基于RBF神经网络的气体流量软测量模型研究

流量信号是热工过程中非常重要的一个信号。由于流量信号存在着非线性、随机性和易受干扰的特点，很难建立起一个准确的测量模型，如传统的3种圆管紊流流速分布的近似模型，基于这些模型的传统测量方法很难测量出准确的流量值。该文提出的基于径向基函数（RBF）神经网络的流量测量模型，采用了带有遗忘因子的梯度下降算法来确定隐层基函数中心的位置和输出层权值的大小。计算结果表明这种模型计算量小、精度高，且算法简单实用。实验结果说明，基于这种模型的流量测量精度较以往模型有很大提高。

Study on Soft-sensing Model of the Gas Flowrate Measurement

The flowrate signal is an important parameter in thermal processes.However,flowrate signals generally contain nonlinearity and randomicity,it is difficult to build accurate models by traditional methods such as the three approximate mathematical models of the velocity distribution of turbulent flow in cylindrical pipes,and so the traditional method of flowrate measurement based on these models was impossible to get accurate results.A novel RBF neural network model is presented in this paper.A gradient descend algorithm with a momentum factor in this network model is introduced to decide the positions of hidden layer RBF centers and output layer weights.The soft-sensing model is established with the novel RBF neural network,the computational results show that the model has less calculation and high precision,and the relevant algorithm is very simple and useful.The experimental results show clearly that the measurement precision of flowrate is greatly improved comparing with traditional methods.