作者机构:
[樊绍胜] College of Electric and Information Eng., Changsha University of Sciences and Technology, Changsha 410077, China;[王耀南] College of Electric and Information Eng., Hunan University, Changsha 410082, China
通讯机构:
[Fan, S.-S.] C;College of Electric and Information Eng., Changsha University of Sciences and Technology, China
会议名称:
Machine Learning and Cybernetics, 2004. Proceedings of 2004 International Conference on
关键词:
online measurement of viscosity;rubber mixing process;fuzzy-GA modeling;similarity assessing
摘要:
Rubber mixing is a complicated process and online measurement of viscosity is very difficult to achieve. To cope with the problem, a soft sensing approach based on fuzzy-GA modeling is proposed. During modeling, T-S fuzzy model is employed to approximate the non-linearity of rubber mixing process, an improved Gustafon-Kessel fuzzy clustering algorithm based on similarity assessing is proposed to determine the optimum number of clusters and real-coded GA (genetic algorithm) is adopted to optimize model parameters. All these techniques make the fuzzy model simple and accurate. Based on the approach, a test is conducted. The results show that the proposed approach provides a result near laboratory measurement, and the error is lower and acceptable. It decreases the time involved with tests in laboratory and can be seen as a powerful tool for online measurement of viscosity.
摘要:
A neural network based predictive control strategy for active power filter is presented in this paper. In the strategy, RBF neural network is employed to predict future harmonic compensating current. In order to make the predictive model much simpler and tighter, an adaptive learning algorithm for RBF network is proposed. Based on the model output, Genetic algorithm is introduced to optimize objective function, which generates proper value of control vector. The neural network based predictive algorithm is used in internal model control scheme to compensate for process disturbances, measurement noise and modeling errors. Simulation test under various conditions is implemented. The results show the neural network based predictive control is more effective and feasible than PI control or digit adaptive control.