DESIGN AND ANALYSIS OF INTELLIGENT CONTROLLERS FOR NON-LINEAR SYSTEMS

Abstract

All the practical systems in the real world are mostly non-linear in nature due to their dynamic behavior. Non-linear systems are extremely complex due to their uncertainties, parameter variations, and other complexities. Soft computing methods are commonly employed to address system dynamics and uncertainties. One of the most effective and successful soft computing methods is optimization. Due to numerous applications, optimization is a significant paradigm. Minimization or maximization are the forms of optimization that occur in nearly all engineering and industrial applications. The necessity to manage complex nonlinear processes with a high degree of uncertainty and satisfy performance requirements are the main drivers of progress in the field of control. Conventional control techniques cannot meet these objectives and have been proven ineffective for complicated nonlinear systems due to their complexity. To enhance the system's performance, an appropriate controller with optimal parameters is needed to be employed. Conventional Proportional-Integral-Derivative (PID) is still utilized in industries and other real world applications because of its simplicity. This thesis uses PID controllers with various optimization techniques, such as Particle Swarm Optimization (PSO) and Teaching Learning Based Optimization (TLBO) algorithm, to control non-linear benchmark systems such as artificial respiratory systems and vehicle cruise control systems. Performance indices were used to evaluate the suggested controllers' effectiveness. The robustness of the controllers was also examined. The suggested techniques were also used for the control and tuning of a non-linear ball and beam system by cascade-optimized PID. The controllers' ability to reject disturbances and variations in parameters was examined. Fractional order PID controllers (FOPIDs), which offer more controller flexibility, are frequently employed for the control of non-linear systems. Whale optimization algorithms (WOA), TLBO, and PSO are used in this thesis to tune the FOPID controller and control the inverted cart pendulum system. Several advantages of artificial neural networks have made them particularly attractive for use in modeling and controlling complex non-linear systems. They are capable of adaptation and self-learning. In this thesis, a new PID-like neural network is proposed, whose weights are optimized by the PSO algorithm. The optimization of weights with PSO has many advantages as compared to conventional back propagation learning since it is not based on gradient calculations. The proposed controller was tested for its effectiveness on a highly non-linear Continuous stirred tank reactor. It was used to control the temperature of the CSTR. vi The proposed controller was also tested for robustness under disturbance application. It was also compared with the optimization-based PID controllers.