To forecast the fault and carry out condition-based maintenance for weapon system, the prognostic method based on Genetic Neural Network (GNN) is studied. The genetic algorithm is improved by adopting real coding, adaptive crossover rate and mutation rate, also the learning algorithm of neural network's weight is ameliorated with the improved genetic algorithm, and the genetic neural network is obtained. The genetic neural network is trained by the detected data of equipments, and then is used to predict the degenerating trend of the characteristic parameters of the equipments. The predicting example shows that the use of the improved neural networks can achieve fault prediction before the time point of faults respectively, and the predicting accuracy and the predicting performance of the genetic neural networks are greatly improved compared with those of the basic neural network, which can enhance the supporting capability of the weapon equipment and realize condition-based maintenance.