基于RBF神经网络辨识氮爆式破碎器模型及其仿真

Identification and Simulation for Modeling of Nitrogen Explosion Breaker Based on RBF Neural Networks

根据神经网络的辨识原理,建立用RBF神经网络辨识氮爆式破碎器模型的方法。利用辨识实验采集的输入输出数据,通过RBF中的最小均方(LMS)算法动态调整神经网络权值,使达到最佳效果。RBF神经网络辨识分为2步:第1步确定径向基核函数的中心c和宽度d,根据输入向量u的特点,确定隐含层神经元q的个数;第2步采用最小均方(LMS)方法,动态调整隐含层到输出层的权重,优化神经网络模型,并对建立的神经网络模型进行仿真。仿真结果表明:LMS算法具有良好的收敛性质,RBF神经网络模型较好的描述了氮爆式破碎器模型。

Based on the acquired-date of input and output, the model of nitrogen explosion breaker was established by using RBF neural network identification method. The weight vector of the neural network was adaptively adjusted by the least mean square （LMS） algorithm in order to achieve optimal results. The RBF identification was divided into two steps： First of all, the center （c） and width （d） of radial basis function was determined, and the number of Hidden layer neuron （q） was determined according to the characteristics of the input vector （u） ; Secondly, the weight vector of the hidden layer to output layer was adaptively adjusted by the LMS, and moreover the neural net- work modeling established was simulated in order to obtain optimized neural network model. The simulation results show the LMS is good of global convergence, the RBF Neural Network modeling is better description of the Nitrogen explosion breaker model.