基于混沌和人工神经网络方法建立计算气固循环流化床局部颗粒浓度时间序列的非线性模型

Nonlinear Model for Generation of the Solids Holdup Time Series in the CFB Riser Based on Combination of Chaos with Artificial Neural Network

基于气固循环流化床是一混沌动力学系统,故采用了耦合混沌理论中的重构相空间方法与人工神经网络的非线性映射功能,建立起提取气固循环流化床颗粒浓度时间序列的非线性模型,并将此模型应用到f100mm×16m,FCC固体颗粒的上行气固循环流化床系统。 由模型产生的局部颗粒浓度时间序列与实验测得局部颗粒浓度时间序列的统计特性、功率谱及非线性动力学特征吻合较好。

A nonlinear model for generation of solids holdup time series in the CFB riser based on combination of chaos with artificial neural network was proposed. In this model Takens phase space reconstruction method was used to reconstruct the attractors, which represent the system hydrodynamics from the single variable time series, and the RBF(radial based function) artificial neural network was used to fit the attractors. The model was trained to simulate the dynamic behavior of the local solids holdup fluctuations measured in a circulating fluidized bed riser with 16m high and 0.10m ID. The experiments were conducted with FCC particles, the superficial gas velocity Ug ranging from 3.5 to 8.0ms-1 and the solids circulating flux Gs ranging from 50 to 200kgm-2 s-1. Time series signals of solids holdup fluctuation were measured at a frequency of 900 Hz using an optical fiber probe located at 8 axial and 11 radial positions, respectively. Then the trained ANN model was used to generate the local solids holdup time series in different local positions and various operating conditions. The local solids holdup time series generated by ANN model were compared with the experiment measured time series by time-averaged statistics characteristics、power spectrum、and nonlinear features. The results show that they coincide with each other satisfactorily.