Artificial neural networks for modeling ammonia emissions released from sewage sludge composting

• Published in: Atmospheric environment (1994) 2012-09, Vol.57, p.49-54
• Main Authors: Boniecki, P., Dach, J., Pilarski, K., Piekarska-Boniecka, H.
• Format: Article
• Language: English
• Subjects: Ammonia; Ammonia emission; Analysis methods; Applied sciences; atmospheric chemistry; Atmospheric pollution; Bark; Carbon; carbon nitrogen ratio; Composting; correlation; Dispersed sources and other; Emission analysis; emissions; Exact sciences and technology; ingredients; Mathematical models; Neural network modeling; neural networks; Pollution; Pollution sources. Measurement results; sawdust; Sewage sludge; Straw; trees
• ISSN: 1352-2310; 1873-2844
• DOI: 10.1016/j.atmosenv.2012.04.036

The project was designed to develop, test and validate an original Neural Model describing ammonia emissions generated in composting sewage sludge. The composting mix was to include the addition of such selected structural ingredients as cereal straw, sawdust and tree bark. All created neural models contain 7 input variables (chemical and physical parameters of composting) and 1 output (ammonia emission). The α data file was subdivided into three subfiles: the learning file (ZU) containing 330 cases, the validation file (ZW) containing 110 cases and the test file (ZT) containing 110 cases. The standard deviation ratios (for all 4 created networks) ranged from 0.193 to 0.218. For all of the selected models, the correlation coefficient reached the high values of 0.972–0.981. The results show that he predictive neural model describing ammonia emissions from composted sewage sludge is well suited for assessing such emissions. The sensitivity analysis of the model for the input of variables of the process in question has shown that the key parameters describing ammonia emissions released in composting sewage sludge are pH and the carbon to nitrogen ratio (C:N). ► The research on ammonia emission from the composted sewage sludge have been made. ► The prediction neural models’ file has been built. ► The optimal models for prediction of ammonia emission were tested and verified. ► It has been stated that pH was a key factor for ammonia emission from composting.