基于神经网络的振荡热管传热性能建模

• 作 者: 崔晓钰 翁建华 M.GROLL
• 机 构: 上海理工大学动力学院,无锡快捷半导体有限公司,斯图加特大学核能与能源系研究所 上海200093,江苏无锡214028,德国斯图加特70569
• 出 处: 化工学报
• 发表日期: 2003年09期
• 关键词: 振荡热管 神经网络 冷凝段 训练误差 传热机理 性能模型 神经元数 蛇形管 注量 加热功率

This paper presents artificial neural network (ANN) modeling of heat transfer performance for a pulsating heat pipe (PHP). The investigated PHP is a vertical closed loop copper/ethanol PHP. Fully connected multi-layer feed forward network is adopted and back propagation momentum algorithm, sigmoid node function are used. In the network, two input nodes correspond to heat load and fill rate and the output is a single node for thermal resistance. The matching of the ANN test output data and the experimental data is satisfying. It can be inferred that the ANN model can be applied to accurately model PHP performance.

This paper presents artificial neural network (ANN) modeling of heat transfer performance for a pulsating heat pipe (PHP). The investigated PHP is a vertical closed loop copper / ethanol PHP. algorithm, sigmoid node function are used. In the network, two input nodes correspond to heat load and fill rate and the output is a single node for thermal resistance. The matching of the ANN test output data and the experimental data is satisfying. It can be inferred that the ANN model can be applied to accurately model PHP performance.