Objective The aim is to prolong the service life of methyl diethanolamine (MDEA) solution and enhance the absorption effect of amine liquid.

Method The electrodialysis method was used to remove thermally stable salts from MDEA waste liquid. Single-factor experiments were conducted to investigate the effects of experimental voltage, operation time, electrode spacing, and polar liquid concentration on solution desalination rate, with variance analysis employed to screen out significant influencing factors. Response surface methodology (RSM) and radial basis function (RBF) neural network model were used to optimize the model on the basis of the experiments.

Result RSM and RBF neural network models were established. The training was conducted with 70% of the Box-Behnken experimental data, and 15% of the experimental data were used for testing and validation, respectively. The coefficients of determination (R²) of the training set, test set, validation set, and the overall data set were 0.998 35, 0.997 83, 0.999 95 and 0.997 79, respectively, indicating a good fitting ability. Comparison of RSM and RBF models showed that the RBF model outperformed the RSM model in all indicators. By coupling the RBF neural network with a genetic algorithm for optimization, the optimal process conditions were obtained after 135 iterations. When the experimental voltage was 28.7 V, the running time was 14.92 h, and the polar liquid concentration was 0.158 mol/L, the solution desalination rate could reach 96.4%. The t-test results showed that there was no significant difference between the predicted value and the actual verification value, which provided a reliable basis for the removal of thermally stable salts in MDEA solution.

Conclusion The determined optimal process parameters can provide a reference for similar devices.