Some Investigation into System Identification using Soft Computing Techniques

Abstract

This thesis induces the inspiration from the conventional methods of system identification and used the soft computing techniques for linear and nonlinear system identification. The motivation behind the system identification is the requirement to design a better system that would fulfill the industrial demands and could be applied for various engineering applications. System identification is the process of building a mathematical model for the dynamic system if ample amount of input and output data is available. There are numerous extremely difficult applications of system identification and modeling in various fields. Telecommunications signal processing, and automatic control are just a few examples of situations where linear and nonlinear system identification have been widely used. Many researchers have used conventional gradient based algorithms like least mean squares method, Newton method and maximally likelihood method for system identification problems. But complexity, local optimal solution higher, computational time and slow convergence make them inadequate. This leads to the use of soft computing techniques in the system identification area. newline In this thesis nature inspired population based optimization algorithms namely teacher learner based optimization, sine cosine algorithm and arithmetic optimization algorithm assisted with Kalman filter are used and validated. In system identification both linear and nonlinear systems are considered for the identification purpose. Linear systems are more flexible and many systems which are truly linear can be well described by linear systems. Generally, finite impulse response and infinite impulse response systems are preferred as linear systems. In the first part of thesis unknown IIR system is identified using the adaptive IIR system. The design principle of the adaptive IIR system identification considered is to evaluate the system parameters by using adaptive algorithm. newline newline