基于神经网络的绝热毛细管新型简化模型

通过对现有各种毛细管仿真模型的分析比较，提出了基于神经网络的新型简化模型——两相区关联模型．该模型针对壅塞两相区各参数间的相关性，建立了无量纲长度与两相区进出口压力间的人工神经网络辨析关系．而且基于等焓假设推导出两相区进出口压力相同的非壅塞态与壅塞态流量关系式．从而获得了快捷而且紧凑的仿真流程．在样本范围内取得了良好的仿真结果，R12和R22壅塞状态下的误差绝对值分别控制在2．14％和2．46％内．

A New Simplified Model of Flow in Adiabatic Capillary Tubes Based on Neural Network

The current capillary simulation models usually have a contradiction between simulation accuracy and simulation speed. After analyzing the current models, a simplified model was developed by applying the non-linear analysis capacity of artificial neural network(ANN). An ANN model was built in the chocked state in two-phase region. As for non-Fanno flow, the mass flux was related to critical mass flux under similar outlet pressure of the chocked state in the assumption that the enthalpy is constant in the course of the flow. According to the simulation results, this model was effective and accurate. The absolute values of simulation deviation of R12 and R22 were limited within 2.14% and 2.46% respectively.