able to gather a knowledge-base about the process control. A current challenge in FLC research is to determine the most suitable fuzzy rules and membership functions of a FLC using data obtained from a given process while it is being manually or automatically controlled. Even, after the learning of the FLC, it is crucial that it can work properly over time, controlling as accurately as possible output variable of the plant, even under operating areas where the train dataset may be not sufficiently representative of the plant, and “unknown changes” of the process. To address these problems, iterative rule learning techniques will be a starting point, where for the unknown operating areaschanges the learning process may create a new fuzzy rule, modify the parameters of an existing one, or merge similar rules. Know issuesproblems with data collection, such as sampling time, missing data, and outliers will be studied due to their influence in the iterative rule learning techniques. Thus, the methodology to be developed should be efficient in terms of performance, adaptivity and robustness.