Design Of Industry Steady-State Optimization Based On CMAC

It has become the focus of people to make the complicated industrial engineering operated steadily,safely,high quality and high-effectively,along with the fact that industry process run increasingly to large-scale and automation.People do not only satisfy the steady operation of production device,but also set higher and higher request to enhance the production efficiency , enhance the quality and production as well as reduce the consumption.It is important guarantee to obtain the economic efficiency in process system to steady state optimization of industrial process.It is precise mathematical model that tradition stable state optimization method based on the model.But many process objects present the non-linearity greatly and the intrinsic mechanism is extremely complex,which makes it is very difficult to promulgate its inherent laws from the mechanism directly.Moreover,along with the enhancement of complex degree of industrial process,it is also very difficult to establish the process system of precise mathematical model.Therefore exploring a new modeling and controlling method,which does not need to have the thorough understanding of the process,becomes more and more important.The present paper has studied the non-linear modeling method based on the neural network,as well as the optimization techniques based on the neural network modeling foundation.The optimization algorithm that is mainly studied in this paper is particle swarm optimization,and the model is cerebellar model articulation controller.These methods can solve the stable state optimization problem in industrial process.The concrete work has been divided into two major parts,the modeling and the optimization. In modeling part,aiming at some characteristics in the modern complex industrial system,such as the multivariable,the non-linearity,the close coupling,the time-variable and time lag,and uncertainty,this paper proposes to use the modern intelligent modeling method - artificial neural networks to establish the system model . The artificial neural networks not only may approach the non-linearity willfully , but also have massively parallel processing,the knowledge distribution saving strongly,self-learning ability greatly,and the fault tolerance well and so on.The CMAC neural network is one kind of typical partial neural network,having the optimal approximation ability,but at present this network is mainly used in the control aspect,this paper uses the network to the optimized aspect.At the same time,aim to the contradiction between storage space and generalization precision,this paper proposed a fuzzy CMAC network that effectively overcome the disadvantages in original model.The simulation test has confirmed its high accuracy and fast convergence.In optimized part,in order to overcome the disadvantages that can fall into local optimization solution and the application scope is not very widespread,uses the GA algorithm to optimize its parameters,and at the search later implements random perturbation . Enables the algorithm to choices its parameters according to practical problems , and can avoiding local optimization.The simulation test has confirmed its high accuracy and fast convergence by standard testing function.Finally takes the hydrogen peroxide cumin(CHP) decomposition unit as the optimized object,gathered and analysis massive field data.Aim to the product installment movement situation determining the controlled variables and the optimized goal.Obtained satisfying effect by using the fuzzy CMAC neural network and the improved particle swarm optimization.