【摘要】 利用计算流体动力学软件FLUENT建立了烟气、容器、保温棉多层热流固耦合模型。模拟得到了大型压力容器热处理升温阶段温度场分布。结果表明,模拟结果与实验相吻合,相对误差在±1.2%以内。基于FLUENT计算结果,以富氧燃烧时火焰温度和容器不同层位置作为输入变量建立含15个隐含层的BP神经网络。采用经过训练的神经网络,预测容器不同层温度,得到的相对误差仅为0.34%,准确性高,省去了大量的FLUENT计算,提高了预测效率。更多还原

【Abstract】 A numerical model which considered the coupling effect of air,pressure vessel and insulation material was developed based on FLUENT software,and the transient temperature field of the large-scale pressure vessel was obtained. The results show that the simulated temperatures meet the experimental ones well,which verifies the accuracy of the model. The relative error is less than 1. 2%. Based on the results of FLUENT calculation,a back-propagation(BP) neural network containing 15 hidden layers is established,in which the flame temperature of oxygen enriched combustion and positions of different layers are the input variables. The temperature distribution of the pressure vessel can be predicted accurately with a relative error of 0. 34%. This method avoids a mass of FLUENT calculation and improves the forecasting efficiency.更多还原