AI Insight
This study demonstrates that tungsten trioxide (WO3) photocatalyst can degrade spiramycin antibiotic in water with 87.6% efficiency under optimal conditions (pH 5.0, WO3 concentration 0.125g/l, spiramycin concentration 0.15g/l). The researchers used an XGBoost machine learning model to analyze 45 experimental data points and identify optimal parameters, achieving high prediction accuracy (R2 = 0.983-0.992). The degradation mechanism primarily involves hydroxyl radicals, followed by photoinduced holes and hydrogen peroxide.
Why it matters
This research addresses the growing environmental problem of antibiotic contamination in water systems, which threatens human and ecological health. The combination of photocatalytic treatment with machine learning optimization offers a more efficient, resource-saving approach to removing emerging pollutants from wastewater compared to traditional trial-and-error methods.
Understand the Science
by Hayat Khan
The intensification of worldwide urbanization and industrial growth has led to a rise in the release of toxic substances into the environment. Antibiotics are widely used as human medicine and animal husbandry, for example annual global consumption (106 kg) of amoxicillin is 11.8, sulfamethoxazole is 2.0, tetracycline is 0.5 and spiramycin is 0.3 etc., therefore, they are often found in surface water and wastewater, thus possess a threat both to human and eco health. Photocatalytic technology can efficiently eliminate highly hazardous, low-concentration, and hard-to-treat contaminant, and tungsten trioxide (WO3) serves as a highly promising alternative photocatalyst. Evaluating the effectiveness of WO3 in photocatalytic decomposition through traditional techniques is resource intensive and complex. Consequently, in this study experimental results are applied for modeling and optimizing the operational parameters on photocatalytic decomposition of spiramycin, an emerging antibiotic contaminant using WO3 as photocatalyst. Three experimental parameters (photoreactor solution pH, WO3 concentration and spiramycin (SPR) initial concentration) were chosen to gather preliminary information. The influencing interaction effects and optimal parameters values were determined by using a machine learning technique of XGBoost employing a data set of 45 dataset points obtained under 3 parametric effects. The model achieved an R2 value of 0.983 and 0.992 on test and training data, respectively. SHAP (SHapley Additive exPlanations) analysis was utilized to elucidate the prediction outcomes, uncovering the importance and influence patterns of the input parameters. The degradation rate of SPR approached ~87.6% under optimal conditions of pH value of 5.0, WO3 concentration of 0.125g/l and pollutant concentration of 0.15 g/l, respectively. In addition, the applied ML method revealed that the most influencing parameter is the reactor solution pH value followed by pollutant concentration on the decomposition of SPR antibiotic. Moreover, we conducted experiments to investigate the role of oxidizing species. Our findings indicate that primarily hydroxyl radicals (OH•) followed by photoinduced holes (h+s) and hydrogen peroxide (H2O2), play significant roles in the breakdown of model contaminant. We also studied the reaction kinetics and proposed the photocatalytic mechanism based on the obtained results. As a study outcome, the XGBoost model exhibits great accuracy and strength, suggesting that machine learning holds considerable practical promise and worth in forecasting the degradation emerging pollutants via photocatalysts.