Keywords
anaerobic digestion, artificial neural networks, data mining, diagnosis, foaming
Start Date
1-7-2008 12:00 AM
Abstract
Activated sludge processes are complex biological systems in which organic matter and nutrients (nitrogen and phosphorous) are removed from wastewater. One of the most common alternatives for the treatment of waste from activated sludge systems is Anaerobic Digestion (AD). AD is even a more complex biological process. One of the most important problems which can appear in anaerobic digesters is foaming. In the literature there is not a big agreement on the foaming causes. Therefore, the aim of the paper is to apply a methodology based on artificial neural networks to determine the most relevant variables for foaming diagnosis in anaerobic digestion. To perform the study real data from a pilot plant located in the LBE in Narbonne were used. Results show inflow rate, total organic carbon and carbon dioxide percentage among the relevant variables which are, according to the literature, some of the factors which influence the presence of foaming. This methodology can be valuable when selecting probes to monitor AD processes or to select the most relevant variables to diagnose the state of the process when foaming is involved.
A neural network approach for selecting the most relevant variables for foaming in anaerobic digestion
Activated sludge processes are complex biological systems in which organic matter and nutrients (nitrogen and phosphorous) are removed from wastewater. One of the most common alternatives for the treatment of waste from activated sludge systems is Anaerobic Digestion (AD). AD is even a more complex biological process. One of the most important problems which can appear in anaerobic digesters is foaming. In the literature there is not a big agreement on the foaming causes. Therefore, the aim of the paper is to apply a methodology based on artificial neural networks to determine the most relevant variables for foaming diagnosis in anaerobic digestion. To perform the study real data from a pilot plant located in the LBE in Narbonne were used. Results show inflow rate, total organic carbon and carbon dioxide percentage among the relevant variables which are, according to the literature, some of the factors which influence the presence of foaming. This methodology can be valuable when selecting probes to monitor AD processes or to select the most relevant variables to diagnose the state of the process when foaming is involved.
