Machine learning-based ion sieve forming strategy: impact on salt lake lithium recovery by fixed beds

Lithium recovery from salt-lake brines is of great significance for addressing the growing demand for lithium resources. Fixed-bed adsorption is a key process in lithium recovery from salt lakes, and the shaping strategy of the adsorbent plays a decisive role in determining fixed-bed performance. This study is the first to apply machine learning to analyze the shaping strategy of fixed-bed adsorbents for lithium recovery from salt lakes. A multi-model analytical framework integrating back propagation (BP), radial basis function (RBF), probabilistic neural network (PNN), generalized regression neural network (GRNN), random forest (RF), and genetic algorithm (GA) models was established to evaluate the effects of material properties, process parameters, and shaping strategies on fixed-bed adsorption performance. The results showed that the liquid-film mass transfer efficiency has a greater influence on fixed-bed performance than the adsorbent material itself. The optimal structural parameters were identified using the radial basis function–genetic algorithm (RBF-GA) model and further validated experimentally by response surface methodology (RSM). The optimized conditions were determined to be a packing diameter of 12.5 mm and a packing density of 0.30. This work provides new insight into the structural design of fixed-bed adsorbents and offers a useful reference for the development and optimization of lithium recovery processes from salt-lake brines.