Traditional probability statistics theory is impossible to obtain failure lifetime data by accelerated test for expensive and complex systems or equipments for real-time work. Due to the variety of system failure modes and the randomness of system deterioration process and the fuzziness of system maintenance threshold, it is difficult to estimate the random deterioration process of a complex repairable system by single parameter. In order to describe system performance deterioration more subjectly, it proposes generalized proportional intensity model(GPIM). considering the effects of various covariates such as performance parameters, environment stress, failure types and maintenance history simultaneously. This method provides a new method to solve the maintenance decision-making problem of complex repairable system. CF6-80C2A5 aero-engine is illustrated as an example for case study to indicate the obvious practical value by the method proposed herein.