Based on the characteristic analysis of existing drag models for different particle concentration segments, three traditional drag models, which are adaptive for specific concentration ranges, are chosen for the drag force reduction caused by particle agglomeration. The smooth function is used to connect the drag models, then a modified drag model is proposed. Coupling with the Eulerian-Eulerian model, the modified drag model is applied for simulation of 2D bubbling fluidized bed. The result shows that the modified model has better prediction of partial pressure drop than Gidaspow and Syamlal-O'Brien models. Along with the increase of superficial gas velocity, the modified drag model can predict more accurately bed expansion. When Ug=0.46 m/s, the modified model leads to a better simulation result of particle concentration than Syamlal-O'Brien model apparently.