An orthotopic linear programming filtering based fault diagnosis method is proposed for the linear systems with unknown but bounded noises. The linear programming equation is used to show the constraints in the recursive operation process and the space expression of the orthotopic is updated during the recursive computation. Each parameter has its own uncertainty interval and the feasible parameter set is described by the orthotopic, with checking whether the filter of orthotopic feasible parameter set is empty or not to finish fault detection. Set membership filters are designed according to different faults, when the orthotopic feasible parameter set is empty, which means as the fault occurred, the model is matched to achieve fault identification. If a fault type is not included in the fault sample, the type of fault is added to the fault sample base. The low dimension and high dimension examples are given to illustrate the performance of the orthotopices and to analysis the space expressions under different fault situations. The simulation results and analysis verify the effectiveness and practicability of the proposed fault diagnosis method.