中文摘要：

介绍了一种BP神经网络的改进Levenberg-Marquardt（LM）算法原理，用这种方法对颗粒碰撞振动系统的阻尼进行了训练和仿真，并将此改进算法与传统算法进行比较．结果表明，该算法稳定、快捷，预测准确，适合应用于对实时性要求比较高的场合，且预测得到的模型与相关文献中的结果一致．

英文摘要：

Using an improved Levenberg-Marquardt algorithm and based on the experimental data, the BP neural network is trained to study and simulate the damping performance of the particle impact vibration system, and the results are compared with those from conventional calculation methods, It is demonstrated that the Levenberg-Marquardt algorithm can greatly speed up the learning process and therefore reduce the training time, and is suitable for real time system identification, The recognized models agree qualitatively well with those in the literatures.