A Simulation Method for Synchronous Evolution of Particle Agglomeration and Coating in Spray Fluidized Bed

By employing a synchronous simulation method for particle agglomeration and coating in spray fluidized beds, coupled with droplet drying and wet collision machine learning for discrimination, the problem of synchronous evolution of agglomerate morphology and coating layer distribution in spray fluidized beds is solved, achieving efficient physical consistency simulation and computational optimization.

CN123116828APending Publication Date: 2026-11-17JIANGSU UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611330872.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing spray fluidized bed particle forming technology, when simulating the coating and agglomeration process in a spray fluidized bed, is difficult to reflect the synchronous evolution characteristics of agglomerate morphology and coating layer distribution in actual processes. In addition, it requires a large amount of computation or is difficult to uniformly describe the residual volume of droplets, dry and wet states, and obstruction areas.

Method used

A synchronous simulation method for particle aggregation and coating is adopted, which combines droplet drying, wet collision machine learning discrimination and Monte Carlo event-driven particle aggregation. By establishing a multi-layer state representation system, including the spatial coordinate matrix of the aggregate, the volume matrix of the droplet, and the thickness matrix of the coating layer, the aggregate structure and coating layer thickness are dynamically updated. Combined with the machine learning model, the collision results are predicted, thereby reducing the computational cost.

Benefits of technology

This improved the physical consistency of spray fluidized bed agglomeration-coating simulation, reduced computational overhead, and enabled simultaneous evaluation of agglomerate morphology and coating morphology.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116828A_ABST
    Figure CN123116828A_ABST
Patent Text Reader

Abstract

This invention discloses a synchronous evolution simulation method for particle agglomeration-coating in a spray fluidized bed. Using agglomerate number k, initial particle number i, and particle surface deposition position number j as unified indices, a multi-layer state characterization system is established, consisting of the agglomerate spatial coordinate matrix AggCell, the droplet current volume matrix DVCell, the droplet maximum volume matrix DVCellMax, the droplet deposition position reachability flag matrix DVCellFlag, the coating layer thickness matrix LTCell, and the coating layer number matrix LNCell. Coupled with droplet drying, wet collision machine learning discrimination, and Monte Carlo event-driven synchronous simulation of particle agglomeration-coating, the results of particle agglomeration, breakup, and non-agglomeration dynamically change with the drying process, improving the physical consistency of the spray fluidized bed agglomeration-coating simulation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of simulation technology for particle coating and agglomeration behavior in spray fluidized beds, and in particular to a synchronous simulation method for particle agglomeration-coating that couples droplet drying, wet collision machine learning discrimination, and Monte Carlo event-driven processes. Background Technology

[0002] Spray fluidized bed particle forming technology enables particle surface coating, agglomeration granulation, and structural control, and is widely used in the preparation of functional particulate materials. Especially in the preparation of energy particles such as lithium battery positive and negative electrode materials and solid electrolytes, particle size distribution, agglomeration morphology, and coating uniformity significantly affect the processing performance and durability of the materials. Therefore, higher requirements are placed on the synergistic control of coating and agglomeration processes in spray fluidized beds. During the formation of spray fluidized bed particles, atomized droplets are deposited, spread, dried, and form a coating layer on the particle surface. Simultaneously, wet particles may agglomerate due to liquid bridging adhesion, and the formed agglomerates may break up in subsequent collisions. There is a significant nonlinear coupling between the droplet drying state, particle collision results, and coating morphology. Calculating the coating or agglomeration process separately is insufficient to reflect the synchronous evolution of agglomerate morphology and coating layer distribution in actual processes.

[0003] Existing methods typically employ the discrete element method to directly simulate wet particle collisions or use empirical rules to describe droplet deposition and drying. The former involves significant computational costs in large-scale Monte Carlo processes, while the latter struggles to integrate local droplet residual volume, wet / dry state, shading region, and post-fragmentation state into a single data structure.

[0004] Therefore, it is necessary to establish a simulation method that can synchronously update the aggregate structure, droplet residual volume, and coating thickness in event-driven simulations. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a synchronous evolution simulation method for particle agglomeration-coating in spray fluidized beds. This method couples droplet drying, wet collision machine learning discrimination, and Monte Carlo event-driven synchronous simulation of particle agglomeration-coating, enabling the results of particle agglomeration, breakup, and non-agglomeration to dynamically change with the drying process, thereby improving the physical consistency of the spray fluidized bed agglomeration-coating simulation.

[0006] To achieve the technical solution of this invention, the technical solution adopted by this invention is as follows:

[0007] A simulation method for the synchronous evolution of particle agglomeration and coating in a spray fluidized bed includes the following steps:

[0008] Step S1: Set up and initialize state variables, including but not limited to spray fluidized bed, droplets, particles and Monte Carlo simulation;

[0009] Using aggregate number k, initial particle number i, and particle surface deposition location number j as unified indexes, a multi-layer state characterization system is established, consisting of aggregate spatial coordinate matrix AggCell, droplet current volume matrix DVCell, droplet maximum volume matrix DVCellMax, droplet deposition location reachability flag matrix DVCellFlag, coating layer thickness matrix LTCell, and coating layer number matrix LNCell.

[0010] Step S2: Randomly select events and advance Monte Carlo physical time. Based on the scaling relationship between the real system and the MC system, calculate the droplet deposition event frequency and particle collision event frequency in the current MC system;

[0011] Step S3: If the current event is a particle collision event, determine the collision result based on the pre-trained wet collision prediction model; if the current event is a droplet deposition event, randomly determine the deposition location according to DVCellFlag and update DVCell and DVCellMax.

[0012] Step S3.1: Train machine learning models for predicting different collision events respectively, using collision parameters as input features of the machine learning models and collision results as output labels; the collision events include: single-particle-single-particle wet collision, single-particle-two-particle aggregate wet collision, and single-particle-multi-particle wet collision.

[0013] If the prediction result of single-particle wet collision is "agglomeration", the center coordinates of the newly added particles are calculated based on the center coordinates of the collided particles and the direction of the collision contact point, and AggCell, DVCell, DVCellMax, DVCellFlag, LTCell, and LNCell are updated simultaneously. If the prediction result is "no agglomeration", the current agglomerate structure remains unchanged, and the volume of the residual droplet at the collision contact position is updated for liquid bridge loss.

[0014] For wet collisions between single-particle and dual-particle aggregates or between single-particle and multi-particle aggregates, if the prediction result is "aggregation", the coordinates of the newly added particle are appended to AggCell{k}, and the corresponding rows in DVCell, DVCellMax, DVCellFlag, LTCell, and LNCell are appended simultaneously. At the same time, the occlusion area is updated based on the newly added contact point. If the prediction result is "no aggregation", no new particle connection relationship is added, and the aggregate spatial coordinate matrix AggCell remains unchanged. At the same time, it is assumed that the liquid bridge in the contact area has suffered partial loss, and the volume of the residual droplet at the contact position is updated by attenuation.

[0015] The wet collision of single-particle and multi-particle aggregates is approximately the collision between a single particle and a local two-particle aggregate of a complete aggregate, and the local collision state is characterized by selected contacting particles and their adjacent structures.

[0016] Step S3.2: When a droplet deposition event occurs, randomly select an aggregate from the current MC system, and randomly select an initial particle from the aggregate; then, according to the droplet deposition reachability flag matrix DVCellFlag on the surface of the initial particle, randomly select a position from the deposition positions that satisfy DVCellFlag{k}(i,j)=1 as the current droplet deposition position.

[0017] Step S4. After executing the collision event and droplet deposition event in step S3, based on the updated AggCell, DVCell, DVCellMax, DVCellFlag, LTCell and LNCell, the droplet drying state, aggregate morphology, fragmentation inheritance state and coating morphology are updated synchronously.

[0018] Step S5: After completing the preset MC physical time or event steps, output the aggregate morphology parameters and coating morphology parameters; the aggregate morphology parameters include, but are not limited to, the initial number of particles, the radius of rotation, the adjacency relationship, the number of fragmentation times, and the particle size distribution; the coating morphology parameters include, but are not limited to, the average coating thickness, the coating coverage rate, the occlusion rate, and the coating thickness variation coefficient COV.

[0019] Furthermore, in step S2, the physical time is calculated. The process is as follows:

[0020] Step S2.1. Calculate the initial total number of particles in the real system, denoted as: ,in, For bed quality, Let the mass of a single initial particle be denoted as . , The initial particle density, The initial particle diameter;

[0021] Step S2.2. Introduce the system scaling factor S, denoted as: , This represents the initial total number of particles in the current MC system;

[0022] Step S2.3. After considering system scaling, the frequency of droplet deposition events in the MC system is denoted as: = S(t), Let the droplet number flow rate in the real system be denoted as . , For spray mass flow rate, The density is the droplet density. The diameter of the droplet;

[0023] Step S2.4. The collision event frequency in the MC system is denoted as... , denoted as: , The initial collision frequency of a single particle. This represents the initial total number of particles in the current MC system;

[0024] Step S2.5. Calculate the total event frequency of the current MC step. , recorded as = + ,

[0025] Step S2.6. Generate two independent uniformly distributed random numbers between 0 and 1. and , Used to determine the type of event that occurred in the current MC step. Used to calculate the physical time of the event's progression; when: If the current event is a particle collision event, then the current event is a droplet deposition event; otherwise, the current event is a droplet deposition event.

[0026] Step S2.7. Physical time of the current event progression Obtained by random sampling according to an exponential distribution, its calculation formula is:

[0027] ;

[0028] Then update MC physics time: .

[0029] Furthermore, the pre-training of the wet collision prediction model:

[0030] 3.1.1. Establish a wet particle collision simulation dataset based on the discrete element method (DEM); use the collision parameters obtained from the discrete element simulation as input features of the machine learning model, and use the collision results as output labels to obtain the DEM collision dataset;

[0031] 3.1.2. Select a machine learning model for classification, train the model using the DEM collision dataset, and obtain a wet collision prediction model.

[0032] Furthermore, let the liquid bridge retention coefficient be... ,in Then the volume of the residual droplet at the contact point after the collision is updated to .

[0033] Furthermore, the droplet deposition process is denoted as follows: Let the selected aggregate be numbered k, the selected initial particle be numbered i, the selected deposition location be numbered j, and the initial volume of a single droplet be... Then the accumulated droplet is added to DVCell{k}(i,j), that is, DVCell{k}(i,j) = DVCell{k}(i,j) + And simultaneously update DVCellMax{k}(i,j)=max[DVCellMax{k}(i,j),DVCell{k}(i,j)]; if the selected particle has no deposition location, then reselect a particle or record this deposition event as an empty event.

[0034] Furthermore, the droplet drying and residual volume renewal process in step S4:

[0035] For droplets deposited on the particle surface, the contact diameter between the droplet and the particle surface during the drying process... The droplet height and surface area exposed to the gas phase remain constant, while the residual droplet volume gradually decreases; assuming the contact radius... = / 2, at a certain moment the residual volume of the droplet is Then the height h of the spherical crown satisfies:

[0036] ;

[0037] Surface area of ​​droplets exposed to the gas phase Represented as:

[0038] ;

[0039] During the MC physical time progression, for the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate, the current residual droplet volume is recorded by DVCell{k}(i,j); let the residual droplet mass and volume at the n-th time step be respectively... and ,in =DVCell{k}(i,j), then the time step is... The residual mass and volume of the droplet are updated as follows:

[0040] ;

[0041] ;

[0042] The residual droplet volume at the next time step is calculated using the above recursive method. Then, update DVCell{k}(i,j) to Subsequently, based on the updated DVCell{k}(i,j) and DVCellMax{k}(i,j), it is determined whether the current deposition drying cycle at this deposition location has been completed.

[0043] Let the threshold for determining drying completion be... , where 0 < <1; when DVCellMax{k}(i,j)>0 and satisfies DVCell{k}(i,j)≤ When DVCellMax{k}(i,j), it is considered that the droplets at that deposition location have completed drying within the current deposition drying cycle.

[0044] Furthermore, the morphology of the aggregates is represented as follows:

[0045] Calculate the k-radius of gyration of the aggregates based on the initial particle position information from AggCell.

[0046] ;

[0047] In the formula and Let represent the position vectors of the i-th and j-th initial particles relative to the centroid of the aggregate, respectively.

[0048] Furthermore, the coating thickness is expressed as:

[0049] After drying at a certain deposition location, the current coating layer number l = LNCell{k}(i,j) is read; the thickness of the single-layer coating is calculated based on the maximum droplet volume recorded at that location during this deposition-drying cycle, DVCellMax{k}(i,j); let the maximum droplet volume at the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate during the current deposition-drying cycle be... The mass fraction of solids in the spray liquid is After drying, the porosity of the coating layer is The volume of the coating layer formed at that location after this drying process is... Represented as:

[0050] ;

[0051] Assuming the volume is the same at each deposition location, the total volume of the outer layer covering particle i is... The thickness of the coating layer formed in this cycle was calculated. :

[0052] ;

[0053] Write the coating thickness into LTCell{k}(i,j,l+1); then update LNCell{k}(i,j)=l+1, and reset DVCell{k}(i,j) and DVCellMax{k}(i,j) to zero, and enter the next deposition-drying cycle.

[0054] Furthermore, the uniformity of coating thickness was determined using the coefficient of variation. Where h is the set of local coverage thicknesses participating in the statistics, For thickness standard deviation, The average thickness; from this, the average coating coverage and coefficient of variation (COV) of the coating thickness for a single initial particle, a single agglomerate, and all agglomerates can be calculated; the coating coverage of the j-th particle surface is determined by the number of droplet deposition sites. and the number of locations where deposited droplets have been received To calculate: .

[0055] Beneficial effects

[0056] 1. By incorporating the residual volume of droplets and the drying state into the wet collision result prediction model, the particle agglomeration, breakup and non-agglomeration results are dynamically changed with the drying process, which improves the physical consistency of the spray fluidized bed agglomeration-coating simulation.

[0057] 2. By using a machine learning classification model trained on DEM data to replace the direct DEM calculation for each collision, the computational overhead in event-driven MC simulation is reduced, and the efficiency of large-sample aggregate evolution simulation is improved.

[0058] 3. By using a unified index structure of AggCell, DVCell, DVCellMax, DVCellFlag, LTCell, and LNCell, the synchronous updating of agglomeration, fragmentation, droplet deposition, occlusion area, and coating thickness is achieved, which facilitates the output of evaluation indicators for agglomerate morphology and coating morphology. Attached Figure Description

[0059] Figure 1 This is a flowchart of the overall simulation method for simultaneous particle agglomeration-coating in a spray fluidized bed, which couples droplet drying with machine learning-based wet collision discrimination.

[0060] Figure 2 This is a schematic diagram of the occlusion effect.

[0061] Figure 3 This is a schematic diagram of the matrix structure of AggCell, DVCell, and LTCell.

[0062] Figure 4 This is a schematic diagram of wet collision parameters between single particles.

[0063] Figure 5 This is a schematic diagram of wet collision parameters for single-particle and double-particle aggregates.

[0064] Figure 6 This is a schematic diagram illustrating the performance of the prediction model for single-particle-double-particle wet collision results.

[0065] Figure 7 This is a schematic diagram illustrating the fragmentation and connectivity of aggregates.

[0066] Figure 8 This is a diagram showing the evolution of the radius of gyration of the aggregates.

[0067] Figure 9 This represents the evolution of aggregate structure and surface coating morphology. Detailed Implementation

[0068] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the scope of protection of the present invention.

[0069] This invention discloses a simulation method for the synchronous evolution of particle agglomeration and coating in a spray fluidized bed, comprising the following steps:

[0070] Step S1: Parameter setting and state variable initialization.

[0071] In one specific embodiment, the basic parameters of the spray fluidized bed, droplets, particles, and Monte Carlo simulation are first set, and a data structure is established for synchronously recording the aggregate structure, droplet residual volume, occlusion area, and coating thickness.

[0072] 1. Parameter settings

[0073] (1) The gas phase and fluidized bed parameters are set as follows: bed mass is 0.45 kg, fluidizing gas inlet temperature is 80 °C, inlet moisture content is 1 g / kg dry gas, and fluidizing gas mass flow rate is 39 kg / h. In the droplet drying calculation, the gas density is taken as 1.2 kg / m³, and the gas dynamic viscosity is taken as 1.84 × 10⁻ 5 Pa·s, the diffusion coefficient of water vapor in air is taken as 2.5 × 10⁻ 5 m² / s.

[0074] (2) The droplet and spray parameters are set as follows: the spray liquid is 5% HPMC solution, the spray mass flow rate is 0.4 kg / h, the droplet diameter is 14 μm, the droplet density is 1097 kg / m³, the contact angle between the droplet and the particle surface is 40°, the porosity of the coating layer after drying is 30%, and the solid content is 5%.

[0075] (3) The particle parameters are set as follows: the initial particle diameter is 140 μm, the particle density is 900 kg / m³ according to the example, and the particle collision frequency is... The value is 10; the number of MC simulated aggregate samples is 1000.

[0076] 2. Establish a multi-layered state representation data structure

[0077] A multi-layer state characterization system was established, consisting of the aggregate spatial coordinate matrix AggCell, the droplet current volume matrix DVCell, the droplet maximum volume matrix DVCellMax, the droplet deposition location reachability flag matrix DVCellFlag, the coating layer thickness matrix LTCell, and the coating layer number matrix LNCell. This system uses the aggregate number k, the initial particle number i, and the particle surface deposition location number j as unified indices, enabling the synchronous updating of aggregate structure evolution, droplet deposition distribution, shading region, and layer-by-layer coating layer thickness changes during MC simulations.

[0078] AggCell is defined as having a size of A cell array of size 1. This represents the total number of elements in AggCell. The k-th cell element, AggCell{k}, stores the spatial coordinates of all initial particles in the k-th aggregate, with dimensions [missing information]. ×3. Among them, This represents the initial number of particles contained in the k-th aggregate. The three columns of the matrix correspond to the x, y, and z spatial coordinates of the initial particles, respectively. Initially... =1.

[0079] DVCell is defined as having a size of A cell array of size ×1. The k-th cell element DVCell{k} stores the current residual droplet volume at each deposition position on the surface of all initial particles in the k-th aggregate, and its dimension is 1. × .in, The total number of deposition sites on the surface of each particle.

[0080] DVCellMax and DVCell have the same index structure, where DVCellMax{k}(i,j) represents the maximum droplet volume reached at the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate during the current deposition drying cycle. When a droplet deposition event increases DVCell{k}(i,j), DVCellMax{k}(i,j) = max[DVCellMax{k}(i,j), DVCell{k}(i,j)] is updated synchronously. When the drying completion criterion is met at this location, DVCell{k}(i,j) and DVCellMax{k}(i,j) are reset to zero, and the next deposition drying cycle begins.

[0081] DVCellFlag has the same index structure as DVCell, where DVCellFlag{k}(i,j) indicates whether the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate allows droplet deposition. When DVCellFlag{k}(i,j)=1, the location is a depositable location; when DVCellFlag{k}(i,j)=0, the location is a non-depositable location. Subsequent droplet deposition events are randomly selected only from the deposition locations that satisfy DVCellFlag{k}(i,j)=1.

[0082] Let the center coordinates of the i-th initial particle be... Particle radius is The unit direction vector corresponding to the j-th surface deposition location is Then the spatial coordinates of the deposition location can be expressed as: When the i-th initial particle agglomerates with the m-th initial particle, their contact direction is... The contact point on the surface of the i-th particle is .like If the deposition location is within the shading area adjacent to the agglomeration contact point, then DVCellFlag{k}(i,j) is set to 0. The shading area of ​​the m-th particle surface is determined by... The same method was used to determine it.

[0083] Half corner The area of ​​the equivalent spherical cap to be shielded can be determined based on the required shielding area. If it is desired that each contact point will be shielded approximately... If there are equal-area deposition sites, then let , can be obtained In this embodiment, =109; if the equivalent shading is based on the area of ​​a single depositional location, then ≈11°, if you want to block about 1 / 6 of the sphere, you can take... ≈48°. The actual value can be determined based on the contact area size and shading range. When the agglomerates break up or the contact relationship changes, the DVCellFlag of the corresponding surface deposition location is recalculated based on the updated particle adjacency relationship.

[0084] LTCell is defined as having a size of A cell array of size ×1. The k-th cell element LTCell{k} stores the layer-by-layer coating thickness at each deposition location on the surface of all initial particles in the k-th aggregate, with dimensions of ×1. × .in, This indicates the maximum number of recording layers in the covering layer.

[0085] LNCell and DVCell share the same two-dimensional index structure, where LNCell{k}(i,j) represents the number of coating layers currently formed at the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate. Initially, LNCell{k}(i,j) = 0. When a location meets the drying completion criterion, the current layer number l = LNCell{k}(i,j) is read, and the thickness of the newly formed coating layer is calculated based on DVCellMax{k}(i,j). Write LTCell{k}(i,j,l+1) to LNCell{k}(i,j) and then update LNCell{k}(i,j)=l+1, and set DVCell{k}(i,j) and DVCellMax{k}(i,j) to zero.

[0086] Combined with appendix Figure 3 The diagram shows the matrix structure of AggCell, DVCell, and LTCell, while DVCellMax, DVCellFlag, and LNCell have the same structure as DVCell, only the stored data is different.

[0087] Before the MC simulation begins, the total number of droplet deposition sites on the surface of each initial particle is determined based on the initial particle size and the contact area after individual droplet deposition. The initial particles are approximately spherical, and their diameter is denoted as . Surface area of ​​a single initial particle Calculate using the following formula:

[0088]

[0089] After being deposited on the particle surface, the droplet approximates a spherical cap shape. Let the diameter of the droplet used for particle surface discretization be... The contact angle between the droplet and the particle surface is The contact diameter of the droplet on the particle surface is then... It can be calculated using the following formula:

[0090]

[0091] area for:

[0092]

[0093] Therefore, the number of droplet deposition sites on each initial particle surface It can be determined by the ratio of the particle surface area to the contact area of ​​a single droplet:

[0094]

[0095] Here, round means taking the nearest integer.

[0096] In this embodiment, the initial particle diameter Take 140 Contact angle Take 40°. If the droplet diameter ratio L = 0.100 is used as the benchmark for particle surface discretization, then the droplet diameter used for surface discretization is:

[0097]

[0098] The droplet contact diameter can be calculated. Approximately 26.80 Surface area of ​​a single initial particle The contact area of ​​a single droplet is approximately 61575 μm². Approximately 564.2 .therefore:

[0099]

[0100] Thus, the surface of each initial particle is divided into 109 droplet deposition sites. Subsequently, the Voronoi subdivision method is used to generate 109 approximately uniformly distributed surface centroids on the spherical particle surface, with each centroid corresponding to a droplet deposition site. The orientation vectors of these 109 surface sites can be pre-saved as a surface mesh file and read and used during the MC simulation.

[0101] Step S2: Randomly select events and advance Monte Carlo (MC) physics time. ;

[0102] Based on the scaling relationship between the real system and the MC system, calculate the droplet deposition event frequency and particle collision event frequency in the current MC system.

[0103] First, the total number of initial particles in the real system is calculated based on the bed mass, initial particle diameter, and particle density. The total number of initial particles in the real system is denoted as... The mass of a single initial particle is denoted as ,but: , .in, For bed quality, The initial particle density, The initial particle diameter is denoted as .

[0104] Since the number of particles in a real system is much larger than the number of samples in a Monte Carlo simulation, a system scaling factor S is introduced. For the t-th Monte Carlo step, the initial total number of particles in the current Monte Carlo system is denoted as [S_t]. The current scaling factor is then expressed as:

[0105]

[0106] in, It is obtained by summing the number of initial particles contained in all aggregates. In the program implementation, if AggCell{k} is the three-dimensional coordinate matrix of the initial particles in the k-th aggregate, then the total number of initial particles in the current MC system can be obtained by summing the number of rows of all AggCell elements.

[0107] The frequency of droplet deposition events is determined by the spray liquid mass flow rate, droplet density, and droplet diameter (distribution). Assuming the droplets are monodisperse spherical, the droplet number flow rate in a real system is denoted as... The formula for its calculation is:

[0108]

[0109] in, The spray mass flow rate is expressed in kg / h. The density is the droplet density. Let be the droplet diameter. Considering system scaling, the frequency of droplet deposition events in the MC system is denoted as . The formula for its calculation is:

[0110] = S(t)

[0111] The frequency of particle collision events is determined based on the number of samples in the MC system and the collision frequency per unit sample. If the initial collision frequency of a single particle is... The frequency of collision events in the MC system is denoted as The formula for its calculation is:

[0112]

[0113] Then calculate the total event frequency of the current MC step. The formula for its calculation is:

[0114] = +

[0115] Generate two independent uniform random numbers between 0 and 1 and .in, Used to determine the type of event that occurred in the current MC step. Used to calculate the physical time of the event's progression. When:

[0116]

[0117] If the event is a particle collision, the current event is considered a droplet deposition event; otherwise, the current event is considered a droplet deposition event. This method ensures that after a large number of MC steps, the proportions of collision events and deposition events approach their respective proportions in the total event frequency.

[0118] Physical time of current event progression Obtained by random sampling according to an exponential distribution, its calculation formula is:

[0119]

[0120] Then update MC physics time:

[0121]

[0122] In this embodiment, for each MC event, only one particle collision or one droplet deposition is performed, and the result is determined according to the event. Advance physical time. To ensure consistency of time logic, before executing this event, first press... The existing residual droplets in the current system are dried and updated; if the current event is a droplet deposition event, the newly added droplets will participate in the subsequent drying and updating from the current moment.

[0123] Step S3: Event Execution

[0124] 3.1 Collision Event Execution

[0125] Pre-training of wet collision prediction models

[0126] In one specific embodiment, a wet particle collision simulation dataset is first established based on the Discrete Element Method (DEM). The surface tension and viscous forces generated by liquid bridges are considered during the wet collision process. By changing parameters such as collision velocity, collision angle, incident angle, and equivalent droplet-particle diameter ratio, particle collision results under different working conditions are obtained. Subsequently, the collision parameters obtained from the DEM simulation are used as input features of a machine learning model, and the collision results are used as output labels to train a collision result prediction model for quickly determining the results of particle collision events in McLeod simulations.

[0127] For single-particle wet collisions, a collision model is constructed between two initial particles. One particle collides with another stationary particle at a set collision velocity. The collision parameters include collision velocity, collision angle, and droplet-to-particle diameter ratio. The collision velocity can be set to 0.005-2 m / s, with intervals of 0.005 m / s; the collision angle can be set to 0° to 90°; and the droplet-to-particle diameter ratio L can be set to 0.010-0.300, with intervals of 0.005. By iteratively changing the above parameters, multiple sets of single-particle wet collision DEM simulations are generated. After each simulation, the collision result is judged based on the final connection state of the two particles after the collision: if the two particles form a stable connection after the collision, it is marked as "agglomeration"; if the two particles separate after the collision, it is marked as "non-agglomeration". For single-particle-to-two-particle agglomerate wet collisions, a collision model is constructed where a single particle collides with a two-particle agglomerate. Before the collision, the spatial relationship between the single particle, the two-particle agglomerate, and the direction of the collision velocity needs to be determined. Input parameters include collision velocity, collision angle, incident angle, and droplet diameter ratio. The collision velocity can be set to a range of 0.005–2 m / s, with intervals of 0.005 m / s; the collision angle can be set to a range of -89° to 89°; the incident angle can be set to a range of 0° to 90°; and the droplet diameter ratio L can be set to a range of 0.010–0.300, with intervals of 0.005. A wet collision DEM simulation dataset of single-particle and two-particle aggregates is generated using the above parameter combinations.

[0128] The initial collision results of wet collisions between single-particle and two-particle aggregates can be categorized into four types: fragmentation, transfer, rebound, and aggregation. Fragmentation indicates that the original two-particle aggregate undergoes structural breakage after the collision; aggregation indicates that the impacting particle forms a new three-particle aggregate with the two-particle aggregate; transfer indicates that the impacting particle forms a new connection with one particle in the aggregate, while the other particle detaches; and rebound indicates that the impacting particle leaves the aggregate after the collision without forming a stable connection. When integrating with MC simulations, "transfer" and "rebound" can be merged into "non-aggregation," therefore the output categories are "aggregation," "fragmentation," and "non-aggregation."

[0129] After obtaining the DEM collision dataset, a weighted K-nearest neighbor model was used to train the collision result prediction model. The model input included collision parameters such as collision velocity, collision angle, incident angle, and droplet diameter ratio; the model output was the collision result category. For the single-particle-single-particle wet collision model, the output categories were "agglomeration" and "non-agglomeration"; for the single-particle-two-particle agglomerate wet collision model, the output categories were "agglomeration," "fragmentation," and "non-agglomeration." During training, Euclidean distance was used as the distance metric between samples, and neighboring samples were weighted using an inverse squared distance weighting method. The number of neighboring samples K could be set to 10.

[0130] To verify the model's predictive ability, the DEM dataset was divided into training and validation sets. For example, a hold-out validation method was used, with 85% of the data used as the training set and 15% as the validation set. After training, the model performance was evaluated using metrics such as confusion matrix, accuracy, true positive rate, and false negative rate. If four original DEM labels were used for evaluation, the numbers 0, 1, 2, and 3 in the figure correspond to fragmentation, migration, bounce, and clustering, respectively. Before the MC simulation, migration and bounce were merged into "non-clustering".

[0131] If a particle collision event occurs, the collided agglomerate k, the initial collided particle i, and the collision contact position j on the particle surface are randomly selected, and the droplet volume at that position is determined accordingly. Determine the collision type: when When >0, it is a wet collision. When the value is 0, it is a dry collision.

[0132] In this embodiment, dry collisions are assumed to be non-agglomerative, meaning the original agglomerate structure remains unchanged, only advancing the MC physics time.

[0133] The droplet diameter ratio required for the wet collision prediction model is calculated from the droplet volume at the current collision location. That is, the equivalent spherical diameter is... The droplet diameter ratio is ,in The initial particle diameter is denoted as . As one of the input parameters for machine learning collision result prediction models.

[0134] For single-particle wet collisions ( =1), the collision velocity is randomly generated in the range of 0.005-2 m / s, and the collision angle is randomly generated in the range of 0° to 90°, and is consistent with... A common input single-particle wet collision prediction model is used. If the prediction result is "agglomeration," the center coordinates of the newly added particle are calculated based on the center coordinates of the collided particle and the direction of the collision contact point, and AggCell, DVCell, DVCellMax, DVCellFlag, LTCell, and LNCell are updated simultaneously. If the prediction result is "no agglomeration," the current agglomerate structure remains unchanged, and the volume of the residual droplet at the collision contact location is updated for liquid bridge loss. Let the liquid bridge retention coefficient be... ,in Then the volume of the residual droplet at the contact point after the collision is updated to In this embodiment, η=0.5.

[0135] For single-particle-double-particle aggregates wet collision ( =2) or single-particle-multi-particle aggregate wet collision ( >2), the collision velocity is randomly generated within the range of 0.005-2 m / s, the collision angle is randomly generated within the range of -89° to 89°, and the incident angle is randomly generated within the range of 0° to 90°, and is consistent with... Common input collision result prediction model.

[0136] For wet collisions between single-particle and multi-particle aggregates, it can be approximated as a collision between a single particle and a local two-particle aggregate within a complete aggregate, with the selected contacting particle and its adjacent structures representing the local collision state. If the prediction result is "aggregation," the coordinates of the newly added particle are appended to AggCell{k}, and the corresponding rows in DVCell, DVCellMax, DVCellFlag, LTCell, and LNCell are appended simultaneously. The occlusion region is also updated based on the newly added contact point. If the prediction result is "no aggregation," no new particle connection relationships are added, and the aggregate spatial coordinate matrix AggCell remains unchanged. Simultaneously, it is assumed that the liquid bridge in the contact area has partially lost its capacity, and the residual droplet volume at the contact location is updated by attenuation. Let the liquid bridge retention coefficient be... ,in Then the volume of the residual droplet at the contact point after the collision is updated to In this embodiment, η=0.5. If the prediction result is "breakage", the broken connection edges are deleted according to the interparticle liquid bridge connection relationship or adjacency matrix, and the connected components of the updated aggregate structure are identified. To keep the number of MC samples fixed, only the largest connected component containing the largest number of initial particles is used as a new aggregate sample to continue participating in subsequent evolution, and the remaining connected components are discarded as broken and detached parts. The retained largest connected component synchronously inherits its corresponding particle coordinates, current droplet volume, maximum droplet volume, deposition reachability flag, coating thickness, and coating layer number information. At the same time, the residual droplet volume near the breakage contact position is updated according to the liquid bridge retention coefficient η.

[0137] 3.2 Execution of Droplet Deposition Events

[0138] When a droplet deposition event occurs, an aggregate is first randomly selected from the current MC system, and an initial particle is randomly selected from within that aggregate. Then, based on the droplet deposition reachability flag matrix DVCellFlag on the surface of the initial particle, a location is randomly selected from the deposition locations that satisfy DVCellFlag{k}(i,j)=1 as the current droplet deposition location. Let the selected aggregate be numbered k, the selected initial particle be numbered i, the selected deposition location be numbered j, and the initial volume of a single droplet be... Then the accumulated droplet is added to DVCell{k}(i,j), that is, DVCell{k}(i,j) = DVCell{k}(i,j) + And simultaneously update DVCellMax{k}(i,j)=max[DVCellMax{k}(i,j),DVCell{k}(i,j)]. If the selected particle has no deposition location, then reselect a particle or record this deposition event as an empty event.

[0139] Step S4. After executing the collision event and droplet deposition event in step S3, based on the updated AggCell, DVCell, DVCellMax, DVCellFlag, LTCell and LNCell, the droplet drying state, aggregate morphology, fragmentation inheritance state and coating morphology are updated synchronously.

[0140] 4.1 Droplet drying and residual volume renewal

[0141] After the droplets are deposited onto the particle surface, they are approximated as spherical cap-shaped droplets attached to the particle surface, and the droplet drying rate is calculated based on the gas-phase mass transfer process. Initial volume of a single droplet. and initial droplet mass They are respectively:

[0142]

[0143] In the droplet drying process, the mass transfer coefficient at the droplet surface is first calculated based on the gas properties and the relative motion state between the droplet and the gas. Let the gas density be... The gas dynamic viscosity is The diffusion coefficient of water vapor in a gas is The gas velocity is The particle velocity is Then the Schmidt number Sc and the droplet Reynolds number They are respectively:

[0144]

[0145]

[0146]

[0147] in, The contact diameter between the droplet and the particle surface is denoted as . Where is the droplet diameter, It represents the contact angle.

[0148] Further calculation of the Sherwood number Sh:

[0149]

[0150]

[0151]

[0152] This yields the gas-phase mass transfer coefficient at the droplet surface. :

[0153]

[0154] Let the saturated moisture content of the droplet surface be... The moisture content of the inlet gas is The mass flow rate of the liquid spray is The dry gas mass flow rate is The vapor phase moisture content Y can be expressed as:

[0155]

[0156] Drying mass flux per unit area of ​​droplets It can be represented as:

[0157]

[0158] For droplets deposited on the particle surface, the contact diameter between the droplet and the particle surface can be considered during the drying process. While remaining constant, the droplet height and surface area exposed to the gas phase gradually decrease with the residual droplet volume. Let the contact radius be... = / 2, at a certain moment the residual volume of the droplet is Then the height h of the spherical crown satisfies:

[0159]

[0160] Surface area of ​​droplets exposed to the gas phase It can be represented as:

[0161]

[0162] During the MC physics time progression, for the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate, the current residual droplet volume is recorded by DVCell{k}(i,j). Let the residual droplet mass and volume at the n-th time step be... and ,in =DVCell{k}(i,j), then the time step is... The residual mass and volume of the droplet are updated as follows:

[0163]

[0164]

[0165] The residual droplet volume at the next time step is calculated using the above recursive method. Then, update DVCell{k}(i,j) to Subsequently, based on the updated DVCell{k}(i,j) and DVCellMax{k}(i,j), it is determined whether the current deposition drying cycle at this deposition location has been completed. Let the drying completion criterion threshold be... ,in =0.1; when DVCellMax{k}(i,j)>0 and satisfies DVCell{k}(i,j)≤ When DVCellMax{k}(i,j), it is considered that the droplets at the current deposition location have completed drying within the current deposition drying cycle and enter the coating thickness writing step.

[0166] 4.2 Morphology of Aggregates

[0167] Calculate the k-radius of gyration of the aggregates based on the initial particle position information from AggCell.

[0168]

[0169] In the formula and Let represent the position vectors of the i-th and j-th initial particles relative to the centroid of the aggregate, respectively. This calculation method first selects the center of all initial particles as reference points, then calculates their positions relative to the centroid, and finally obtains the average value among all particles to obtain the characterization value of the gyration radius.

[0170] 4.3 Coating thickness and coverage

[0171] After drying at a certain deposition location, the current coating layer number l = LNCell{k}(i,j) is read. The thickness of the single-layer coating is calculated based on the maximum droplet volume recorded at that location during this deposition-drying cycle, DVCellMax{k}(i,j). Let the maximum droplet volume at the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate during the current deposition-drying cycle be... The mass fraction of solids in the spray liquid is After drying, the porosity of the coating layer is The volume of the coating layer formed at that location after this drying process is... It can be represented as:

[0172]

[0173] Assuming that the volume (thickness) is the same at each deposition location, the volume of the entire outer layer covering particle i is... .so

[0174]

[0175] The thickness of the coating layer formed in this cycle was calculated. Then, write it into LTCell{k}(i,j,l+1); then update LNCell{k}(i,j)=l+1, and reset DVCell{k}(i,j) and DVCellMax{k}(i,j) to zero, and enter the next deposition-drying cycle.

[0176] The uniformity of coating thickness can be measured using the coefficient of variation. Where h represents the set of local coverage thicknesses participating in the statistics. For thickness standard deviation, The average thickness is given. From this, the average coating coverage and coating thickness (COV) of a single initial particle, a single agglomerate, and all agglomerates can be calculated.

[0177] The surface coverage of the j-th particle can be determined by the number of droplet deposition sites. and the number of locations where deposited droplets have been received To calculate:

[0178] .

[0179] Step S5: Output Results and Process Evaluation

[0180] After completing the preset MC physical time or event steps, the system outputs agglomerate morphology parameters and coating morphology parameters. Agglomerate morphology parameters may include initial particle number, radius of gyration, adjacency relationship, number of breakages, and particle size distribution; coating morphology parameters may include average coating thickness, coating coverage, occlusion rate, and coating thickness COV. These results can be used to evaluate the impact of process parameters such as spray flow rate, droplet size, gas temperature, humidity, fluidizing gas flow rate, and particle collision frequency on the agglomeration-coating synchronization process.

[0181] The spray flow rate was shown to be 0.4. At that time, the normalized gyration radius of the aggregate ( The changes over time, and the two moments of 500 s and 1000 s. Cumulative distribution. Figure (a) shows... The changes over time, where the orange square test measured 0.4 using a particle size analyzer. Normalized diameter of agglomerates at spray flow rate ( The variation curves of the MC simulation and experimental values ​​show a high degree of agreement, verifying the accuracy and reliability of the MC model in predicting agglomerate size. In the initial stage of agglomeration in the spray fluidized bed (before 500 s), all gyroscope radii show a rapid upward trend, indicating a high probability of wet collisions and agglomeration collisions between particles, leading to a rapid increase in initial agglomerate size. As time progresses, agglomeration becomes the dominant mechanism, accompanied by a small amount of breakage, and the gyroscope radius fluctuates upward. After 1500 s, the rate of increase in gyroscope radius gradually slows down and stabilizes, indicating that particle collision and agglomeration have gradually reached an equilibrium state, and the agglomerate structure tends to stabilize. Comparing Figures (b) and (c), from 500 s to 1000 s, and The growth rates were 31.9%, 27.6%, and 22.3% respectively, and the distribution range decreased from 0.57 to 0.49, indicating that the uniformity of aggregate size has improved.

[0182] 0.4 The aggregates were mainly characterized by particle aggregation in the range of 500 s to 1500 s, with the initial number of particles... The number of particles increased from 8 to 32, and the normalized gyroscope radius increased from 2.18 to 4.62. Between 1500 and 2000 s, the aggregate underwent particle breakage; after 1500 s, the edge blocks of the aggregate fractured, and the initial number of particles in the aggregate at 2000 s... The normalized gyroscope radius decreased to 2.91, and the droplet deposition rate during agglomeration was not affected by agglomerate breakage. The coating coverage gradually increased from 32.23% to 83.52%.

[0183] Figure 9 This represents the evolution of aggregate structure and surface coating morphology.

[0184] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A simulation method for the synchronous evolution of particle agglomeration and coating in a spray fluidized bed, characterized in that, Includes the following steps: Step S1: Parameter setting and state variable initialization, the parameters including but not limited to spray fluidized bed, droplets, particles and Monte Carlo simulation; Using aggregate number k, initial particle number i, and particle surface deposition location number j as unified indexes, a multi-layer state characterization system is established, consisting of aggregate spatial coordinate matrix AggCell, droplet current volume matrix DVCell, droplet maximum volume matrix DVCellMax, droplet deposition location reachability flag matrix DVCellFlag, coating layer thickness matrix LTCell, and coating layer number matrix LNCell. Step S2: Randomly select events and advance Monte Carlo physical time. Based on the scaling relationship between the real system and the MC system, calculate the droplet deposition event frequency and particle collision event frequency in the current MC system; Step S3: If the current event is a particle collision event, then determine the collision result based on the pre-trained wet collision prediction model; If the current event is a droplet deposition event, then the depositable location is randomly determined according to DVCellFlag and DVCell and DVCellMax are updated; Step S3.1: Train machine learning models for predicting different collision events respectively, using collision parameters as input features of the machine learning models and collision results as output labels; the collision events include: single-particle-single-particle wet collision, single-particle-two-particle aggregate wet collision, and single-particle-multi-particle wet collision. If the prediction result of single-particle wet collision is "agglomeration", the center coordinates of the newly added particles are calculated based on the center coordinates of the collided particles and the direction of the collision contact point, and AggCell, DVCell, DVCellMax, DVCellFlag, LTCell, and LNCell are updated simultaneously; if the prediction result is "non-agglomeration", the current agglomerate structure remains unchanged, and the volume of the residual droplet at the collision contact position is updated for liquid bridge loss. For wet collisions between single-particle and dual-particle aggregates or between single-particle and multi-particle aggregates, if the prediction result is "aggregation", the coordinates of the newly added particle are appended to AggCell{k}, and the corresponding rows in DVCell, DVCellMax, DVCellFlag, LTCell, and LNCell are appended simultaneously. At the same time, the occlusion area is updated based on the newly added contact point. If the prediction result is "no aggregation", no new particle connection relationship is added, and the aggregate spatial coordinate matrix AggCell remains unchanged. At the same time, it is assumed that the liquid bridge in the contact area has suffered partial loss, and the volume of the residual droplet at the contact position is updated by attenuation. When the prediction result is "breakage", the broken connection edge is deleted, the connected component is identified, the connected component is retained / recorded, and AggCell, DVCell, DVCellMax, DVCellFlag, LTCell, and LNCell are inherited or updated simultaneously. The wet collision of single-particle and multi-particle aggregates is approximately the collision between a single particle and a local two-particle aggregate of a complete aggregate, and the local collision state is characterized by selected contacting particles and their adjacent structures. Step S3.2: When a droplet deposition event occurs, randomly select an aggregate from the current MC system, and randomly select an initial particle from the aggregate; then, according to the droplet deposition reachability flag matrix DVCellFlag on the surface of the initial particle, randomly select a position from the deposition positions that satisfy DVCellFlag{k}(i,j)=1 as the current droplet deposition position; Step S4. After executing the collision event and droplet deposition event in step S3, based on the updated AggCell, DVCell, DVCellMax, DVCellFlag, LTCell and LNCell, the droplet drying state, aggregate morphology, fragmentation inheritance state and coating morphology are updated synchronously. Step S5: After completing the preset MC physical time or event steps, output the aggregate morphology parameters and coating morphology parameters; the aggregate morphology parameters include, but are not limited to, the initial number of particles, the radius of rotation, the adjacency relationship, the number of fragmentation times, and the particle size distribution; the coating morphology parameters include, but are not limited to, the average coating thickness, the coating coverage rate, the occlusion rate, and the coating thickness variation coefficient COV.

2. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 1, characterized in that, Calculate physical time in step S2 The process is as follows: Step S2.

1. Calculate the initial total number of particles in the real system, denoted as: ,in, For bed quality, The mass of a single initial particle is denoted as . , The initial particle density, The initial particle diameter; Step S2.

2. Introduce the system scaling factor S, denoted as: , This represents the initial total number of particles in the current MC system; Step S2.

3. After considering system scaling, the frequency of droplet deposition events in the MC system is denoted as: = S(t), Let the droplet number flow rate in the real system be denoted as . , For spray mass flow rate, The density is the droplet density. The diameter of the droplet; Step S2.

4. The collision event frequency in the MC system is denoted as... , denoted as: , The initial collision frequency of a single particle. This represents the initial total number of particles in the current MC system; Step S2.

5. Calculate the total event frequency of the current MC step. , recorded as = + , Step S2.

6. Generate two independent uniformly distributed random numbers between 0 and 1. and , Used to determine the type of event that occurred in the current MC step. Used to calculate the physical time of the event's progression; when: If the current event is a particle collision event, then the current event is a droplet deposition event; otherwise, the current event is a droplet deposition event. Step S2.

7. Physical time of the current event progression Obtained by random sampling according to an exponential distribution, its calculation formula is: ; Then update MC physics time: .

3. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 1, characterized in that, Pre-training of the wet collision prediction model: 3.1.

1. Establish a wet particle collision simulation dataset based on the discrete element method (DEM); use the collision parameters obtained from the discrete element simulation as input features of the machine learning model, and use the collision results as output labels to obtain the DEM collision dataset; 3.1.

2. After obtaining the DEM collision dataset, select a machine learning model for classification, train the model using the DEM collision dataset, and obtain a wet collision prediction model.

4. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 1, characterized in that, Let the liquid bridge retention coefficient be... ,in Then the volume of the residual droplet at the contact point after the collision is updated to .

5. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 1, characterized in that, The droplet deposition process is denoted as follows: Let the selected aggregate be numbered k, the selected initial particle be numbered i, the selected deposition location be numbered j, and the initial volume of a single droplet be . Then the accumulated droplet is added to DVCell{k}(i,j), that is, DVCell{k}(i,j) = DVCell{k}(i,j) + And simultaneously update DVCellMax{k}(i,j)=max[DVCellMax{k}(i,j),DVCell{k}(i,j)]; if the selected particle has no deposition location, then reselect a particle or record this deposition event as an empty event.

6. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 1, characterized in that, The droplet drying and residual volume renewal process in step S4: For droplets deposited on the particle surface, the contact diameter between the droplet and the particle surface during the drying process... While remaining constant, the droplet height and surface area exposed to the gas phase gradually decrease with the residual droplet volume; assuming the contact radius... = / 2, at a certain moment the residual volume of the droplet is Then the height h of the spherical crown satisfies: ; Surface area of ​​droplets exposed to the gas phase Represented as: ; During the MC physical time progression, for the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate, the current residual droplet volume is recorded by DVCell{k}(i,j); let the residual droplet mass and volume at the n-th time step be respectively... and ,in =DVCell{k}(i,j), then the time step is... The residual mass and volume of the droplet are updated as follows: ; ; The residual droplet volume at the next time step is calculated using the above recursive method. Then, update DVCell{k}(i,j) to ; Subsequently, based on the updated DVCell{k}(i,j) and DVCellMax{k}(i,j), it is determined whether the current deposition drying cycle at this deposition location has been completed. Let the threshold for determining drying completion be... , where 0 < <1; when DVCellMax{k}(i,j)>0 and satisfies DVCell{k}(i,j)≤ When DVCellMax{k}(i,j), it is considered that the droplets at that deposition location have completed drying within the current deposition drying cycle.

7. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 1, characterized in that, The morphology of the aggregate is represented as follows: Calculate the k-radius of gyration of the aggregates based on the initial particle position information from AggCell. ; In the formula and Let represent the position vectors of the i-th and j-th initial particles relative to the centroid of the aggregate, respectively.

8. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 1, characterized in that, The thickness of the coating layer is expressed as: After drying at a certain deposition location, the current coating layer number l = LNCell{k}(i,j) is read; the thickness of the single-layer coating is calculated based on the maximum droplet volume recorded at that location during this deposition-drying cycle, DVCellMax{k}(i,j); let the maximum droplet volume at the j-th deposition location on the surface of the i-th initial particle in the k-th aggregate during the current deposition-drying cycle be... The mass fraction of solids in the spray liquid is After drying, the porosity of the coating layer is The volume of the coating layer formed at that location after this drying process is... Represented as: ; Assuming the volume is the same at each deposition location, the total volume of the outer layer covering particle i is... The thickness of the coating layer formed in this cycle was calculated. : ; Write the coating thickness into LTCell{k}(i,j,l+1); then update LNCell{k}(i,j)=l+1, and reset DVCell{k}(i,j) and DVCellMax{k}(i,j) to zero, and enter the next deposition-drying cycle.

9. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 8, characterized in that, The uniformity of coating thickness is determined by the coefficient of variation. Where h is the set of local coverage thicknesses participating in the statistics, For thickness standard deviation, The average thickness; from this, the average coating coverage and coefficient of variation (COV) of the coating thickness for a single initial particle, a single agglomerate, and all agglomerates can be calculated; the coating coverage of the j-th particle surface is determined by the number of droplet deposition sites. and the number of locations where deposited droplets have been received To calculate: .

10. The method for simulating the synchronous evolution of particle agglomeration and coating in a spray fluidized bed according to claim 1, characterized in that, The update of the broken inheritance state includes: based on the particle number mapping relationship before and after the breakage, synchronously retaining the DVCell, DVCellMax, DVCellFlag, LTCell and LNCell data corresponding to each initial particle in the retained connected component, and recalculating the droplet deposition position reachability flag matrix DVCellFlag corresponding to the contact relationship of each particle in the broken agglomerate.