Object-oriented signal modeling using an adaptive neuro-fuzzy inference system
This research explores the use of fuzzy rules for capturing the underlying dynamics of object-oriented signals and for extracting relationships between signals. An Adaptive Neuro-Fuzzy Inference System (ANFIS) is used to generate the fuzzy rules. The ANFIS is able to model nonlinear, somewhat chaotic, signals which would be otherwise very difficult to model using traditional modeling techniques. A Matlab implementation of the ANFIS is developed, and the ANFIS is trained using a Genetic Algorithm. The ANFIS is trained and validated using highly nonlinear exothermic reaction process data. This modeling approach proves to be well suited for the modeling and prediction of object-oriented signals whose characteristics are poorly understood and shows good potential for relationship extraction between pairs of signals. The approach works well for data sets which were available, and should scale well. It is reasonable to believe that using this modeling technique for controlling object-oriented signals is feasible since the use of fuzzy sets has been established as being very appropriate for controls applications.
Thesis97H42.pdf
3.34 MB
Unknown
9642a45dc48cb00e36782d1d813c84ab