Abstract: Avionics equipment is playing an increasingly important role in the integration of avionics systems. Due to the harsh airborne environment, such as high and low temperature, strong impact, acid corrosion and other factors, the fault features of avionics equipment have strong randomness, complexity, and coupling with external stress, which leads to the support technology of avionics equipment becoming a research hotspot. Aiming at the classical problems of engineering, this paper proposes a fault prognosis method based on the fusion of transferred multi-deep-belief network (DBN) models. First, Transfer learning and Dropout strategy are used to improve the feature extraction capability of the model. Secondly, the optimized genetic algorithm effectively determines the fusion weight of each DBN model according to the real-time data. Finally, the complete hybrid framework was adopted to estimate the complete residual life of the equipment. To verify the performance of the proposed method, the experiment was performed according to the data of the power supply equipment. The results show that the proposed method has higher accuracy and stability than traditional support vector machines and classical DBN models which is helpful to realize the maintenance based condition for avionics equipment.

Key words: avionics equipment, deep belief network, fusion strategy