具有随机系数和带间隙的自激振动问题的人工神经网络解法

利用人工神经网络技术(BP网络)研究具有随机系数和带有间隙的初轧机自激振动问题,提出了一种Runge-Kutta法和人工神经网络相结合的求解方法。即利用数值计算和BP网络建立随机系数和稳态振幅之间的关系,从而直接计算出稳态振幅的统计特性。计算结果表明:采用所提出的方法求解,通用性好且可提高计算精度,并得到了间隙与稳态振幅的均值有关,与标准离差无关的有用结论。

Solution on Self-excited Vibration of the System with Random Coefficients and Clearance by Artificial Neural Network

An artificial neural network (Back Propagation network) is applied to solve the problem of self-excited vibration of the system with random coefficients and clearance for a rolling mill and a new method is presented in which Runge-Kutta method and BP network are applied successively. First, the relations between random coefficients and the stationary amplitude of the self-excited vibration are deduced by numerical calculation and a BP network, then, the statistical property of the amplitude can be calculated directly. It is shown that the method is generally adaptable in engineering and the precision of the solution is improved greatly. The result illustrates that the mean value of stationary amplitude is dependent on clearance while its root-mean-square value is not.