Simulation Screening and Process Optimization Methods for Anhydrous Hydrogen Fluoride Desulfurization Adsorbents

Abstract

This invention discloses a simulation screening and process optimization method for anhydrous hydrogen fluoride desulfurization adsorbents, relating to the field of anhydrous hydrogen fluoride technology. The method includes: constructing a database to establish a basic screening library; constructing an intelligent screening engine to correct the force field parameters of materials in the basic screening library to capture the high-polarity solvent effect; calculating and extracting adsorption selectivity, Henry's law coefficient, and working capacity indicators to screen a primary set of candidate materials; selecting core candidate materials; determining the optimal adsorbent model; and converting the microscopic adsorption data of the optimal adsorbent model into macroscopic process parameters, importing them into a macroscopic process simulator for prediction, and completing process optimization. This invention simulates the strong polarity solvent effect by constructing a corrected force field environment, combined with an active learning strategy to achieve high-throughput screening of adsorbent materials; it optimizes adsorption sites using electronic-level refinement and modification strategies, and couples microscopic adsorption data with a macroscopic mass transfer model to achieve accurate prediction and optimization of the adsorption process.

Description

Technical Field

[0001] This invention relates to the field of anhydrous hydrogen fluoride technology, specifically to a method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbents. Background Technology

[0002] Anhydrous hydrogen fluoride, as an important basic raw material in the fluorochemical industry, has a decisive impact on the quality of downstream products due to its purity. During industrial production, anhydrous hydrogen fluoride often contains trace amounts of sulfur impurities. The presence of these impurities can lead to catalyst poisoning or affect the performance of the final product. Currently, adsorption is a commonly used technique for removing trace sulfur impurities.

[0003] However, existing adsorbent screening and process design processes have significant limitations. First, when screening adsorbents for anhydrous hydrogen fluoride systems, existing simulation methods often fail to accurately simulate the interference of strongly polar solvent environments on the adsorption process, leading to a serious discrepancy between predicted adsorption selectivity and actual test results. Materials with theoretically excellent performance often exhibit near-complete loss of adsorption capacity under real-world conditions. Second, traditional full-scale computational screening methods suffer from enormous computational resource consumption and excessively long cycles, making it difficult to quickly locate target structures from a vast pool of novel porous materials. Furthermore, in the transition from microscopic screening to macroscopic process design, existing technologies often neglect the thermal effects during adsorption and their impact on bed stability distribution, resulting in discrepancies between predicted breakthrough times and actual operating data, hindering effective guidance for adsorption tower process optimization. Summary of the Invention

[0004] The purpose of this invention is to provide a method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbents, thereby solving the problems existing in the background technology.

[0005] To address the aforementioned technical problems, this invention provides a method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbents, comprising the following steps:

[0006] S1. Construct a database of porous material crystal structures, perform geometric topological analysis on the material models in the database, calculate the pore size limit diameter and the maximum pore diameter, set the pore size screening threshold to 3.0 Å to 5.0 Å, remove structures with pore size limit diameters smaller than this threshold that do not meet the requirements of sulfide molecular dynamics diameter, and establish a basic screening library.

[0007] S2. A smart screening engine is built for anhydrous hydrogen fluoride solvent environment. The interaction energy between HF molecules and framework atoms is calculated at the density functional theory level using generalized gradient approximation or hybrid functionals. Lennard-Jones potential parameters are fitted to correct the general force field, or an implicit solvent model with dielectric constant set to 50-84 is introduced to correct the force field parameters of materials in the basic screening library in order to capture the high polarity solvent effect.

[0008] S3. Under the modified force field environment, the working conditions are set at 273K to 333K and 0.1bar to 5.0bar. A competitive adsorption scenario of anhydrous hydrogen fluoride and sulfur impurities is constructed with a molar ratio of 1000:1 to 10000:1. High-throughput grand canonical Monte Carlo simulation is performed under grand canonical ensemble μVT. The number of equilibrium steps and production steps are set to 1×105 to 5×107. The adsorption selectivity, Henry's law coefficient and working capacity index are calculated and extracted. The top 5% to 10% of primary candidate materials are screened out.

[0009] S4. Perform electronic-level refinement on the primary candidate material set, calculate the change in electron cloud density of adsorption sites using density functional theory, calculate the differential charge density through Bader charge or Mulliken charge analysis, set an adsorption energy threshold to determine whether a chemisorption mechanism exists, and select the core candidate materials.

[0010] S5. Based on the structure-activity relationship, a modification strategy is implemented to generate core candidate materials. Functional groups are introduced at the coordinate unsaturated metal sites or organic ligand side groups of the framework according to the geometric constraints of bond lengths of 1.5 Å to 2.5 Å. The performance is then verified by high-throughput giant canonical Monte Carlo and density functional theory joint simulation to determine the optimal adsorbent model.

[0011] S6. Transform the microscopic adsorption data of the optimal adsorbent model into macroscopic process parameters, fit a multi-parameter adsorption isotherm model, and import it into the macroscopic process simulator. Calculate the mass transfer coefficient based on the linear driving force model, perform breakthrough curve and cycle performance prediction, and complete process optimization.

[0012] Preferably, in step S2, when constructing the intelligent screening engine, the variance of the electrostatic potential distribution of the material skeleton, the energy of local sulfur affinity sites, and the hydrogen fluoride-skeleton interaction energy are extracted as the core descriptor set for subsequent feature mapping and model training.

[0013] Preferably, between steps S3 and S4, there is also an active learning enhancement step, which constructs a Gaussian process regression or neural network model, trains the model using a small amount of high-precision computational data, quickly predicts the properties of massive materials, and automatically iterates the model through uncertainty sampling to predict materials with low confidence.

[0014] Preferably, in step S5, the specific operation of introducing functional groups is as follows: virtually grafting amino groups onto the organic linkers of the material skeleton or introducing cuprous ions into open metal sites, setting the coverage of the modified groups to 10% to 50%, and adjusting the spatial orientation of the groups through a geometric optimization algorithm to minimize steric hindrance energy and enhance the specific adsorption of sulfur impurities.

[0015] Preferably, in step S6, the adsorption data obtained from the microscopic simulation is fitted to the parameters of the Langmuir model or the Toth model and imported as an input file into the breakthrough curve simulator.

[0016] Preferably, in step S6, when performing breakthrough curve and cycle performance prediction, a non-isothermal and non-adiabatic fixed-bed adsorber model is established, and the influence of adsorption heat on the temperature distribution of the adsorption tower bed is coupled. The dynamic response of apparent gas velocity in the range of 0.5 cm / s to 10.0 cm / s, feed temperature in the range of -20℃ to 50℃, and bed height in the range of 0.5 m to 5.0 m to breakthrough time and mass transfer zone length is simulated and investigated, and the structural stability after 50 to 500 adsorption-desorption cycles is predicted.

[0017] Preferably, it also includes an experimental verification step, in which the top 3 to 5 materials in the calculation are selected for synthesis and small-scale testing;

[0018] The synthesis steps include: dissolving the metal precursor salt and organic ligand in a molar ratio of 1:0.5 to 1:5 in a polar solvent, ultrasonically dispersing for 10 min to 60 min, placing in a high-pressure reactor and carrying out a solvothermal reaction at 80℃ to 180℃ for 12 h to 72 h, and after the reaction, gradually cooling at a rate of 1℃ / min to 5℃ / min, separating the product by centrifugation, performing 3 to 10 solvent exchanges with a low-boiling-point solvent, and finally performing gradient heating and vacuum activation at 100℃ to 250℃ for 4 h to 24 h to remove solvent molecules from the pores;

[0019] Small-scale trials include comparing the penetration curves measured in experiments with the simulated predicted curves, and verifying the reliability of the screening process by plotting parity check diagrams.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] By introducing a correction mechanism and a continuous medium model for specific polar environments, the microscopic effects in complex solvent systems can be accurately captured. This overcomes the limitations of conventional simulation methods in predicting insufficient accuracy when dealing with highly polar environments, enabling the screened materials to exhibit good stability in practical applications. It effectively addresses the competitive interference of solvent molecules on impurity adsorption, ensuring the reliability and scientific rigor of the material screening process.

[0022] By employing intelligent sampling and iterative prediction strategies, the allocation of computing resources is optimized while maintaining screening accuracy. This enables the rapid identification of potential high-performance material structures and avoids invalid calculations for low-confidence candidates. It not only shortens the conversion cycle from theoretical models to laboratory verification but also increases the throughput of processing massive amounts of candidate materials, providing efficient tool support for the rapid development of novel adsorption materials.

[0023] By coupling microscopic adsorption parameters with macroscopic heat and mass transfer models, a more realistic process simulation environment is established. By considering heat migration during adsorption and its impact on equilibrium, the performance of the adsorption bed during dynamic operation can be predicted more accurately. This provides a theoretical basis for the precise optimization of process parameters, helps to avoid risks in industrial operation, and improves the predictability of adsorption process design. Detailed Implementation

[0024] Example 1

[0025] This embodiment provides a simulation screening and process optimization method for anhydrous hydrogen fluoride desulfurization adsorbents. This method fully includes all technical features and is a preferred embodiment of the present invention.

[0026] In step S1 of this embodiment, the pore size sieving threshold is set to 3.0 Å to achieve physical sieving using steric hindrance effect. This threshold setting can effectively exclude dense structures that cannot accommodate sulfur impurity molecules, ensure that sulfide molecules can enter the pores, and at the same time reduce the amount of invalid calculations, thus constructing a basic screening library with potential mass transfer channels.

[0027] In step S2 of this embodiment, an implicit solvent model with a dielectric constant of 50 is introduced to simulate the high dielectric environment of anhydrous hydrogen fluoride. Specific simulation parameters are set as follows: van der Waals interactions are handled using the Lennard-Jones 12-6 potential energy model with a cutoff radius of 12.8 Å; electrostatic interactions are handled using the Ewald summation method with an accuracy of [missing value]. When fitting Lennard-Jones potential parameters to correct the universal force field, the least squares method is used as the optimization algorithm. The objective function is set to minimize the root mean square error between the single-point energy calculated by DFT and the potential energy calculated by the force field, thereby ensuring that the force field parameters can accurately reproduce the quantum mechanical level interaction surface. Subsequently, the variance of partial charge, local aperture distribution, and electrostatic potential distribution of the framework atoms are extracted. As the core descriptor; the formula for calculating the variance of the electrostatic potential distribution is:

[0028]

[0029] The total number of sampling points on the Connolly surface or van der Waals surface;

[0030] Used to quantify the charge nonuniformity of the framework surface; these descriptors can quantify the competitive adsorption strength between strongly polar HF molecules and the framework, thereby correcting the deviation of the universal force field in extreme solvent environments.

[0031] In step S3 of this embodiment, a low-temperature and low-pressure operating condition (273K, 0.1bar) is set to simulate the operating environment of the fine desulfurization section; a Gaussian process regression model is introduced as an active learning strategy; the input feature vector of the model includes the following obtained in steps S1 and S2: maximum pore size LCD, limiting pore size PLD, specific surface area ASA, and the corrected electrostatic potential distribution variance. The model's output target variable is the HF / sulfide selectivity obtained from GCMC simulation. By utilizing the Bayesian optimization principle and the expectation-enhancing acquisition function, material regions with higher uncertainty are prioritized for screening while ensuring accuracy. This significantly reduces the number of costly GCMC calculations and achieves an exponential improvement in screening efficiency. The GCMC simulation in this step is performed using the RASPA2.0 software platform.

[0032] In step S4 of this embodiment, Bader charge analysis can reveal the amount of charge transfer between sulfur impurities and adsorption sites at the electronic level, thereby distinguishing between physical adsorption and chemical adsorption, and ensuring that the screened materials have sufficiently strong binding energy to resist solvent erosion; the electronic structure calculation in this part is completed using VASP 5.4 or Gaussian 16 software.

[0033] In step S5 of this embodiment, amino ( Using ) as a modifying group and setting a low coverage of 10%, the aim is to utilize the lone pair electrons of the amino group to form Lewis acid-base interactions with sulfur impurities, while avoiding pore blockage caused by excessive coverage, thereby achieving the best balance between adsorption capacity and mass transfer rate, and finally constructing an optimal adsorbent model with both high activity and low resistance.

[0034] In step S6 of this embodiment, it is crucial to establish a non-isothermal, non-adiabatic model based on the optimal adsorbent model output in step S5, because the HF desulfurization process is accompanied by significant thermal effects. By coupling the adsorption heat, the influence of bed hotspot migration on breakthrough time can be accurately predicted, thereby avoiding the risk of thermal runaway in actual operation. Simulation of low temperature and low flow rate conditions verifies the high selectivity potential of the adsorbent in fine treatment scenarios. The macroscopic process simulation in this step is performed using AspenAdsorption V11 software.

[0035] In the verification phase of this embodiment, to verify the accuracy of the screening results, the top-ranked candidate material was selected for physical synthesis. Its structure was identified as an amino-functionalized UiO-66 derivative, specifically using zirconium tetrachloride as the metal precursor salt. The organic ligand selected is 2-aminoterephthalic acid ( ). The solvent used was DMF (N,N-dimethylformamide); the specific procedure was as follows: weigh 1.16 g (5 mmol) With 0.45g (2.5mmol) The solution was dissolved in 150 mL of LDF and ultrasonically dispersed for 30 min to form a homogeneous precursor solution intermediate. The solution was then transferred to a 200 mL Teflon-lined stainless steel reactor, sealed, and placed in an oven. The reactor was heated to 120 °C at a heating rate of 2 °C / min and kept at this temperature for 24 h to crystallize. After the reaction, the solution was slowly cooled (1 °C / min) to room temperature to reduce crystal defects and obtain a highly crystalline material, ensuring that the experimental performance was highly comparable to the theoretical calculation model.

[0036] Example 2

[0037] This embodiment provides a simulation screening and process optimization method for anhydrous hydrogen fluoride desulfurization adsorbent, which is another parameter combination method based on the technical solution of Embodiment 1;

[0038] In this embodiment, the pore size screening threshold is increased to 3.5 Å, with the aim of screening out framework structures with larger pore volumes to meet the adsorption requirements of macromolecular organic sulfur impurities.

[0039] In this embodiment, the SMD solvent model is used and the dielectric constant is set to 60, which further refines the description of the solvent continuous medium and captures the solvation energy differences in different polarity regions.

[0040] In this embodiment, the ambient temperature and pressure (298K, 1.0bar) conditions are simulated, and a neural network model is used to process nonlinear feature mapping, thereby realizing rapid prediction and iteration of the properties of a large number of materials.

[0041] In this embodiment, Mulliken charge analysis is used to evaluate the changes in electron distribution caused by orbital overlap, which helps to determine the chemisorption mechanism.

[0042] In this embodiment, cuprous ions ( And set a 20% coverage, using With sulfur atoms The complexation mechanism significantly enhances the ability to specifically identify trace sulfur;

[0043] In this embodiment, the Toth model is used to fit the adsorption data, which more accurately describes the adsorption behavior of non-uniform surfaces; the mathematical expression of the Toth model is:

[0044]

[0045] To balance the adsorption amount;

[0046] This represents the saturation adsorption capacity.

[0047] It is the adsorption affinity constant;

[0048] For pressure;

[0049] A parameter characterizing the surface inhomogeneity of the adsorbent. ;

[0050] This parameter is temperature-dependent; simulation results verify the effect of the modification strategy on improving sulfur capacity under normal industrial operating conditions.

[0051] In this embodiment, the verification object is the screened optimal material, which is confirmed to be the cuprous ion-functionalized HKUST-1 (Cu-BTC) material; the specific synthesis and modification process is as follows:

[0052] Matrix synthesis: Weigh 2.41 g (10 mmol) of copper nitrate trihydrate ( ) and 2.10g (10mmol) benzoic acid ( The product was dissolved in a mixed solvent consisting of 25 mL of ethanol and 25 mL of deionized water, stirred evenly, and placed in a reaction vessel. The reaction was carried out at 110 °C for 24 h. After centrifugation, washing with ethanol and vacuum drying, Cu-BTC matrix material was obtained.

[0053] Introduction of cuprous active sites: Cuprous ions were introduced using a liquid-phase impregnation method. The dried matrix material was weighed and dispersed in anhydrous acetonitrile, and 1.0 g of cuprous chloride was added. The mixture was stirred and impregnated for 12 to 24 hours under nitrogen protection. The acetonitrile was then used to target the active sites. The stabilizing effect prevents oxidation; after impregnation, filter and wash three times with fresh acetonitrile to remove residual free copper salts on the surface;

[0054] Activation treatment: Finally, the material is placed in a vacuum activation furnace and activated at a gradient temperature of 160℃ to 200℃ for 12 hours to remove acetonitrile solvent molecules from the pores, thereby exposing the coordinate-unsaturated... Active sites; This embodiment successfully utilized the above-described post-synthetic modification strategy to achieve the desired results. With sulfur atoms The complexation mechanism significantly enhances the adsorbent's specific recognition ability for trace sulfur and prevents the loss of active components during the recycling process.

[0055] Example 3

[0056] This embodiment provides a simulation screening and process optimization method for anhydrous hydrogen fluoride desulfurization adsorbents, focusing on process verification under medium pressure conditions.

[0057] In this embodiment, a threshold of 4.0 Å is set to balance specific surface area and mass transfer resistance, and to screen out material structures suitable for medium flow rate conditions.

[0058] In this embodiment, the interaction energy is calculated using GGA functionals, which provides sufficiently accurate potential energy surface data for force field correction while ensuring computational efficiency.

[0059] In this embodiment, for the medium pressure condition of 313K and 2.5bar, the number of GCMC steps was increased to ensure statistical convergence in a high-density fluid environment;

[0060] This step continues to use the electronic-level refinement strategy to ensure that the microscopic electronic properties of the screened materials meet the adsorption requirements.

[0061] In this embodiment, the amino coverage is increased to 30% to cope with higher concentrations of impurities by increasing the density of active sites, while minimizing steric hindrance energy through geometric optimization;

[0062] In this embodiment, the accuracy of the force field parameter fitting method in handling complex solvent effects was verified by simulation, which is particularly instructive for predicting the mass transfer zone length under medium pressure conditions.

[0063] The synthesis process in this embodiment is further optimized. Specifically, for the screened MIL-101(Cr) derivative, 4.00 g (10 mmol) of chromium nitrate nonahydrate and 3.32 g (20 mmol) of terephthalic acid were weighed, dispersed in 80 mL of deionized water, and a small amount of hydrofluoric acid was added to form a high-concentration hydrothermal reaction precursor. By extending the reaction time, the integrity of the crystals was improved, which met the requirements for higher cycle stability.

[0064] Example 4

[0065] This embodiment provides a simulation screening and process optimization method for anhydrous hydrogen fluoride desulfurization adsorbents, focusing on performance evaluation under high temperature and high pressure conditions;

[0066] In this embodiment, the pore size threshold is further increased to 4.5 Å to meet the requirements of rapid diffusion of gas molecules under high temperature and high pressure.

[0067] In this embodiment, the B3LYP hybrid functional is selected for calculation. Although the computational cost is high, it provides more accurate electronic structure information and is suitable for scenarios with extremely high requirements for force field parameters.

[0068] In this embodiment, the simulation conditions of 323K and 3.5bar are simulated, and the number of simulation steps is greatly increased to 10 million to overcome the statistical noise caused by the decrease in adsorption probability at high temperature.

[0069] This step ensures that the material maintains stable chemisorption characteristics even under harsh operating conditions;

[0070] In this embodiment, Coverage increased to 40%, aiming to create a high-density [structure / system]. Complexing networks were used to verify the stability and adsorption performance of the high-coverage modification strategy under high temperature and high pressure conditions.

[0071] In this embodiment, the breakthrough curve at high flow rates was accurately predicted by coupling the Toth model with the non-isothermal fixed bed model, confirming the industrial application potential of the material under harsh conditions.

[0072] This embodiment employs more stringent synthesis and activation conditions. Specifically, for the screened modified UiO-67 material, 2.70 g (10 mmol) was weighed. 5.40 g (30 mmol) of biphenyl dicarboxylic acid ligand was dissolved in a mixture of 300 mL LDM and 10 mL acetic acid, with acetic acid acting as a regulator to control the crystal nucleus growth rate to ensure the rigidity of the framework structure under high load modification.

[0073] Example 5

[0074] This embodiment provides a simulation screening and process optimization method for anhydrous hydrogen fluoride desulfurization adsorbents, aiming to explore the ultimate processing capacity under extreme operating conditions;

[0075] In this embodiment, a wide aperture threshold of 5.0 Å is set to minimize mass transfer resistance and meet the needs of high-throughput processing.

[0076] In this embodiment, the dielectric constant is set to 84 to simulate an extreme high polarity environment and test the robustness boundary of the force field correction method.

[0077] In this embodiment, an extremely high-intensity GCMC simulation was performed for the extreme operating conditions of 333K and 5.0 bar to ensure the reliability of the data under extremely low adsorption probability.

[0078] This step verifies the electronic structure stability of the material under extreme conditions;

[0079] In this embodiment, an ultra-high amino coverage of 50% was attempted to explore the limiting balance between the steric hindrance of the modified group and the number of adsorption sites.

[0080] In this embodiment, the extreme processing capacity and long cycle life of the adsorbent under extreme operating conditions were simulated and verified, providing safety boundary data for industrial scale-up;

[0081] The synthesis process in this embodiment is the most stringent. In order to achieve a high amino coverage of 50%, 2.33g (10mmol) was weighed out during the synthesis stage. 9.05 g (50 mmol) of 2-aminoterephthalic acid was dissolved in 500 mL of high-purity DMF. The extremely high ligand excess ratio (1:5) was intended to promote coordination saturation of defect sites, thereby preparing an adsorbent framework with the highest thermal and chemical stability.

[0082] Comparative Example 1

[0083] This comparative example provides a conventional adsorbent screening method. Compared with Example 2, this comparative example does not use an implicit solvent model or force field parameter correction in step S2, but directly uses a universal UFF force field for GCMC simulation. In step S3, the competitive adsorption scenario of HF and sulfur impurities is not set, and only single-component gas adsorption simulation is performed. The remaining steps are consistent with Example 2. This comparative example aims to verify the necessity of force field correction and competitive adsorption simulation for specific solvent environments, and to reveal the failure mechanism of universal force fields in strongly polar solvent systems.

[0084] Comparative Example 2

[0085] This comparative example provides an adsorbent screening method. Compared with Example 2, this comparative example does not include an active learning enhancement step between steps S3 and S4, but instead performs full GCMC simulation calculations on all materials in the basic screening library. The remaining steps are consistent with Example 2. This comparative example aims to verify the effect of introducing Gaussian process regression or neural network models on improving screening efficiency and to evaluate the time cost difference between full calculation and intelligent sampling.

[0086] Comparative Example 3

[0087] This comparative example provides an adsorbent screening method. Compared with Example 2, this comparative example does not establish a non-isothermal and non-adiabatic fixed-bed adsorber model in step S6, but only estimates the breakthrough time based on the microscopic adsorption amount, without coupling the influence of adsorption heat on the bed temperature distribution. The remaining steps are consistent with Example 2. This comparative example aims to verify the importance of thermal effect coupling in macroscopic process simulation and reveal the breakthrough time prediction deviation caused by ignoring adsorption heat.

[0088] To further illustrate the beneficial effects of the technical solution of the present invention, the performance of the above embodiments 1-5 and comparative examples 1-3 was compared under the same or corresponding test conditions.

[0089] 1. HF / sulfide selectivity simulation calculation: Based on GCMC simulation results, the mole fraction ratio of sulfur impurities in the adsorbed phase and the gas phase is calculated using the following formula:

[0090]

[0091] and These represent the molar fractions of sulfur impurities and hydrogen fluoride in the adsorbed phase, respectively.

[0092] and These represent the mole fractions of sulfur impurities and hydrogen fluoride in the gas phase, respectively.

[0093] 2. Transmission time prediction error verification: Compare the simulated predicted transmission time with the experimentally measured value; the specific experimental conditions are: using the inner diameter... A 10mm thick, 150mm long stainless steel micro-fixed bed adsorption column was used, with an adsorbent loading of 2.0g and quartz sand filling both ends for fixation. The test gas flow rate was controlled at 50mL / min (standard conditions), and the inlet sulfur concentration was 10ppm (within the range of 10 ... (For balancing gas), the operating temperature is consistent with the simulation temperature of each embodiment; the breakthrough point is set at the moment when the outlet sulfur concentration reaches 0.1 ppm;

[0094] 3. Total computation time statistics: Statistics on the CPU time required for the entire process from the establishment of the screening library to the completion of process optimization, measured in hours per core;

[0095] 4. Working capacity prediction: Calculate the difference in adsorption capacity under adsorption pressure and desorption pressure.

[0096] For each embodiment and comparative example, a simulation process was executed on a high-performance computing cluster. Each data point was calculated three times and the average value was taken to eliminate statistical errors. For the experimental verification part, after the corresponding materials were synthesized, a breakthrough experiment was carried out on a self-built micro fixed bed evaluation device, and the outlet sulfur content was analyzed online using a gas chromatograph.

[0097] Table 1

[0098]

[0099] As can be seen from the data analysis in Table 1, the simulation screening and process optimization method for anhydrous hydrogen fluoride desulfurization adsorbents provided by this invention shows significant advantages in both selectivity prediction accuracy and computational efficiency. With the introduction of force field correction strategy and active learning algorithm, the overall efficiency of the screening process has achieved a qualitative leap.

[0100] Specifically, comparing Example 2 and Comparative Example 1, it can be seen that Example 2, by introducing a force field correction and implicit solvent model for the HF environment, achieves a selectivity as high as 2100 with a breakthrough time prediction error of only 2.8%. In contrast, Comparative Example 1, due to the use of a general force field and the neglect of solvent competition effects, results in extremely low predicted selectivity with an error of up to 45.6% compared to the experimental value. This huge difference confirms from a mechanistic perspective that in a strongly polar solvent environment, solvent molecules form a dense solvation layer on the adsorption framework surface, severely interfering with the approach and adsorption of impurity molecules. The force field correction strategy of this invention effectively captures this microscopic solvation effect, thereby solving the technical pain point that conventional methods cannot accurately describe the adsorption behavior of complex systems.

[0101] Comparing Example 2 with Comparative Example 2, it can be seen that the two are consistent in the prediction results of selectivity and working capacity. However, Example 2 introduces an active learning efficiency enhancement step, which significantly reduces the total computation time from 380 hours in Comparative Example 2 to 52 hours, improving efficiency by more than 7 times. This shows that the introduction of Gaussian process regression or neural network model can avoid a large number of low-confidence invalid calculations through intelligent sampling strategy while ensuring screening accuracy, significantly reducing the computation cost and realizing high-throughput screening of massive materials.

[0102] Comparing Example 2 and Comparative Example 3, it can be seen that although Comparative Example 3 is consistent with Example 2 in terms of micromaterial screening, the breakthrough time prediction error reaches 28.4% because the adsorption heat effect was not coupled during the process optimization stage. This indicates that the exothermic effect in the HF adsorption process will significantly change the temperature distribution of the bed and thus affect the adsorption equilibrium. The micro-macro cross-scale coupling model established in this invention can more realistically reflect the dynamic heat and mass transfer behavior in the industrial adsorption tower and provide more reliable parameter support for actual process design.

[0103] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbents, characterized in that, Includes the following steps: S1. Construct a database of porous material crystal structures, perform geometric topological analysis on the material models in the database, calculate the pore size limit diameter and the maximum pore diameter, set the pore size screening threshold to 3.0 Å to 5.0 Å, remove structures with pore size limit diameters smaller than this threshold that do not meet the requirements of sulfide molecular dynamics diameter, and establish a basic screening library. S2. A smart screening engine is built for anhydrous hydrogen fluoride solvent environment. The interaction energy between HF molecules and framework atoms is calculated at the density functional theory level using generalized gradient approximation or hybrid functionals. Lennard-Jones potential parameters are fitted to correct the general force field, or an implicit solvent model with dielectric constant set to 50-84 is introduced to correct the force field parameters of materials in the basic screening library in order to capture the high polarity solvent effect. S3. Under the modified force field environment, with temperatures ranging from 273K to 333K and pressures from 0.1 bar to 5.0 bar, a competitive adsorption scenario of anhydrous hydrogen fluoride and sulfur impurities is constructed, with molar ratios ranging from 1000:1 to 10000:

1. High-throughput grand canonical Monte Carlo simulations are performed under a grand canonical ensemble μVT, with both the equilibrium step number and the production step number set to [value missing]. to Next, the adsorption selectivity, Henry's law coefficient and working capacity index were calculated and extracted to screen out the top 5% to 10% of primary candidate materials. S4. Perform electronic-level refinement on the primary candidate material set, calculate the change in electron cloud density of adsorption sites using density functional theory, calculate the differential charge density through Bader charge or Mulliken charge analysis, set an adsorption energy threshold to determine whether a chemisorption mechanism exists, and select the core candidate materials. S5. Based on the structure-activity relationship, a modification strategy is implemented to generate core candidate materials. Functional groups are introduced at the coordinate unsaturated metal sites or organic ligand side groups of the framework according to the geometric constraints of bond lengths of 1.5 Å to 2.5 Å. The performance is then verified by high-throughput giant canonical Monte Carlo and density functional theory joint simulation to determine the optimal adsorbent model. S6. Transform the microscopic adsorption data of the optimal adsorbent model into macroscopic process parameters, fit a multi-parameter adsorption isotherm model, and import it into the macroscopic process simulator. Calculate the mass transfer coefficient based on the linear driving force model, perform breakthrough curve and cycle performance prediction, and complete process optimization.

2. The method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbent according to claim 1, characterized in that, In step S2, when constructing the intelligent screening engine, the variance of the electrostatic potential distribution of the porous material framework, the energy of local sulfur affinity sites, and the hydrogen fluoride-framework interaction energy are extracted as the core descriptor set for subsequent feature mapping and model training.

3. The method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbent according to claim 1, characterized in that, Between steps S3 and S4, there is also an active learning enhancement step, which constructs a Gaussian process regression or neural network model, trains the model using a small amount of high-precision computational data, quickly predicts the properties of massive materials, and automatically iterates the model through uncertainty sampling to predict materials with low confidence.

4. The method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbent according to claim 1, characterized in that, In step S5, the specific operation of introducing functional groups is as follows: virtually grafting amino groups onto the organic linkers of the material skeleton or introducing cuprous ions into open metal sites, setting the coverage of the modified groups to 10% to 50%, and adjusting the spatial orientation of the groups through a geometric optimization algorithm to minimize steric hindrance energy and enhance the specific adsorption of sulfur impurities.

5. The method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbent according to claim 1, characterized in that, In step S6, the adsorption data obtained from the microscopic simulation are fitted to the parameters of the Langmuir model or the Toth model and imported into the breakthrough curve simulator as an input file.

6. The method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbent according to claim 1, characterized in that, In step S6, when predicting the breakthrough curve and cycle performance, a non-isothermal and non-adiabatic fixed-bed adsorber model is established. The influence of adsorption heat on the temperature distribution of the adsorption tower bed is coupled. The dynamic response of apparent gas velocity in the range of 0.5 cm / s to 10.0 cm / s, feed temperature in the range of -20℃ to 50℃, and bed height in the range of 0.5 m to 5.0 m to breakthrough time and mass transfer zone length is simulated and investigated. The structural stability after 50 to 500 adsorption-desorption cycles is also predicted.

7. The method for simulation screening and process optimization of anhydrous hydrogen fluoride desulfurization adsorbent according to claim 1, characterized in that, It also includes experimental verification steps, selecting the top 3 to 5 materials in the calculation for synthesis and small-scale testing; The synthesis steps include: dissolving the metal precursor salt and organic ligand in a molar ratio of 1:0.5 to 1:5 in a polar solvent, ultrasonically dispersing for 10 min to 60 min, placing in a high-pressure reactor and carrying out a solvothermal reaction at 80℃ to 180℃ for 12 h to 72 h, and after the reaction, gradually cooling at a rate of 1℃ / min to 5℃ / min, separating the product by centrifugation, performing 3 to 10 solvent exchanges with a low-boiling-point solvent, and finally performing gradient heating and vacuum activation at 100℃ to 250℃ for 4 h to 24 h to remove solvent molecules from the pores; Small-scale trials include comparing the penetration curves measured in experiments with the simulated predicted curves, and verifying the reliability of the screening process by plotting parity check diagrams.

Images