基于MEA-BP神经网络的多孔质轴承参数优化

Parameter Optimization of Porous Bearing Based on MEA-BP Neural Network

为了提高圆盘类多孔质静压止推轴承的静态特性,采用思维进化算法(MEA)优化反向传播(BP)神经网络,创建圆盘类多孔质静压止推轴承静态特性的预测模型,完成轴承静态承载力、静刚度以及耗气量的高精度预测,其预测误差分别低于2%、5%以及5%。基于此模型,以最大化静态承载力和静刚度、最小化耗气量为优化目标值,采用遗传算法对轴承的多孔质参数组合进行多目标优化,实现轴承参数的快速优化设计。优化结果表明:此轴承静态承载力提高了64.53%,静刚度提高了31.93%,耗气量降低了56.52%。

In order to improve the static characteristics of disc porous aerostatic thrust bearings,mind evolutionary algorithm(MEA)was used to optimize the back propagation(BP)neural network to create the prediction model of the static characteristics of disc porous aerostatic thrust bearings.The high precision prediction of static bearing capacity,static stiffness and gas consumption was completed,and the prediction errors were less than 2%,5%and 5%respectively.Based on this model,to maximize static bearing capacity and static stiffness and minimize gas consumption as optimization objectives,the multi-objective optimization of the bearing porosity parameter combination was carried out by using genetic algorithm,and the rapid optimization design of bearing parameters was realized.The optimization results show that the static bearing capacity increases by 64.53%,the static stiffness increases by 31.93%,and the gas consumption decreases by 56.52%.