Document Type

Theses, Ph.D


Available under a Creative Commons Attribution Non-Commercial Share Alike 4.0 International Licence


The local model (LM) network is considered for the control of complex nonlinear systems. Both controller design and system analysis techniques are investigated for the Purpose of the development of an overall global controller with guaranteed stability and performance, based on the control methods and theories well developed for linear systems. In particular, the influence of the offset term of affine LM networks on the performance and stability of closed-loop systems is investigated. Assuming the system changes ‘slowly’ enough, an integrator can be utilised in the controller design by considering the offset term as ‘constant’. Gain-scheduled local controller (LC) networks based on feedback control and generalised predictive control methods are proposed for the control of a highly nonlinear simulated process, the continuous stirred tank reactor. Test results show reasonably good performance, but also expose the weakness of the interpolation procedure. Considering the nonlinear dynamics inherent in the interpolation procedure, the stability issues association with blending affine systems are investigated. Assuming the corresponding linear blending system is exponentially stable, the affine blending system will be bounded. The ultimate bound is determined for both open loop systems and closed-loop systems via affine state feedback control. The velocity-based (VB) LM network has enhanced the capability to capture the dynamics of nonlinear systems compared to normal LM networks. To have the best access to dynamical information from the VB model, a state feedback integral controller is skilfully proposed for the controller design. This approach overcomes the difficulty in the practical implementation of velocity-based approaches. In addition, a discrete-time version of the velocity-based LM network structure is proposed for the purpose of extending the VB approach to the discrete time domain.