质量功能展开中关联关系确定的RBF方法

RBF Method of Determining Functional Relationships for Quality Function Deployment

针对质量功能展开中存在顾客需求和工程特性之间的关联关系不确定和模糊等特质,提出了质量功能展开中关联关系的径向基函数(Radial Basis Function,RBF)神经网络方法。采用三层RBF神经网络,由质量屋中顾客需求组成RBF神经网络的输入层神经元,中间层选用高斯核函数,工程特性组成RBF神经网络输出层神经元,由顾客需求和工程特性关联关系评价样本集组成网络的训练样本集,通过网络训练的方式来获得最优的顾客需求与工程特性的关联关系。最后,结合天然光采光产品开发,进行了实例分析,说明该方法具有计算速度快,拟合精度高的特点。

As the functional relationships between customer requirements and engineering characteristics in Quality Function Deployment （QFD） were uncertain and fuzzy, Radial Basis Function （RBF） to determine functional relationships for QFD was presented. The RBF neural network with three layers was adopted, which includes input layer, hidden layer and output layer. The customer requirements and engineering characteristics in QFD constituted the input and output of the RBF Neural Network respectively, and the nonlinear function of hidden layer was given as the Gaussian function. The training set of RBF neural network was composed by sample set of functional relationships between customer requirements and engineering characteristics, and the optimal relationships were constructed through the neural network training. Finally, a case study of developing light guide product illustrated that the method was feasible and effective with faster calculate speed and high simulation precision.