AI-Driven Framework for Enhancing Water Quality Engineering Experimentation

• Closed-loop AI feedback accelerates error correction and boosts metacognitive skills. • SHAP-optimized extreme trees model achieves 97.9% prediction accuracy. • AI-powered teaching increase active exploration time and innovation capability. • AI-driven pedagogical reconfiguration improves curriculum design efficiency. • AI enhances real-time data processing pipeline and reduces operational errors. Artificial Intelligence (AI) holds transformative potential for cultivating high-quality talent in higher education, particularly in engineering experimental pedagogy. However, traditional water quality engineering courses face challenges in dynamic responsiveness, personalized learning, and data-intensive experimental workflows. This study addresses these gaps by establishing an integrated AI-enhanced framework for water quality engineering experimental courses. A mixed-methods approach was employed across eight laboratory modules. The framework combined project-based learning, Bloom’s taxonomy, and AI-driven tools (SHAP-optimized Extreme Trees/Random Forest/Decision Tree algorithms, NLP/image recognition, closed-loop feedback systems). Undergraduate cohorts using AI-integrated methods (n=64) were compared with conventional method (CM) cohorts (n=67). Metrics included technical proficiency, data accuracy, innovation capability, engagement, error rates, and teacher workload. AI integration significantly enhanced pedagogical efficacy, evidenced by a 41% increase in experimental design efficiency alongside 33.8% (p<0.01) and 35.6% (p<0.01) improvements in hypothesis formulation and data interpretation accuracy, respectively. Data processing accuracy exceeded 85%, accompanied by a 57.9% reduction in processing time, while innovation capability rose by 28.3% (p<0.01) and operational errors decreased by 40%. Concurrently, student engagement increased by 62.5% (p<0.05) with significant metacognitive skill gains (Cohen’s d = 0.72), and teacher workload declined by 37.1%, freeing 3.1 weekly hours per instructor. These outcomes were driven primarily by real-time closed-loop feedback and personalized learning pathways. This study provides a replicable “AI + Experimental Courses” paradigm that synergizes human expertise with AI capabilities to overcome data robustness and emotional intelligence challenges. It advances sustainable AI-education integration, offering a scalable model for engineering education reform aligned with sustainable development goals.