A fluid filter abnormality detecting method is provided. The method includes considering the geometry of the fluid filter, the physical properties of the fluid, the porosity of the fluid filter, the particulate density, the fluid flow and the pressure difference to obtain an operation model of the fluid filter, and obtaining an initial particulate quantity through the operation model. Furthermore, in the present disclosure, a kalman detector is used for estimating the particulate quantity at an estimating time, and the particulate quantity is compared with a pre-determined value to determine the abnormality of the fluid filter. Thereby, a high accuracy can be achieved through the method of the present disclosure.