The invention provides an online quantitative recognition analysis method for a loose fault of a bearing support of a rotor system. The method comprises the following steps: 1) establishing a bolt model with spiral threads; 2) acquiring a change rule of load distribution between junction surfaces of screw connection following the loose amount; 3) acquiring the change rule of normal contact stiffness and damp of the junction surfaces of screw connection following the loose amount; 4) acquiring the change rule of equivalent stiffness and damp at the loose position following the loose amount; 5) establishing a dynamical model taking the loose bearing support as an object, thereby acquiring a relation between the loose amount of the bearing support and the oil film pressure endured by the bearing support; 6) acquiring the relation between the thickness distribution of a lubricating film and the loose amount; 7) monitoring the change rule of the thickness distribution of the lubricating film in a sliding bearing online in real time; 8) quantitatively recognizing and analyzing a loose fault signal of the bearing support online. According to the method provided by the invention, a dynamical model establishing method and a fault carrier signal with a shorter route and few influence factors are supplied for the loose fault of the bearing support.