Optimising Venturi flume oxygen transfer efficiency using uncertainty-aware decision trees.

Area of Science:

• Environmental Engineering
• Fluid Mechanics
• Water Treatment

Background:

• Optimizing oxygen transfer is crucial for wastewater treatment and aquaculture.
• Venturi flumes are commonly used for flow measurement and aeration.
• Predictive modeling can enhance the understanding and efficiency of these systems.

Purpose of the Study:

• To optimize standard oxygen transfer efficiency (SOTE) in Venturi flumes.
• To investigate the impact of parameters like discharge per unit width (q), throat width (W), and gauge readings (H).
• To compare the performance of various machine learning models for SOTE prediction.

Main Methods:

• Analysis of a comprehensive experimental dataset using multiple linear regression (MLR), multiple nonlinear regression (MNLR), gradient boosting machine (GBM), extreme gradient boosting (XRT), random forest (RF), M5 (pruned and unpruned), random tree (RT), and reduced error pruning (REP).
• Model performance evaluation using correlation coefficient (CC), root mean square error (RMSE), and mean absolute error (MAE).
• Uncertainty analysis, one-way analysis of variance, sensitivity, correlation, and SHapley Additive exPlanations (SHAP) analyses.

Main Results:

• M5_Unprun model demonstrated the highest performance with CC=0.9455, RMSE=0.1918, and MAE=0.0030.
• GBM model also showed strong performance (CC=0.9372, RMSE=0.2067, MAE=0.0006).
• Throat width (W) and gauge readings (H) were identified as the most influential factors affecting SOTE.

Conclusions:

• M5_Unprun and GBM models provide reliable and robust predictions for SOTE in Venturi flumes.
• The study highlights the importance of specific geometric and operational parameters for optimizing oxygen transfer.
• Advanced machine learning techniques offer significant improvements over traditional methods for predicting SOTE.