基于基函数展开的双隐层过程神经元网络及其应用

Research and application of process neural network with two hidden-layer based on expansion of basis function

提出一类基于基函数展开的双隐层过程神经元网络模型.过程神经元隐层完成对输入信息过程模式特征的提取和对时间的聚合运算,非时变一般神经元隐层用于提高网络对系统输入输出之间复杂关系的映射能力.在输入空间中引入一组函数正交基,将输入函数和网络权函数表示为该组正交基的展开形式,利用基函数的正交性简化过程神经元聚合运算.以旋转机械故障诊断和油藏开发过程采收率的模拟为例,验证了模型和算法的有效性.

A class of process neural network model with two hidden-layer based on expansion of basis function is brought forward. The hidden layer of process neuron performs the aggregation operation of time, while the hidden layer of generic neuron raises the mapping capability of network to the complex relation between the system input and output. By introducing a group of function orthogonal basis into the input space, the input functions and the network weight functions are expressed in the expansion form. Using the orthogonality of basis function simplifies the aggregation operation of process neuron. The machinery fault diagnosis and process simulation of oil reservoir development show the effectiveness of the algorithm and model.