Development of control strategies for single input single output nonlinear systems
University of New Brunswick
Industrial-scale systems exhibit nonlinear behaviour such as changing system gain, time constants and overshoot, as well as other types of nonlinear behaviour include dead-zones, dead time and discontinuous system response. Some specialized industrial systems also require that the process track a desired trajectory, which may be nonlinear in nature. Typical control strategies rely on obtaining linear approximations to nonlinear systems and then designing a set of controllers to achieve good closed loop performance. Such control approaches are often complex and application-specific, thus limiting their usability in industrial applications, specially when considering the desire to track a profile. The doctoral investigation focused on developing control algorithms specifically designed to handle nonlinear systems. An important consideration is the applicability of the proposed control algorithms to real-world systems. The investigation begins by considering the proposed nonlinear compensator. Using a known nonlinear system, a compensator is designed such that the closed loop system exhibits desirable behaviour in terms of a predetermined rate of decay, stability and zero steady state error. The approach was implemented in several nonlinear systems with good closed loop performance. The second proposed approach is called the nonlinear predictive controller. Modifications to the predictive controller architecture were made to improve tracking and system modelling such that it is able to handle nonlinear systems. Two types of control move optimization are offered: least squares optimization is efficient for a simple system while golden section optimization places few constraints on system type and order but requires more computational resources. Improved robustness means the nonlinear predictive controller is able to achieve excellent closed loop performance with accurate tracking. Its applicability to industrial systems was demonstrated by implementing the scheme to control injection speed and mould opening and closing in an injection moulding machine. Its versatility is also studied as it was used to control system exhibiting complex nonlinear behaviour.