Intelligent Detection Method Of Bulking Sludge Based On A Self-organizing Recurrent Radial Basis Function Neural Network

Sludge bulking is always the tough problem in the safe operation of the urban wastewater treatment system.This urban wastewater treatment system uses the activated sludge process.However,sludge bulking has high incidence rate and the wide coverage.When sludge bulking happens,it will lead to not up to effluent standard,and even result in the failure of the whole process.In order to avoid the sludge bulking,an intelligent diagnosis method is desiderated to effectively identify the incidence of sludge bulking and guaranteeing the normal operation for wastewater treatment process.In this paper,according to the mechanism of sludge bulking and partial least squares algorithm,the correlation variables of SVI are obtained.A self-organizing recurrent RBF neural network(IOA-RRBFNN)based on information-oriented algorithm(IOA)is proposed.The SVI soft-computing model,based on IOA-RRBFNN,could predict sludge volume index(SVI)online.Meanwhile,the cause variable identifying(CVI)algorithm is developed.So the intelligent detecting system for sludge bulking can be complished.The works in the paper are as follows:1.The selection of variables correlated SVI.The PLS algorithm is proposed to extract the correlation variables.Firstly,analysing the main influencing factors of the sludge setting,the parameters associated with SVI are extracted.Then,by the running data of wastewater process,correlation parameters(MLSS?DO?COD?T?BOD? TN)are extracted by PLS method.2.The design of the IOA-RRBFNN.An IOA-RRBFNN with capability of nonlinear processing is desinged.The independent contribution degree index from hidden neurons to output as well as the concept of information processing degree of hidden neurons is put forward.Through the theory of information processing degree of hidden neurons,the self-organizing mechanism is proposed.And the mechanism realizes the optimum structure of recurrent RBFNN.Moreover,an improved(Levenberg-Marquardt,LM)algorithm has been proposed to adjust the parameters of IOA-RRBFNN,which realized the fast convergence.Two nonlinear system modeling experiments reveals that the proposed IOA-RRBFNN has a more compact structure and more highly accurate prediction.3.The SVI soft-computing model study based on IOA-RRBFNN.In view of obtaining the SVI values real-time,the correlation variables of the SVI as input of the IOA-RRBFNN proposed is applied to SVI soft-computing model design.The design of SVI soft-computing model is applied to actual wastewater treatment process preparation platform,the simulation results show that the SVI soft-computing model complemented high prediction accuracyof SVI.4.The design of intelligent detecting method for sludge bulking.For realizing the effective identification of sludge bulking,a CVI is put forward for intelligent diagnosis method to sludge bulking.Firstly,through the comparison of residual error of the actual values of wastewater treatment plant and the predicted SVI values of IOA-RRBFNN neural network,sludge bulking is judged.Secondly,the CVI algorithm identify fault variables of the sludge bulking.The intelligent detectiong method has provided support to identify fault variables,thus effectively reduce the incidence rate of sludge bulking.