Research On Nonlinear Adaptive Control Methods For Block-oriented Models

Block-oriented models are a class of typical nonlinear models that have special structures.Such models are built up from nonlinear static mappings and linear time-invariant dynamic systems in various forms of interconnection.Block-oriented model structures have attracted considerable attentions in the past decade since they have a remarkable ability to represent nonlinear dynamics of many industrial processes.Meanwhile,they possess another significant merit in terms of the flexible separation characteristics between nonlinear static parts and linear dynamic parts.Essentially compared with traditional nonlinear dynamic representations,the special structure properties are advanced and attractive,which bring in some advantages,such as ease of model development,possibility to incorporate a prior process knowledge and suitability of being used for control.Based on the parameterized Hammerstein,Wiener and Hammerstein-Wiener block-oriented models,this research intends to propose some novel adaptive control schemes by the use of measurable input-output signals and available model structure properties.All these adaptive control algorithms are motivated by the self-tuning control framework.Under some reasonable assumptions,this research has also achieved the closed-loop stability and the output tracking properties for each nonlinear adaptive control system.Besides,some representative simulation experiments including industrial nonlinear processes are provided to verify the effectiveness of each nonlinear adaptive controller.Briefly,the main contents of this research are summarized as follows:1.A robust adaptive control algorithm is proposed for uncertain Hammerstein block-oriented models with unstable zero dynamics,non-Gaussian noises and unmodeled dynamics.The adaptive control scheme is derived from a modified performance index.The uncertain system parameters are updated by a novel recursive identification algorithm with a weighted factor.The unmodeled dynamics estimation is introduced as a feedback to compensate the model error.Motivated by the self-tuning adaptive control theory,it is proved that the closed-loop system stability can be guaranteed under certain conditions.Simulation examples including a chemical process control problem are studied to test the applicability.Besides,a deadzone compensating adaptive control algorithm is also proposed for a class of special Hammerstein models with deadzone input nonlinearity,which is applied to a simulated motor speed control problem.2.A robust indirect adaptive sliding mode control algorithm is proposed for uncertain Wiener block-oriented models with unstable zero dynamics,non-Gaussian noises,unmodeled dynamics and distinct nonlinear behaviors.The parameterization Wiener models are obtained by incorporating a linear transfer function with a radial basis function network.The adaptive control scheme is derived from a modified sliding surface with online adjustment.The uncertain system parameters together with the upper bound of the unmodeled dynamics are updated by a novel recursive identification algorithm with a weighted factor.The weighting polynomials of the sliding surface are calculated online based on the pole assignment technique.The unmodeled dynamics estimation is introduced as a feedback to compensate the model error.Moreover,by the use of the time-varying operation,it is proved that the closed-loop system stability can be guaranteed under certain conditions.Simulation examples including two chemical process control problems are studied to test the applicability.3.It is well known that Hammerstein or Wiener models can be represented as bilinear parameterization models.Compared with these simple block-oriented models,the analysis for Hammerstein-Wiener models is much more difficult since no bilinear Hammerstein-Wiener parameterization models have been reported-To address this key issue,this research presents a novel parameterization model which is motivated by the key term separation principle.The uncertain system parameters and the unmeasurable internal variables are updated by a novel recursive identification algorithm with a weighted factor.Based on the online estimations,the control law is derived from the certainty equivalence principle.Since there exist estimated variables in the information vector,a novel theoretical framework about adaptive control is proposed.Under reasonable conditions,it is proved that the closed-loop system stability is guaranteed in a local domain.Furthermore,some important modifications are also discussed,such as how to enlarge the locally stable domain,how to extend this algorithm to a colored noise situation and so on.The effectiveness of this adaptive controller is verified in some representative simulation examples.