Machine learning and experimental study on the activity decrease of VW/Ti for SCR at ultra-high temperature: The influence mechanism and regulation strategy

Abstract

Selective catalytic reduction (SCR) with ammonia (NH₃) is an established method for NO_x elimination. However, the catalytic activity of conventional VW/Ti catalysts declines sharply at ultra-high temperatures (above 700 °C). To address this limitation, this study employed machine learning (ML) to optimize high-temperature SCR catalysts. A gradient boosting decision tree (GBDT) model with Shapley Additive Explanations (SHAP) identified calcination temperature, crystal phase, and surface acidity as the key factors influencing catalytic performance. Based on these insights, a designstrategy was developed for VW/Ti catalysts under ultra-high temperature conditions and successfully identified a catalyst exhibiting excellent resistance to high-temperature degradation through evaluation and characterization. Experimental validation confirmed that Si doping significantly improves catalytic activity and thermal stability, with the VW/5Ti1Si-800 catalyst demonstrating superior performance. This study highlights the effectiveness of combining ML and experimental approaches to mitigate high-temperature deactivation in SCR catalysts.