提出在BP网络的输入层与隐层或径基函数网络的输入层与隐层之间增加一个只有 2个结点的线性函数层(Z层 ) ,以构成基于BP网和基于径基函数网的降维映射网络。这两种网络均将多维输出Y (可为一维 )与多维输入X之间的非线性映射关系转变成与二维向量Z之间的非线性映射关系。网络学习后 ,就可以在由Z构成的二维映射平面上描绘出输出向量的等值线 ,通过这些等值线可全景式地、准确可靠地确定出样本数据集的最优操作区域 ,实现混凝土配合比优化设计 This paper proposes to add a linear function layer (Z layer) with only two nodes between the input layer and hidden layer of the input layer and the hidden layer or RBF network of the BP network to form a BP neural network based on the BP network and the RBF network Dimension reduction mapping network. Both of these networks transform the non-linear mapping between multi-dimensional output Y (which may be one-dimensional) and multi-dimensional input X into a non-linear mapping with two-dimensional vector Z. After learning from the network, contour lines of the output vectors can be drawn on the two-dimensional mapping plane formed by Z. Through these contour lines, the optimal operation area of ​​the sample data set can be determined panoramicly, accurately and reliably, Concrete mix optimization design