Abstract: Targeting the problems of low accuracy and long diagnosis time of conventional neural networks, a neural network fault diagnosis method based on improved genetic algorithm is proposed. By means of simplifying the genetic algorithm coding method, neural network weights and threshold values are acquired through decoding optimal fitness functions, thus a fault diagnosis model for bogie bearing based on neural network optimized by improved genetic algorithm is established. To verify the effectiveness of the model, simulation calculation and comparison are conducted with other diagnostic models. The results indicate that the neural network bogie bearing fault diagnosis model based on the improved genetic algorithm is timesaving and highly accurate.