Research And Simulation Of Control Strategy For Fuzzy Neural Network Based On Improved FCM

With the development of science and technology,the structural complexity of the control system is continuously increasing,it is difficult to establish an accurate mathematical model of the controlled object,which greatly limits the overall control effect of the system.Compared to other control strategy construction methods,the fuzzy system does not require the precise mathematical model of the controlled object,combines process information with function approximation,and adopts easy-to-understand language rules,providing a rule-based systematic control strategy implementation framework.Therefore,the fuzzy system is widely used in the control problems of nonlinear,time-varying and pure-lag systems.However,the fuzzy system itself lacks the ability to extract rules and learn knowledge.Combined with other intelligent technologies,the hybrid fuzzy model with the complementary expression ability and learning ability has become one of the research hotspots in this field.In fuzzy system construction,the fuzzy system and neural network are used in the selection of membership function parameters and extraction of knowledge rules.Fuzzy C-means(FCM)clustering algorithm is introduced into fuzzy neural network and FCM clustering algorithm is used.The features of the membership function are extracted to obtain the parameters of the membership function,the number of fuzzy subsets,and the number of rules,and the initial structural model of the fuzzy neural network is obtained.After that,the model parameters are optimized through the learning algorithm,the final system model is obtained,and the system model is constructed.Simulation verification was performed.The main work of the dissertation is as follows:According to the fuzzy C-means(FCM)clustering algorithm randomly selecting the initial clustering center and the number of clusters,it is vulnerable to isolated points and affects the clustering effect.This paper proposes an initialization method to optimize the initial clustering center and the number of clusters,and uses the improved Mahalanobis distance to perform distance calculations.The covariance of each clustering result is used to estimate the sample covariance,and when the covariance matrix is irreversible,it is replaced with a pseudo-inverse matrix to achieve a more efficient representation of the distance between data.Compared with the traditional algorithm using Euclidean distance operator,the proposed method avoids the repeated calculation of related attributes and the influence of different attribute dimensions on the clustering results.Finally,the improved FCM algorithm is verified by UCI data,and the validity and reliability of the algorithm are verified.For the problem of parameter setting based on empirical or random values in the initial model of fuzzy neural network,this paper proposes that the improved FCM clustering algorithm can be used to set the parameters of the initial model of the fuzzy neural network,and an efficient initial model structure parameter can be obtained.The learning algorithm is used to optimize the parameters,and the final system model is obtained based on the optimized structural parameters.Based on the improved FCM fuzzy neural network,simulations were performed in time series prediction,typical second-order plus delay systems and PMSM position control systems.In the simulation of time series prediction,the effects of conventional fuzzy,conventional neural network and improved fuzzy neural network are compared and analyzed.In the typical second-order plus-lag system simulation,the comparison with PID,conventional fuzzy,fuzzy PID and conventional fuzzy neural network is made.In the simulation of the PMSM position control system,the PI algorithm is used and compared with the control effect.In the PMSM motor position control,the laboratory's existing experimental equipment is used to verify the effectiveness and performance of the algorithm's actual application.In this way,the verification from the algorithm itself to its application in the typical control system is completed,and finally the application of the actual control system of the motor is verified.The simulation and experimental results show that the improved fuzzy neural network algorithm is superior to other algorithms.