基于神经网络的车身零件检具概念设计

Concept design of measuring fixtures for auto-body parts by neural network

采用人工神经网络进行车身覆盖件检具概念设计,以检测特征的7个分量作为神经网络的输入,以检具类型分量作为输出,对构成的神经网络用生产中的100个实例作为样本进行训练,达到误差平方和小于0.001的目标,得到检具概念设计神经网络模型,并通过车身一零件检具概念设计为例验证了该方法的有效可行,从而达到在一族相似零件的众多检具概念设计方案中进行优选的目的。

Measuring fixtures for auto-body parts are designed by neural network based on discussing measuring features of auto-body parts, and the input of neural network model is composed of 7 sets of measuring features, the output is composed of 4 types of Measuring fixtures. In order to obtain error （between output of sample and model） less than 0.001, the neural network model is trained by 100 sets of example dates, then a example is used to validate Neural network model for selecting types of Measuring fixtures, and results show that the model can select optimal type of Measuring fixture for measuring features of auto-body part.