Abstract: CFD in designing and optimizing the structure of the trickle bed gas-liquid distributor have the disadvantages of the slow speed and high calculation time cost. In this paper, we selected the support vector regression machine (SVR) model as the proxy model for prediction. In order to obtain a fast and accurate SVR model, the particle swarm optimization (PSO) algorithm was used to optimize the parameters of support vector regression (SVR), including the penalty coefficient C, the parameter g of the kernel function, and the insensitive loss coefficient ε. Then, using the PSO-SVR hybrid model as the data source coupled with the response surface method (RSM), the structural parameters corresponding to the minimum liquid uneven distribution were calculated. The CFD results showed good agreement with the values predicted by the response surface method, and the liquid distribution unevenness was 0.159 and 0.162, respectively. The research shows that the hybrid PSO-SVR-RSM method provides an effective tool for guiding the optimal design of gas-liquid distributors in trickle beds.

Key words: gas-liquid distributor, support vector regression, particle swarm, response surface methodology, CFD