Screening Framework Of MXene Single Nanoporous Membrane Seawater Desalination System Based On Ensemble Learning And Optimization Algorithm

In the face of the severe global freshwater scarcity,developing energy-efficient and low-cost seawater desalination technologies is imperative.Reverse Osmosis（RO）technology is considered one of the most promising solutions to address the world’s freshwater crisis,and it has been widely applied to large-scale distributed seawater desalination.MXene is a new type of two-dimensional material composed of transition metal carbides and nitrides layered compounds.Due to its unique properties,including excellent stability and charge characteristics,MXene RO membrane with nanopores is considered a promising membrane for seawater desalination.However,due to the limitations of experimental conditions,there is currently no way to precisely design various nanopore structures and operational conditions,which hinders the rational design of MXene single nanopore membranes.Additionally,the inherent complexity of the desalination process and the various adjustable properties of the membrane and nanopores themselves make it challenging to accurately predict the desalination performance of new materials.To effectively address the above issues,this paper innovatively proposes a screening framework for MXene single nanopore membrane seawater desalination system based on the combination of Ensemble Learning and Optimization Algorithm.The main contents include:1.Using Molecular Dynamics（MD）simulation to establish 288 MXene single nanopore membrane seawater desalination systems with different environmental design conditions（driving pressure and saltwater concentration in the feed solution）and membrane nanopore design conditions（pore area,pore shape,membrane material,and membrane charge coefficient size）.The water flux and salt rejection rate of each system are calculated to represent the desalination performance of each system,and an MXene single nanopore membrane seawater desalination performance database is established.We found that it is challenging to achieve both high water flux and high salt rejection rate in the desalination system.Ti3C2OH2 has the strongest filtration capacity among the four membrane materials（Ti₃C₂,Ti₃C₂F₂,Ti₃C₂O₂,and Ti₃C₂OH₂）under the same conditions,but it also has the poorest ion retention capacity.2.Based on the MD result data,two ensemble machine learning methods,Random Forest（RF） and XGboost,are used to establish MXene single nanopore membrane seawater desalination performance regression prediction models.The optimal hyperparameters are determined through five-fold cross-validation and grid search methods.The results show that both models have good predictive ability for water flux and salt rejection rate,and the R2 value can reach above 0.9.Through the feature importance analysis of the model,we conclude that the driving pressure and charge coefficient are the two most important factors that affect the water flux and salt rejection rate of MXene single nanopores.The partial dependence analysis results show that the driving pressure and pore area are proportional to the water flux but inversely proportional to the salt rejection rate,while the charge coefficient size and concentration show the opposite trend.3.Establish an optimization model based on Genetic Algorithm（GA）and Particle Swarm Optimization（PSO）in a scholarly manner.Introduce machine learning models as fitness functions to evaluate the candidate desalination systems at each iteration,and use this as the basis for rapid screening of the optimal MXene desalination system.Results show that PSO has a faster convergence rate than GA and can quickly find the global optimal solution.Through MD simulations,we validate that the Ti₃C₂O₂＿Charge1.2 system（pore shape:circular,MXene membrane material:Ti₃C₂O₂,pore area:71.4（?）2,applied pressure:330MPa,concentration:0.4 M,charge coefficient:1.2）selected by the PSO method is the most excellent:water flux:97.3 number/ns;salt rejection rate:100%.Molecular mechanism analysis reveals that the O-terminal groups in Ti₃C₂O₂＿Charge1.2 have a large negative charge,which can produce electrostatic interactions with ions,thereby blocking ions from entering the pores.Additionally,a reasonable distribution of pore charges ensures that ions do not stay in the pores for a long time,obstructing water molecules from passing through.By adjusting the charge coefficient,this electrostatic effect can be further amplified.The results demonstrate that the MXene single nanopore membrane screening framework based on XGboost and PSO algorithms proposed in this paper can quickly screen out the optimal seawater desalination system that meets specific performance targets.This provides a new approach for future research on the application of MXene in seawater desalination.