- Thesis topic: Machine Learning Prediction of Membrane Bioreactor Effluent Quality with Explainable Insights from Microscopic Image Analysis

- Doi: 

- Abstract: 

        The application of artificial intelligence in wastewater treatment field has shown potential in enhancing monitoring and prediction capabilities. This study investigates the integration of AI-driven microscopic image analysis for predicting effluent in lab-scale membrane bioreactors The objective is to enhance decision-making and process optimization in sludge management through the use of advance deep learning techniques. A convolutional neural network-based object detection model, YOLOv8, was employed to extract features from microscopic images of sludge samples. The initial object detection model achieved a mean average precision at IoU threshold 0.5 (mAP50) of 75% with precision of 72% indicating promising potential for accurate microorganism identification. Three predictive scenarios and four machine learning models (linear regression, random forest, XGB, and ExtraTrees) were applied to predict key effluent quality indicators such as biochemical oxygen demand, chemical oxygen demand, ammonia nitrogen, and total phosphorus. The scenarios include operation parameter such as pH, dissolved oxygen, alkalinity, mixed liquor suspended solid, sludge volume index, and influent characteristic. Operational parameters generally provided best prediction accuracy, while combined parameters showed improved performance in predicting certain parameters. The findings demonstrate the potential of AI can serve as a valuable tool for optimizing efficiency of monitoring wastewater treatment performance by providing more accessible and automated approach.

Keywords: artificial intelligence, membrane bioreactor, microscopic data, wastewater treatment.