The improvement of soft computing methods and its applications in electric drives

ABSTRACT PhD Thesis THE IMPROVEMENT OF SOFT COMPUTING METHODS AND ITS APPLICATIONS IN ELECTRICAL DRIVES Mehmet KARAKÖSE Firat University Graduate School of Natural and Applied Sciences Department of Electrical and Electronics Engineering 2005, Page: 205 The development of soft computing that consist of many methodologies has attracted considerable research interest over the past decade. Each of these methodologies is capable of model and enables solutions to real world problems which can not be modeled or too difficult to model, mathematically. Fuzzy logic, artificial neural networks, genetic algorithms, probabilistic reasoning, machine learning, chaos theory, support vector machines, data mining, artificial immune systems and rough sets can be given in the soft computing concept. Although soft computing has an important application potential in areas such as control, signal processing, classification, decision making, pattern recognition, robotics and system modeling, it has some common drawbacks. The aim of this thesis is to investigate the soft computing theory, propose new approaches in its components and apply to solve some problems in electrical drives with soft computing methodologies. In this thesis, four new approaches and three electrical drive applications in soft computing methods were studied. In theoretical studies, firstly, a novel fuzzy system approach called block based fuzzy controllers that is constructed by several fuzzy controllers with nine fuzzy rules to carry out control tasks was proposed to design robust, efficiency and low complexity controllers. Secondly, new three membership functions as sinus modulated, parabolic and inverse parabolic for fuzzy logic systems were proposed. Thirdly, new improvements in fuzzy multicriteria decision-making methods using type-1 and type-2 fuzzy sets were used. Main features of these proposed methodologies are the conventional fuzzy integral gets adaptation capability with fuzzy density values, and a decision making method with a combination of fuzzy integral and XVIIItype-2 fuzzy sets improves the flexibility of the control systems. Lastly, a type-2 fuzzy based activation function for multilayer feedforward neural networks was proposed. Instead of other activation functions, the proposed approach uses a type-2 fuzzy set to accelerate backpropagation learning. In electrical drive application section, first study includes a fuzzy filter that is heuristic- knowledge-based algorithm to detect and discard feedback signal error and noise. For this purpose, a fuzzy filter is designed and tested on hysteresis controlled induction motor currents. Second study based on determining output of compensated voltage model with fuzzy algorithm. The proposed algorithm can be used to accurately measure the sensorless vector controlled induction motor flux including its magnitude and phase angle. Final study proposes a fuzzy logic based smooth transition method between flux models for low speed operation of a stator flux oriented induction motor drive. The major task of this transition algorithm is to eliminate jerks on the torque during the transition between the flux estimation models. Briefly, the theory and electrical drives applications of soft computing methodologies were studied in this thesis. Keywords: Soft computing, intelligent systems, hybrid soft computing, electrical drives, sensorless vector control. XIX