China Mechanical Engineering ›› 2026, Vol. 37 ›› Issue (6): 1393-1401.DOI: 10.3969/j.issn.1004-132X.2026.06.012

Previous Articles    

Research on Dynamic Behavior Prediction of Droplet Collision and Diffusion Based on Improved Grey Wolf Algorithm

YU Zhiyong1(), LI Xin2(), WANG Zhengmao1   

  1. 1.School of Mechanical Engineering,Shandong University of Technology,Zibo,Shandong,255049
    2.School of Mechanical Engineering,Hangzhou Dianzi University,Hangzhou,224005
  • Received:2025-12-26 Online:2026-06-25 Published:2026-07-17
  • Contact: LI Xin

基于改进灰狼算法的液滴碰撞扩散动态行为预测研究

于智勇1(), 李昕2(), 王正茂1   

  1. 1.山东理工大学机械工程学院, 淄博, 255049
    2.杭州电子科技大学机械工程学院, 杭州, 224005
  • 通讯作者: 李昕
  • 作者简介:于智勇,男,2000年生,硕士研究生。研究方向为机电一体化。发表论文1篇。E-mail:yuzhiyong2025@163.com
    李昕*(通信作者),男,1992年生,博士研究生、讲师。研究方向为液固耦合。发表论文5篇。E-mail: 42995@hdu.edu.cn
  • 基金资助:
    杭州电子科技大学校基金(GK249909299001-027)

Abstract:

To explore the dynamic behaviors of droplet impact and spreading and to further improve the efficiency of designing process parameters for droplet spreading, a dataset for an intelligent optimization algorithm was established based on a finite element model, and a prediction model based on extreme learning machine optimized by improved grey wolf optimizer(ELM-IGWO) was proposed. Using the initial droplet diameters, velocities, heights, and material contact angles as inputs, and maximum spreading diameters and central thicknesses of droplets as outputs, the dynamic behaviors of droplet impact and spreading were predicted. By comparing with the back propagation neural network(BP), back propagation neural network optimized by genetic algorithm(BP-GA), extreme learning machine(ELM), and extreme learning machine optimized by grey wolf optimizer(ELM-GWO)models, it is found that the proposed algorithm has the greater accuracy in predicting the dynamic behaviors of droplet impact and spreading. In addition, the processing parameters of droplet impact and spreading were optimized by using an improved multi-objective non-dominant sorting genetic algorithm based on artificial neural networks(ANN-IMNSGA-Ⅱ),the experimental results show that the average relative error between the predicted maximum spreading diameters and the experimental data is as 3.54%, and the prediction error for the central thicknesses is as 5.71%.

Key words: droplet impact, finite element model, intelligent optimization algorithm, prediction model

摘要:

为探究液滴撞击与铺展动力学行为,提高液滴铺展工艺参数设计效率,基于有限元模型建立了智能优化算法的数据集,并提出了一种基于改进灰狼优化算法的极限学习机(ELM-IGWO)预测模型。以液滴初始直径、速度、高度以及材料的接触角作为输入,液滴最大铺展直径和中心厚度为输出,对液滴撞击与铺展的动力学行为进行了预测。通过与反向传播神经网络、遗传算法优化反向传播神经网络、极限学习机以及传统的灰狼优化极限学习机模型进行比较,发现所提算法在预测液滴撞击和扩散的动态行为方面具有更高的准确性。此外,利用基于人工神经网络的改进型多目标非支配排序遗传算法对液滴撞击与铺展的工艺参数进行优化,实验结果表明,预测的最大铺展直径与实验数据的平均相对误差为3.54%,中心厚度的预测误差为5.71%。

关键词: 液滴铺展, 有限元模型, 智能优化算法, 预测模型

CLC Number: