论文部分内容阅读
运用近年来迅速发展的神经网络技术，成功构造了飞行器总体设计过程中具有重要作用的总体质量估算网，升力系数估算网，以及阻力系数估算网，结果表明训练后的参数估算网比通常使用的近似估计公式具有更高的精度．同时，根据所涉及问题的数据特点，为了提高对网络的训练精度，对现有ＭＢＰ 算法作了进一步改进，仿真结果证明改进的ＭＢＰ 算法具有更高的训练效率．这一思路和方法可适用于机械及航空航天其它产品的总体概念设计过程．
Using the rapid development of neural network technology in recent years, the overall mass estimation network, the lift coefficient estimation network and the drag coefficient estimation network which have an important role in the overall design of the aircraft have been successfully constructed. The results show that the trained parameter estimation network has a better performance than the commonly used The approximate estimation formula has higher accuracy. At the same time, according to the data characteristics of the involved issues, in order to improve the training accuracy of the network, the existing MBP algorithm is further improved. The simulation results show that the improved MBP algorithm has higher training efficiency. This line of thought and methodology can be applied to the overall concept design process for other mechanical and aerospace products.