论文部分内容阅读
论述了使用神经网络求解结构动力学反问题的基本原理和方法，并针对ＢＰ网络在使用中存在的主要问题，提出了网络快速学习算法，以及训练样本正交化处理方法和反解结果精度的自适应修正等一系列解决方法。应用上述理论进行了设备动态诊断的数值仿真和试验研究，结果令人满意。
In this paper, the basic principles and methods of neural network to solve the inverse problem of structural dynamics are discussed. Aiming at the main problems existing in the use of BP network, a fast network learning algorithm is proposed, and the orthogonalization of training samples and the accuracy of the solution Adaptive correction and a series of solutions. Using the above theory, the numerical simulation and experimental study on dynamic diagnosis of equipment are carried out with satisfactory results.