A comprehensive reliability allocation method, which is applicable to the cases of lack of data in design stage, as well as the complexity of the influence factors and uncertainty of expert evaluation, is presented based on the grey system theory. The key of this method is to modify the weights of influence factors which are associated with expert scoring data by grey correlation analysis before reliability indices are assigned linearly to each subsystem.Based on grey evaluation with 4 levels, the application of this method is illustrated with the reliability allocation of a rotary machine under the consideration that complexity, technical level, importance and environment condition are chosen as influence factors. Compared with the actual faults statistics results, the validity and feasibility of the method is verified.