Neuro-controller Of Cement Rotary Kiln Based On Adaptive Critic Designs

With the development of the Chinese cement manufacturing,the new kind cement rotary kiln has been widely used.The Rotary kiln decomposition is the most important craft tache in the cement production line,and its running status affects the output,quality,energy consumption and environment pollution.It is hard to describe exactly the production processes in the cement rotary kiln,which include fuel combustion,heat transfer,and chemical composition of clinker.The most important reason is that there are chemical reaction,physical reaction and mineralogical reactions simultaneously during the complicated heat transfer process.Stable control of temperature in rotary kiln is critically important.The sintering process of the cement has the characteristics of non-linearity,ultivariable, close coupling,large time lag,and time varying.This causes the kiln-status of the rotary kiln system to be complex.Just as what mentioned above,it is difficult to describe the rotary kiln with an accurate mathematical model.The traditional control method is no longer suitable to control the rotary kiln system.Artificial neural networks(ANNS)are good at identifying and controlling complex nonlinear systems.As they are suitable for multi-variable applications,they can easily identify the interactions between the inputs and outputs.It has been shown that a multilayer feedforward neural network using deviation signals as inputs can identify the complex and nonlinear dynamics of the cement rotary kiln with adequate accuracy to design a controller.A few domestic publications have reported on the design of ANN controllers for the cement rotary kiln.Dynamic programming is a very useful tool in solving optimization and optimal control problems.However,it is often computationally untenable to run true dynamic programming due to the backward numerical process required for its solution,i.e.,the well-known "curse of dimensionality".Adaptive critic designs(ACD)is a novel nonlinear optimization and control algorithm based on actual data not with the exact mathematical model by integrating neural network, reinforcement learning and dynamic programming(DP).It provides an effective way to overcome curse of dimensionality.This dissertation introduces two novel nonlinear optimal neuro-controllers which are based on adaptive critic design and use the structure of heuristic dynamic programming(HDP)and action-dependant adaptive critic designs(ADACD).The HDP and ADACD neuro-controllers have features such as a simple frame-work,etc., moreover,they can bring the online study into effect.The simulation results show that the HDP and ADACD neuro-controllers are possible to keep the temperature of sintering zone stability in a certain range,and the temperature can meet the requirements of cement clinker production.Simulation results also show that neuro-controller with the ACD have the potential to control the cement rotary kiln.