Feature coefficient prediction of micro-channel based on artificial neural network

作者： | Liu Huang, Weirong Nie, Xiaofeng Wang, Teng Shen |

作者单位： |
^{1}Nanjing University of Science and Technology |

刊名： | Microsystem Technologies, 2017, Vol.23 (6), pp.2297-2305 |

来源数据库： | Springer Nature Journal |

DOI： | 10.1007/s00542-016-3067-0 |

关键词： | Particle Swarm Optimization; Artificial Neural Network; Artificial Neural Network Model; Inertial Force; Computational Fluid Dynamic Simulation; |

英文摘要： | In order to study the flow damping in micro-channels, unsteady Bernoulli equation was adopted to derive the motion equation. Artificial neural network (ANN) was adopted to predict the feature coefficient in the motion equation. Firstly, the motion equation of liquid column, flow in micro-channel, under inertial force, was derived. Then, the numerical mapping relationship between the feature parameters and the feature coefficient of micro-channel was modeled using ANN. Moreover, a hybrid optimization algorithm was developed to train the ANN model, which based on back propagation, particle swarm optimization and genetic algorithm. Finally, by taking the rectangular cross section straight micro-channel as an example, the theoretical approach was demonstrated. The training samples were... |

原始语种摘要： | In order to study the flow damping in micro-channels, unsteady Bernoulli equation was adopted to derive the motion equation. Artificial neural network (ANN) was adopted to predict the feature coefficient in the motion equation. Firstly, the motion equation of liquid column, flow in micro-channel, under inertial force, was derived. Then, the numerical mapping relationship between the feature parameters and the feature coefficient of micro-channel was modeled using ANN. Moreover, a hybrid optimization algorithm was developed to train the ANN model, which based on back propagation, particle swarm optimization and genetic algorithm. Finally, by taking the rectangular cross section straight micro-channel as an example, the theoretical approach was demonstrated. The training samples were... |