Research And Application Of The Rough Fuzzy Neural Networks Inverse Based On "Assumed Inherent Sensor"

Soft Sensor Technique is an effective approach which is used to solve some actual puzzles,caused by the unavailability of measures in the controlling process variables measuring or the high price of measures,in industrial process control systems.Neural network inverse theory integrates neural network approximation ability,self-learning ability and fault tolerance characteristics of nonlinear systems with inverse system method. Through the use of neural network approximation of the inverse system with original system,used for dynamic measurements,inverse system method will be on the theoretical study and the dynamic characteristics of the sensor project application will have a positive far-reaching significant.Rough set theory is a theoretical method which is applied to study the expression,learning and inducing of incomplete and uncertain data. It can reduce the knowledge-expression space;cancel the redundant information via describing the importance of different attribute in knowledge expression without prior knowledge.In this paper down,"Soft-sensing technology——Neural Network inverse theory——Rough Set Theory——The theory of rough fuzzy neural network inverse method proposed will be rough fuzzy neural network inverse method is applied to soft sensor in" line of thinking on the subject of the author at the Institute to do the work a detailed introduction.The use of rough set theory in which the attribute reduction and rule extraction methods to build neural network inverse paste is the focus of this thesis researchIn this paper,as an example of the measured data of erythromycin fermentation process.The input sample data are processed firstly of the rough fuzzy neural network inverse.Article separately detailed description of carried out the error data on the sample treatment, normalized,and through Matlab simulation software,such as treatment results after the simulation,to illustrate the effectiveness of the method.Finally,after a rough set theory to fuzzy after the fuzzy data set attribute reduction and rule extraction to construction of the rough fuzzy neural network approximation "includes sensors Inverse",so as to achieve on difficult-line measurement data soft sensor of the important parameters of mycelium concentration,total sugar concentration,and the product of the concentration in the erythromycin fermentation process. Receive the prediction results with the same soft-sensing data of ANN inverse prediction results were compared.The simulation results showed that rough neural network in prediction accuracy and generalization ability is on the more improvement,basically meet the needs of a further optimize the fermentation process control.The results show that the modeling method in the prediction accuracy and generalization ability on must have superiority.