基于神经网络的加固计算机振动仿真系统

为降低加固计算机的振动实验成本,提高产品的设计效率,提出了应用神经元BP网络对加固计算机振动仿真的方法.通过对加固计算机振动系统简化模型进行分析,采用最普遍的梯度算法确定BP网络的拓扑结构,利用ANSYS软件分析得到的数据和已有的振动实验数据进行对比分析,对网络模型进行修正,实现了对加固计算机振动的幅频特性进行预测.最后,将预测仿真得到的数据和对样机进行实际振动荻取的实验数据对比分析,表明了该算法的有效性.

Vibration simulation system of reinforced computer based on neural network

To reduce vibration test cost of reinforced computer's and improve the product design efficiency, a method of using BP networks in vibration simulation of reinforced computers is put forward. Analyzing simplified model of reinforced computer vibration system, widely used gradient algorithm is used to confirm the BP network topology. Comparing the data based on ANSYS software with the data of real vibration experiment, the neural network model is amended. Therefore, amplitude-frequency characteristic of reinforced computer vibration could be predicted. Finally, comparing and analyzing the data of prediction and simulation system with data from prototype vibration experiment, the feasibility of the method is validated.