Research On Stochastic Distribution Control Of Output Fiber Length Distribution Of Pulp Refining Process Based On Iterative Learning Control

In the process of chemical mechanical pulping,the pulp refining process is a very important production unit,its operating performance and output fiber morphology distribution directly affects the energy consumption and paper quality.Because of such as wood fiber rich raw materials in the pulp refining process after kneading disc extrusion and a series of complex process,causes the output fiber morphological distribution(such as fiber length distribution)with non-gauss distribution characteristics,so it will not be able to control the mean and variance of fiber morphology to accurately control the distribution of fiber morphology.Stochastic distribution control theory is an effective method to solve the output dynamic stochastic system with non-Gauss distribution,its core is the modeling and control of the probability density curve of system output fiber distribution.According to the actual engineering problems,supported by the National Natural Science Foundation of China(61333007)and the project of the National Natural Science Foundation of China(614730646),research on the control of the output fiber morphology of the pulp refining process based on the iterative learning control is carried out.The main work of this thesis is as follows:(1)In the aspect of modeling,this thesis compares the advantages and disadvantages of the Gauss basis function and the B spline basis function in the model approximation,then we choose the Gauss basis function to approximate the probability density curve of the distribution of the output fiber and decouple weights.Based on the root mean square Gauss basis function model of fiber distribution,that is,represent the square root of the probability density curve of the fiber morphology distribution by the Gauss basis functions and its weights.And the control of the distribution probability density curve of the output fiber morphology is transformed into the control of the weights of the Gauss basis functions.(2)In the aspect of control,in order to achieve the goal of controlling the fiber morphology of the probability density distribution curve,on the one hand,the perfect approximation of the target curve and the weight decoupling of the selected Gauss basis functions are needed,on the other hand,should guarantee the decoupling of the weights without error tracking.Therefore,the double closed loop iterative control algorithm based on the output fiber morphology distribution of the grinding process is studied.The inner loop is based on iterative learning control method to adjust parameters of Gauss basis function model,the outer loop uses the Gauss basis function model obtained by the inner loop to decouple the fiber morphological distribution probability density curve,then uses subspace identification,the state space model of the main input variables and the output weights of the pulp refining process is obtained.Using outer loop iterative learning control algorithm to control the output weights,so as to control the output of fiber distribution probability density curve.Finally,the effectiveness of the control algorithm is verified though simulation results.(3)In order to improve the fast convergence of the iterative learning control algorithm and its applicability,this thesis do not based on the modeling of the input variables and the output weights,uses the prior knowledge of the input variables and the error of output weights,proposes an iterative learning control algorithm based on geometric analysis,which improves the whole double closed loop iterative learning control algorithm.Finally,the effectiveness of the control algorithm is verified though simulation results.