<p>The research on the dynamic behavior of droplet collision and diffusion based on piezoelectric jets is of great significance in fields such as energy, materials science and biomedical engineering. However, in different application scenarios, the requirements for the maximum diffusion diameter and thickness of droplets are different. To address the bottlenecks of finite element modeling in terms of computational efficiency and model universality, we innovatively propose a hybrid computing framework that combines finite element simulation and intelligent optimization algorithms. Firstly, establish the finite element model of droplet impact diffusion; Secondly, the Extreme Learning Machine prediction model based on the improved Grey Wolf optimization (ELM-IGWO) algorithm was proposed; Taking the initial diameter, velocity, height and contact Angle of the droplet as inputs and the maximum diffusion diameter and central thickness as outputs, the dynamic behaviors of droplet impact and diffusion were predicted. Through comparison, it is found that the prediction method proposed in this paper has the highest accuracy in predicting the dynamic behavior of droplet impact and diffusion. Finally, through the verification experiment of high-speed camera design, the results show that the average relative error between the predicted maximum diffusion diameter and the experimental data is 3.54%, and the prediction error of the central thickness is 5.71%.</p>