Combined with the actual operation data of a domestic civil aircraft and the simulation and validation data of aircraft performance software, the landing distance was compared and analyzed to provide quantitative support for the improvement of the aircraft"s operation capability. The key conditions Required for performance simulation calculation are extracted from the actual operation data to obtain the Required Landing Distance (RLD) and distance difference under the same conditions. The landing-related data is discretized by the quartile method, and the modeling methods are compared and selected. Finally, the network structure is learned by PC algorithm. Bayesian estimate learning network parameters to construct the risk model of actual landing distance exceeding RLD. The posterior probability and maximum posterior hypothesis of exceeding RLD state are obtained by using Bayesian network accurate reasoning. The results show that the probability of exceeding RLD can be effectively reduced by flattening to ground time of about 7s and grounding to ground speed of 40 knots between 18-24s, and some suggestions are put forward to reduce the risk of exceeding RLD by combining relevant parameters.