基于神经网络的电梯导轨校直计算

以电梯导轨校直过程中支点跨距、初始最大挠度为输入参数,校直行程和校直后测点的极差为输出,采用有限元优化设计方法得到的数据作为样本,建立了神经网络模型并使用样本对其进行了训练,利用随机数据对训练后的神经网络进行了测试.结果表明,将经过充分训练的神经网络应用于电梯导轨校直行程计算可以得到精确的计算结果,同时神经网络也可以反映校直后的导轨弯曲形式,为电梯导轨校直专家系统提供了一定参考.

Calculation for straightening of elevator guide rail based on neural network

A neural network was built for the straightening process of the elevator guide rail with the supporting points span and the maximal initial deflection as the input parameters,the straightening stroke and the range of measuring points after straightening as the output parameters.And the data obtained by finite element optimization design method shall be as a specimen to build a neural network model,which will be trained with the specimen.Then the neutral network after training shall be tested by random data,which results indicate the neutral network after training up shall be applied in straightening stroke calculation of the elevator guide rails to obtain an accurate result.Meanwhile,the deformation of the guide rail after being straightened could be reflected by the results of the neural network,which may provide reference for straightening expert system of the elevator guide rail.