Accurate Weight Prediction of a New Aircraft’s Electrical System Based on Neural Network

Weight prediction is essentially important in the overall design of an aircraft. A conservative weight prediction will result in uncompetitive performance of the aircraft, while an optimistic one will result in the compromise of weight limit and aircraft performance. According to the composition and characteristics of aircraft electrical system and the experience of other aircraft design, a three-step weight prediction method of the electrical system is proposed. The weight prediction of electrical system is divided into three parts: the weight prediction of newly developed airborne equipment, the weight optimization and prediction of cables, the weight prediction of equipment and cable installations. Firstly, the weight of the newly developed airborne equipment is predicted by introducing two neural network structures: generalized regression neural network (GRNN) and fitnet neural network, then the neural networks are optimized to reduce the prediction deviation. For the best case, the weighted deviation reaches 3.40%. Secondly, the electric cable weight is accurately predicted through the optimization of cable weight by using three-dimensional (3D) numerical model and empirical method. Thirdly, the auxiliary weight of the installation of cables and airborne equipment include mounting brackets, protective covers, grounding, etc. are accurately predicted by using empirical data from weight database. The results show that the weight prediction process basically realizes accurate prediction of the total weight of the electrical system, and therefore improves the reliability of aircraft technical scheme and the accuracy of aircraft performance calculation.