Aim The RFB (radial hats function) netal network was studied for the model indentificaiton of an ozonation/BAC system. Methods The optimal ozone's dosage and the remain time in carbon tower were analyzed to build...Aim The RFB (radial hats function) netal network was studied for the model indentificaiton of an ozonation/BAC system. Methods The optimal ozone's dosage and the remain time in carbon tower were analyzed to build the neural network model by which the expected outflow CODM can be acquired under the inflow CODM condition. Results The improved self-organized learning algorithm can assign the centers into appropriate places , and the RBF network's outputs at the sample points fit the experimental data very well. Conclusion The model of ozonation /BAC system based on the RBF network am describe the relationshipamong various factors correctly, a new prouding approach tO the wate purification process is provided.展开更多
Based on 1,3-propanediol production from batch fermentation of glycerol by Klebsiella pneurnoniae, a multistage dynamic system and its parameter identification are discussed in this paper. The batch fermentation proce...Based on 1,3-propanediol production from batch fermentation of glycerol by Klebsiella pneurnoniae, a multistage dynamic system and its parameter identification are discussed in this paper. The batch fermentation process is divided into three stages exhibiting different dynamic behaviors and characteristics, from which a corresponding nonlinear multistage dynamic system is built. We then propose a parameter identification optimization model whose objective function is the average relative error. The model is solved by particle swarm optimization weighted by inertia, and the result shows that the relative error of our proposed model is 2-10%smaller than those of existing models.展开更多
文摘Aim The RFB (radial hats function) netal network was studied for the model indentificaiton of an ozonation/BAC system. Methods The optimal ozone's dosage and the remain time in carbon tower were analyzed to build the neural network model by which the expected outflow CODM can be acquired under the inflow CODM condition. Results The improved self-organized learning algorithm can assign the centers into appropriate places , and the RBF network's outputs at the sample points fit the experimental data very well. Conclusion The model of ozonation /BAC system based on the RBF network am describe the relationshipamong various factors correctly, a new prouding approach tO the wate purification process is provided.
基金Acknowledgments This work was supported by the National Natural Science Foundation of China （Grant No. 10871033）, ＂863＂ Program （No. 2007AA02Z208） and ＂973＂ Program （No. 2007CB71430c）.
文摘Based on 1,3-propanediol production from batch fermentation of glycerol by Klebsiella pneurnoniae, a multistage dynamic system and its parameter identification are discussed in this paper. The batch fermentation process is divided into three stages exhibiting different dynamic behaviors and characteristics, from which a corresponding nonlinear multistage dynamic system is built. We then propose a parameter identification optimization model whose objective function is the average relative error. The model is solved by particle swarm optimization weighted by inertia, and the result shows that the relative error of our proposed model is 2-10%smaller than those of existing models.
