The needs of software maintenance and the importance of maintenance project management increase rapidly in the military avionics industry. Although few previous studies related to the maintenance effort estimation were proposed, they had two drawbacks:(1) the needs of some input parameters which are hard to measure precisely at early stage, (2) the passive support of “what-if” analysis by exploring the impact of changes among input parameters for more realistic estimation. In this paper, we suggest probabilistic effort estimation model for military avionics software maintenance by considering structural characteristics. We elicited the four aspects to the maintenance effort including structural characteristics, developed probabilistic maintenance effort estimation model based on Bayesian network with a data set of 76 military avionics software maintenance projects in the Republic Of Korea AirForce(ROKAF).