基于数据驱动的蒸发器中流体流动沸腾传热性能预测方法
A data-driven approach for predicting fluid flow boiling heat transfer performance in an evaporator
韩莎莎 1李兴涛 1何畅 2张冰剑 1陈清林1
作者信息
- 1. 中山大学 材料科学与工程学院,广东 广州 510006;广东省石化过程节能工程技术研究中心,广东 广州 510275
- 2. 中山大学 化学工程与技术学院,广东 珠海 519082;广东省石化过程节能工程技术研究中心,广东 广州 510275
- 折叠
摘要
为了准确预测蒸发器中流体流动沸腾传热性能,基于实验数据驱动,采用机器学习中多层感知器(MLPN)为通用逼近器,粒子群算法(PSO)用于优化网络权重和偏置,建立PSO-MLPN模型预测 12 种流体流动沸腾传热特性.神经网络输入参数包括管内径、质量流密度、热流密度、气相质量分数、液体物性参数组合和其他参数(Martinelli 数,摩尔质量和汽化潜热),输出参数为流动沸腾传热系数.网络隐藏层数和神经元数由模型精度和灵敏度分析决定.结果表明,预测值与实验值的平均绝对误差、均方根偏差和决定系数分别为 3.74%、10.09%和 0.984,其中约 90%数据的偏差在±20%以内,与传统经验关联式对比也证明了预测模型的优越性.研究提出的 PSO-MLPN 模型可为今后两相流传热性能研究提供参考,并可用于实际蒸发器设计优化.
Abstract
To accurately predict fluid flow boiling heat transfer performance in an evaporator,a PSO-MLPN model was proposed to predict the flow boiling heat transfer characteristics of 12 fluids based on experimental data-driven approach,which used a multilayer perceptron neural network(MLPN)in machine learning as a general approximator,and particle swarm optimization(PSO)was used to optimize the weight and bias.The input parameters of the neural network included tube inner diameter,mass flow rate,heat flux,vapor quality,liquid physical property parameters combination and other parameters(Martinelli number,molar weight and vaporization latent heat),and the output parameter was the flow boiling heat transfer coefficient.The number of hidden layers and neurons was determined by the model accuracy and sensitivity analysis.The results show that the mean absolute error,root mean squared error and determination coefficient between the predicted and experimental value are 3.74%,10.09%and 0.984 respectively,and deviations of about 90%data are within±20%.The comparison with the traditional empirical correlation also proves the superiority of the prediction model.The proposed PSO-MLPN model can provide guidance for next study on two-phase flow heat transfer performance,which is conducive to the optimization of practical evaporators design.
关键词
流体/流动沸腾/神经网络/传热系数/蒸发器Key words
fluid/flow boiling/neural network/heat transfer coefficient/evaporator引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据