Polyhedron spatial feasible set filting based fault diagnosis is proposed for switched systems with unknown noise term. Firstly, using the regularized spatial geometric structure of the zonotope to ananlyze the intersection of the strip and the strip that obtained by disassembling the polyhedron spatial, the parameter feasible set in the non-switching state is obtained by using the smallest feasible set of zonotope to wrap the parameter feasible set. Then, whether the system parameter is changed by dectecting the compatibility of the polyhedron spatial and the strip space. The parameter changes are divided into switching and faults for analysis, and then the matching consistency of the strip space and the sub-model is analyzed. The filter is designed by matching the fault model to achieve fault diagnosis. Finally, the simulation results and analysis verify the effectiveness and feasibility of the proposed fault diagnosis method.