Soft-computing Model For Sludge Volume Index Based On Spiking Self-organizing Recurrent RBF Neural Network

Sludge bulking, one of the bottleneck problems that restricts the development of wastewater treatment plants(WWTPs). Since sludge bulking has the features of complicated mechanism, numerous inducing factors and the interactions among them, it is hard to establish a precise mathematical model. The sludge volume index(SVI), one of key indexes to quantify sludge bulking, represents the settling performance of activated sludge, and has been widely used to describe the degree of sludge bulking. To achieving the SVI measurement on-line, a SVI soft-computing model based on spiking SR-RBF neural network is proposed, and the real-time SVI values are obtained, the sludge bulking can be recognized.The main contents of the study are as follows:1. According to mechanism analysis and history data of sludge bulking, a set of secondary variables is chosen for the SVI soft-computing model. The secondary variables selection is the key step in SVI soft-computing model. Future study of sludge bulking concept and the main factors, analysis of sludge bulking mechanism model, the 12 influent factors are employed to describe SVI. Meanwhile, using PCA method to analyzed the 12 influent factors, and finally 6 parameters are selected, including Qin、BOD、COD、DO、pH、TN.2. Since it is difficult to online adjust the structure of recurrent RBF neural network, a spiking self-organizing recurrent RBF neural network is designed. According to the brains’ information transfer mode and biological spiking neurons’ model, a growing and pruning mechanism is designed, which adjusts the structure of the recurrent RBF neural networks. Simultaneously, a self-organizing gradient descent algorithm is proposed to training the SR-RBF neural network parameters and improves the recurrent RBF neural network’s performance. The nonlinear system modeling simulation results show that the proposed self-organizing mechanism can online adjust the structure of recurrent RBF neural network, and a more accurate prediction has been got.3. A kind of SVI soft-computing model that based on spiking SR-RBF neural network is established to solve the SVI online measure problem. In order to realize SVI’s online measurement, the spiking-based self-organizing RBF neural network is applied to the designed SVI soft-computing model. As the structure of spiking-based self-organizing RBF neural network can be adjusted on-line, and all the parameters are trained by second order LM algorithm, the proposed SVI soft-computing model got a high precision, and a faster convergence speed, simulation results have verified the effectiveness of the proposed model.4. A SVI soft-computing platform is developed. The platform mainly includes the database module, login module, data processing module, model training and simulation module and results query module. Firstly, using Visual Studio 2010 software designed the interface, and it provides the users for network model selection, parameters initialization, and modeling results query. Then, the Matlab and Mysql softwares are used to write the background program, and realized the soft-computing model calculation and the users’ information management. Finally, through the information transmission among users’ information management module, data processing module, neural network selection module and so on, realized the SVI prediction and display, achieved the purpose of sludge bulking recognition visualization.