Integrating the advantages of the variable precision rough set(VPRS) and optimization method based extreme learning machine(OMELM), a method of variable precision rough set-OMELM is proposed to diagnose the failure of the tiltrotor aircraft flight control system. An attribute reduction algorithm is presented according to data collection characteristics of fault output signals. The attribute reduction algorithm defines a variable precision rough entropy. The attribute importance is defined from the perspective of mutual information increment in information theory. The OMELM classifier is constructed to classify the reduction attribute feature. Finally, simulation is carried out based on XV-15, and the results show the high average recognition rate and the short diagnosis time of the proposed method.