Automatic detection of exogenous respiration end-point using artificial neural network

Area of Science:

• Environmental microbiology
• Biotechnology
• Analytical chemistry

Background:

• Aerobic bacteria respiration shifts from endogenous to exogenous upon introduction of biodegradable material (e.g., wastewater).
• Respirometry records these shifts in a respirogram, enabling expert identification of biodegradation phases.
• The exogenous respiration phase quantifies easily biodegradable material (short-term BOD or BOD(ST)), essential for wastewater treatability assessment with COD.

Purpose of the Study:

• To develop an automated method for detecting the end-point of exogenous respiration in respirograms.
• To enable on-line monitoring of wastewater treatability through accurate end-point detection.
• To investigate the efficacy of neural networks for this automated detection task.

Main Methods:

• Utilized respirometry to generate respirograms reflecting bacterial respiration changes.
• Trained a neural network model to identify the end-point of the exogenous respiration phase.
• Evaluated the neural network's performance on a dataset of respirograms.

Main Results:

• The trained neural network successfully detected the correct end-point of exogenous respiration in most cases studied.
• Demonstrated the potential for automated, on-line detection of wastewater treatability parameters.
• Achieved promising results in identifying critical transition points in bacterial respiration.

Conclusions:

• Neural networks show significant promise for automating the detection of exogenous respiration end-points.
• This automated approach is a key step towards real-time wastewater treatability monitoring.
• The findings support the integration of machine learning in environmental monitoring applications.