Molecular simulation method of powder explosion inhibitor inhibiting micro-reaction of gas and coal dust composite explosion

Abstract

The present application relates to a kind of powder explosion inhibitor inhibits the molecular simulation method of microcosmic reaction of gas coal dust composite explosion, comprising: the initial model of gas / solid mixed reaction system is constructed;Machine learning force field is obtained by training;Balanced uniform reaction model is obtained using machine learning force field optimization initial model, carries out ab initio molecular dynamics simulation under NVT or NPT ensemble, specific temperature and pressure conditions;Data post-processing and calculation are carried out to simulation result, the pyrolysis evolution process is analyzed by ChemTraYzer method, product type and structure information, constructs reaction trajectory network and extracts key element reaction and free radical, carries out sensitivity analysis by Chemkin postprocessor, based on DFT and TST theory, the weak interaction and electronic structure, electric potential distribution in the process of molecular dynamics evolution are calculated, the inhibition mechanism of the selected powder explosion inhibitor and the evolution law of the inhibition of gas / coal dust combustion and explosion are revealed.

Description

Technical Field

[0001] The invention belongs to the technical field of gas / coal dust composite explosion suppression simulation, and relates to a molecular simulation method for suppressing microscopic reactions of a gas / coal dust composite explosion by a powder explosion suppressant. Background Art

[0002] During coal production and processing, deformation and damage inevitably generate large amounts of gas / coal dust mixtures. Once exposed to a fire source, these mixtures can easily cause combined gas and coal dust explosions, posing a serious threat to underground property, equipment, and personnel. Compared to single-phase gas or coal dust explosions, the synergistic interactions and evolutionary characteristics of combined gas / coal dust explosions and explosion suppressants are more complex, generating high temperatures and overpressure while simultaneously releasing large quantities of toxic and hazardous gases and particles, leading to more severe consequences. Effectively preventing and controlling combined gas / coal dust explosions, developing highly effective, targeted powdered explosion suppressants, and elucidating their microscopic suppression mechanisms and evolutionary behaviors have become critical scientific challenges. To address this issue, researchers have investigated the macroscopic properties and explosion suppression performance of various powdered explosion suppressants using methods such as 20L explosive balls and vertical explosion pipes. However, current macroscopic experimental studies have failed to reveal the microscopic dynamics of how powdered explosion suppressants suppress combined gas / coal dust explosions. In addition, due to the limitations of traditional force fields, it is difficult to accurately describe the forces between atoms in the gas / solid mixed reaction system of different powder explosion suppressants and gas and coal dust. Molecular dynamics simulations based on traditional force fields cannot truly simulate the synergistic effect and evolution process of powder explosion suppressants in suppressing gas / coal dust composite explosions. Molecular dynamics (MD) simulations based on reaction force fields are an effective method for determining molecular structure, material design, and performance prediction at the atomic scale. MD simulations fully consider the influence of factors such as electric field, potential energy, temperature, and pressure on molecules. In addition, the optimized force field obtained by machine learning training can reasonably describe the forces between atoms in the simulation system, and the simulation results obtained by calculation are more in line with reality. In recent years, molecular dynamics simulation based on reaction force fields has been widely used to reveal chemical reaction mechanisms and combustion and pyrolysis prediction research, specifically including macroscopic reactant and product analysis, thermodynamic properties, atomic motion trajectories and chemical reaction paths. The application of molecular dynamics simulation based on machine learning force fields in microscopic reaction simulation provides new ideas for the preparation of high-efficiency targeted powder explosion suppressants and the revelation of their suppression mechanism and evolution characteristics of gas / coal dust composite explosions. It has important scientific significance for enriching and developing the "multiphase explosion theory".

[0003] The advantage of reaction molecular dynamics simulation is that it explores the interaction mechanism between molecules in the reaction system from a microscopic perspective, and then reveals the microscopic principles of macroscopic phenomena such as explosion, pyrolysis, flame retardancy, and explosion suppression, which makes up for the limitations of traditional experiments with high costs, long cycles, and low efficiency. The invention patent application with publication number CN111912876A discloses a method for selecting explosion suppression media based on the coupling relationship between explosion pressure and free radicals. According to the explosion chain reaction mechanism and thermodynamic principles, a mathematical model of the peak explosion pressure of combustible gas and the free radical concentration is established. The explosion suppression medium is selected by weight ratio, realizing the combination of macroscopic explosion energy and microscopic particle effects; the invention patent application with publication number CN115831241A discloses a method and system for predicting cellulose pyrolysis reaction products. Based on the iterative training of a multi-layer neural network model, a deep learning potential function of cellulose pyrolysis is obtained and molecular dynamics simulation is performed to predict the pyrolysis products of cellulose; the invention patent application with publication number CN114783537A discloses a method for selecting and evaluating the performance of a drag reducing agent and its application, and uses molecular dynamics simulation to establish a drag reducing agent. The optimal parameter model of the resistance reducing agent is established, and the molecular simulation optimization results are verified by indoor pipe flow friction test to determine the resistance reduction rate of the optimal resistance reducing agent; the invention patent application with publication number CN115898510A discloses a composite powder explosion suppressant for suppressing coal dust composite explosion and its preparation method, melamine polyphosphate, fly ash, ammonium dihydrogen phosphate, calcium silicate, and montmorillonite are compounded into a composite powder explosion suppressant according to weight fraction, and the composite material plays a synergistic role to achieve the ultimate purpose of explosion suppression; the invention patent application with publication number CN114510822A discloses an environmentally friendly method for suppressing gaseous solid decomposition products, combining density functional theory and reaction molecular dynamics simulation to analyze the influence and contribution of each related reaction on the generation or removal of carbon-containing solid decomposition products, thereby realizing the method of changing variables and multivariable collaborative optimization to suppress the components of solid decomposition products.

[0004] However, the above prior art still has the following deficiencies:

[0005] The invention patent with publication number CN111912876A discloses a calculation method for determining the weight ratio of the free radical gain effect on the explosion pressure based on a mathematical model of the combustible gas explosion pressure peak and the free radical concentration. It only considers H, OH, and CH2O free radicals as gain weights, while in the actual reaction process, there are more types and quantities of free radicals that have a gain effect on the explosion pressure; the invention patent with publication number CN115831241A discloses a method for predicting cellulose pyrolysis products based on deep learning potential functions and molecular dynamics simulations. It can only predict the types and structural information of the final pyrolysis products, but fails to reveal the pyrolysis evolution process and formation mechanism of cellulose; the invention patent with publication number CN114783537A discloses the advantages of the drag reducing agent. The selection and performance evaluation method is evaluated by constructing an optimal parameter molecular dynamics model. However, the factors affecting the performance of the drag reducer are complex and diverse, and the established optimal parameter model cannot objectively and comprehensively evaluate the performance of the drag reducer; the compound powder explosion suppressant disclosed in the invention patent with publication number CN115898510A achieves the ultimate explosion suppression purpose through the explosion suppression effect of each material, and fails to reveal the coupling mechanism of the composite powder material in the process of suppressing the explosion reaction; the multivariable collaborative optimization suppression method disclosed in the invention patent with publication number CN114510822A only considers the influence of the carbon-containing solid reaction itself on the decomposition products, while the actual gas solid decomposition characteristics are affected by changes in the external space environment and cannot simulate the real situation of the solid decomposition products.

[0006] In summary, existing numerical simulation theoretical models and evaluation methods struggle to realistically simulate the explosion conditions and influencing variables encountered in actual projects, lacking a detailed theoretical understanding of the explosion or suppression process. Furthermore, the formulation and evaluation of composite powder explosion suppressants, their physical and chemical coupling mechanisms, and their suppression mechanisms require further research. Furthermore, few patent applications have established molecular dynamics models that reasonably describe the forces and reactions between atoms in the system, ignoring the impact of the reaction force field itself on model accuracy. On-site explosion suppression experiments, however, require high safety standards and suffer from high costs, large scale, and poor reproducibility.

[0007] Therefore, there is an urgent need for a molecular simulation method that can realistically simulate the gas / coal dust composite explosion reaction process under complex actual engineering conditions and the process of powder explosion suppressants inhibiting the explosion reaction. Summary of the Invention

[0008] In view of this, the purpose of the present invention is to provide a molecular simulation method for the microscopic reaction of powder explosion suppressants in suppressing gas and coal dust composite explosions, which is used to study the microscopic mechanism of the selected powder explosion suppressants in suppressing gas / coal dust composite explosions. This method can simulate the decomposition and evolution process of gas / coal dust composite explosions and the explosion reaction suppressed by powder explosion suppressants under periodic boundary conditions. By using the ChemTraYzer method, density functional theory (DFT) combined with transition state theory (TST) and Chemkin post-processor, a detailed reaction trajectory network is constructed and the types and structural information of key elementary reactions and free radicals are extracted, thereby revealing the coupled reaction laws of the gas / solid mixed reaction system and the explosion suppression characteristics and mechanism of action of the powder explosion suppressant. In addition, the machine learning force field based on ADF / BAND and Quantum Espresso method or inputting experimental parameters to generate DFT training set training can more reasonably describe the interaction between atoms in the mixed reaction model, and the obtained simulation results, model and reaction process are closer to reality. The present invention enriches the theoretical model of gas / coal dust explosion suppression and disaster reduction technology, and ultimately achieves the purpose of ensuring coal mine production safety and the safety of people's lives and property.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] A molecular simulation method for suppressing the microscopic reaction of a powder explosion suppressant in a gas-coal dust composite explosion is based on a machine learning force field-based molecular dynamics simulation method and data post-processing and calculation methods. The method specifically includes the following steps:

[0011] S1: Obtain the unit cell or molecular information of the powder explosion suppressant, the type and structure of gas, the molecular structure and chemical characteristics of coal, establish the crystal or cluster model of the powder explosion suppressant, the CH4 / O2 / N2 molecular model and the coal macromolecular model, and calculate the volume parameters of the simulation box with periodic boundary conditions;

[0012] S2: Perform geometric optimization and simulated annealing calculations on the molecular model to obtain the molecular structure after preliminary optimization of atomic spacing, bond angles and dihedral angles, and establish the initial model of the gas / solid mixed reaction system of powder explosion suppressant, CH4 / O2 / N2 molecules and coal macromolecules;

[0013] S3: For the initial model of the gas / solid mixed reaction system, generate a DFT training set based on the ADF or BAND & Quantum Espresso method or experimental data, and train to obtain a machine learning force field that can reasonably describe the interactions between atoms in the reaction model. Where DFT represents density functional theory, ADF represents DFT calculation software, and BAND represents periodic system calculation program;

[0014] S4: Optimize the initial model based on the machine learning force field to obtain a balanced reaction model, and perform ab initio molecular dynamics simulations under specific temperature and pressure conditions in the NVT or NPT ensemble; NVT represents the canonical ensemble and NPT represents the isothermal and isobaric ensemble;

[0015] S5: Perform data post-processing and calculation on the molecular dynamics simulation results. Use the ChemTraYzer method to analyze the pyrolysis evolution process, product types and structural information, construct a reaction trajectory network and extract key elementary reactions and free radicals. Perform sensitivity analysis using the Chemkin post-processor. Use DFT combined with TST theory to calculate the weak interactions, electronic structure and potential distribution during the molecular dynamics evolution process. This reveals the inhibition mechanism of the selected powder explosion suppressant and its evolution law in inhibiting gas / coal dust explosions. TST represents the transition state.

[0016] Furthermore, in step S1, the powder explosion suppressant crystal or cluster model, the CH4 / O2 / N2 molecular model and the coal macromolecular model are respectively established using Materials Studio (MS) software.

[0017] The information of powder explosion suppressant unit cell or molecular parameters and atomic positions is obtained through the Materials Project material database.

[0018] The powder explosion suppressant or cluster model is established by processing the unit cell or molecular model through cell expansion, sectioning, polymerization, etc. using Materials Studio (MS) software.

[0019] The gas type and structure information was obtained by testing with a gas analyzer model LRGA-3100, and was established by constructing CH4 / O2 / N2 with different volume concentrations using Materials Studio (MS) software.

[0020] The macromolecular structure and chemical characteristics of coal are analyzed by industrial element analysis, CRD microcrystalline structure analysis, FTIR functional group analysis, 13 The data were obtained through experimental analysis tests such as C NMR carbon spectrum analysis and XPS element distribution analysis, and established by Materials Studio (MS) software.

[0021] The calculation formula for the volume parameter of the simulation box with periodic boundary conditions is:

[0022]

[0023] Among them, V A represents the volume parameter of the simulation box under periodic boundary conditions; Mr represents the relative molecular mass of the gas / solid mixed reaction system in the system; V Hrepresents the volume of the pyrolysis chamber; m represents the mass of the gas / solid mixture in the system; N A represents Avogadro's constant.

[0024] Furthermore, in step S2, the geometry optimization and simulated annealing are to perform preliminary optimization on the powder explosion suppressant crystal or cluster model, CH4 / O2 / N2 molecular model and coal macromolecular model using the Dreiding force field of the Forcite module of Materials Studio (MS) software to obtain a molecular model with stable and reasonable atomic spacing, bond angles and dihedral angles;

[0025] The initial model of the gas / solid mixed reaction system is a gas / solid system model of a periodic boundary condition simulation box and a mixture of powder explosion suppressant crystals or clusters, CH4 / O2 / N2 molecules and coal molecules established using the AmorphousCell (AC) module of Materials Studio (MS) software.

[0026] Furthermore, in step S3, the machine learning force field is based on the ADF or BAND & Quantum Espresso method or input experimental parameters to perform parallel calculations, cross-validation, and iterative expansion of the training set to generate a DFT training set, thereby obtaining a machine learning force field that can reasonably describe the interactions between atoms in the reaction model; the machine learning force field has minimal differences in parameters such as energy, bond length, and bond angle between the DFT or experimental parameters, and specifically includes the following steps:

[0027] S31: Let the number of iterations k be 1, and the original training set be the DFT training set generated by the kth iteration;

[0028] S32: Randomly select K mutually exclusive and equal-sized training subsets from the original training set and perform K rounds of training and cross-validation;

[0029] S33: Importing various structural parameters such as bond lengths and bond angles of the gas / solid mixed reaction system in each training subset as reference data into the multi-layer neural network model of the corresponding training subset to obtain the system energy, atomic charge and force conditions of the molecular model of each gas / solid mixed reaction system;

[0030] S34: Calculate the loss function value of the multi-layer neural network model corresponding to each training subset based on the system energy, atomic charge and force conditions of each gas / solid mixed reaction system molecular model, that is, the value describing the difference between the force field calculation system and the DFT calculation system after considering the weight of each sample;

[0031] S35: When the loss function value is greater than the preset loss function threshold, return to step S33, update the model parameters corresponding to each training subset and further optimize until the loss function value is less than the preset loss function threshold, and the loss function is considered to be minimized, that is, the training obtains a so-called optimal force field;

[0032] The calculation formula of the loss function value is:

[0033]

[0034] Among them, W(E i ,E j ) represents the loss function value; N1, N2, and N3 are the atomic numbers of powder explosion suppressant, gas, and coal molecules in the gas / solid mixed reaction system, respectively; Δx k is the root mean square error of the force on the kth atom in the molecular configuration of the gas / solid mixed reaction system in each training subset, m and n are the mth and nth atoms respectively; ΔI is the root mean square error of the system energy corresponding to each training subset; α and β are the correction coefficients of the atomic force and the system energy respectively; E i and E j They represent the time-related pre-exponential factors of the force term and energy term respectively; C is a constant;

[0035] S36: Output the minimum force field of the loss function corresponding to each training subset as the machine learning force field of the kth successful iteration corresponding to each training subset;

[0036] S37: Calculate the model deviation of each atom and the maximum model deviation of the gas / solid mixed reaction system molecular model in sequence according to the atomic charge and force conditions of the gas / solid mixed reaction system molecular model in the training subset;

[0037] The calculation formula of the model deviation is:

[0038]

[0039] Among them, S k is the model deviation value of the kth atom calculated based on the machine learning optimized force field; is the average force on the kth atom; is the average charge intensity of the kth atom; σ is the maximum standard deviation of the force on the atom; γ is the deviation correction coefficient; M is a constant;

[0040] S38: Select the atomic charge and force conditions of the gas / solid mixed reaction system molecular model corresponding to the maximum model deviation as the expanded training set for k+1 iterations, and return to step S32 until the maximum model deviation of the gas / solid mixed reaction system molecular model is less than the preset deviation threshold, and the training is considered successful.

[0041] Furthermore, step S4 specifically includes: performing NVT or NPT ensemble relaxation or energy minimization on the initial model based on the machine learning force field to obtain a balanced and uniform molecular dynamics model;

[0042] The NVT ensemble is a canonical ensemble, which means it has a certain number of particles (N), volume (V) and temperature (T), and only the Thermostat is set without the Barostat;

[0043] The NPT ensemble is an isothermal and isobaric ensemble, which means it has a certain number of particles (N), pressure (P) and temperature (T), and the Thermostat and Barostat are set at the same time;

[0044] The relaxation or energy minimization is to place the atoms and molecules in the reaction system in a balanced and uniform position under a suitable force field, thereby obtaining a molecular dynamics model that is closer to reality.

[0045] Furthermore, in step S4, ab initio molecular dynamics simulation is a method based on bond order theory to describe the dynamic evolution of chemical bonds between atoms in a complex system. The continuity equation of bond order and interatomic distance is shown as follows:

[0046]

[0047] Among them, BO, γ ij , γ o 、P bo are the bond order, interatomic distance, equilibrium bond length and empirical parameters between atoms respectively; this equation generates a differentiable potential energy surface through the characteristic transitions between σ, π and ππ bonds, thereby calculating the interatomic forces.

[0048] Furthermore, in step S4, the total energy of the reaction system is the sum of all energy items, and the expression is shown as follows:

[0049] E system =E bond +E over +E under +E lp +E val +E tor +E vdWaals +E Coulomb

[0050] Among them, E system 、E bond 、E over 、E under 、E lp 、E val 、E tor 、E vdWaals and E CoulombThey represent total energy, bond energy, overcoordination penalty, undercoordination stability, lone pair energy, valence angle energy, torsion angle energy, van der Waals energy and Coulomb energy respectively.

[0051] Furthermore, in step S5, data post-processing and calculation are performed on the molecular dynamics simulation results, specifically including:

[0052] The ChemTraYzer method is used to track the motion trajectories of molecules and atoms as they change during molecular dynamics simulations. The pyrolysis evolution, types, and structures of reactants, intermediates, and final products are analyzed. A detailed molecular reaction pathway is constructed, and reaction rate constants are calculated. Key elementary reactions and free radicals are extracted, and the coupling reaction patterns of the selected powdered explosion suppressant and gas / coal dust mixture are analyzed.

[0053] Density functional theory (DFT) combined with transition state theory (TST) was used to calculate the weak interactions, electronic structure, and potential distribution during the molecular dynamics evolution process. The Gibbs free energy of reactants and products was calculated and the activation energy was obtained. This allowed the molecular reaction pathway to be further optimized and refined, revealing the suppression characteristics and evolution process of the selected powder explosion suppressant on gas / coal dust explosions.

[0054] Sensitivity analysis was performed using the Chemkin post-processor to reveal the influence of the selected powder explosion suppressant on the key elementary reactions and free radicals in the gas / coal dust composite explosion process.

[0055] The beneficial effects of the present invention are that the method of the present invention can simulate the decomposition and evolution process of gas / coal dust composite explosions and powder explosion suppressants suppressing explosion reactions under periodic boundary conditions, adopt a machine learning method to train the reaction force field, and use the trained machine learning force field to perform molecular dynamics simulation, thereby realizing low-cost, high-precision, fast, and scalable application of the explosion suppression mechanism and evolutionary behavior of gas / coal dust composite explosion suppressants, which has important reference value for the research and development and optimization of powder explosion suppressants used for the prevention and control of gas / coal dust composite explosions. Specifically, the present invention uses the ChemTraYzer method to track the motion trajectories of molecules and atoms as they change with molecular dynamics simulation compensation, analyzes the thermal decomposition evolution process and type and structural information of reactants, intermediates and final products, constructs a detailed molecular reaction path and calculates the reaction rate constant, extracts key elementary reactions and free radicals, and analyzes the coupling reaction law of the selected powder explosion suppressant and the gas / coal dust mixed system; calculates the weak interactions and electronic structure and potential distribution in the molecular dynamics evolution process through density functional theory (DFT) combined with transition state theory (TST), calculates the accurately optimized Gibbs free energy of reactants and products and obtains the activation energy, further optimizes and improves the molecular reaction path, and reveals the inhibitory characteristics and evolution process of the selected powder explosion suppressant on gas / coal dust explosion; and performs sensitivity analysis through the Chemkin post-processor to reveal the degree of influence of the selected powder explosion suppressant on key elementary reactions and free radicals in the gas / coal dust composite explosion process. Furthermore, by using the ADF / BAND and Quantum Espresso methods or inputting experimental parameters for parallel computation, cross-validation, and iteratively expanding the training set to generate a DFT training set, a machine learning force field is developed that can reasonably describe the interactions between atoms in the reaction model. The molecular dynamics (MD) simulation results and models based on this machine learning force field are more similar to the actual reaction process. This invention enriches the theoretical model of gas / coal dust explosion suppression and disaster reduction technology, ultimately achieving the goal of ensuring coal mine production safety and the safety of people's lives and property.

[0056] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0058] Figure 1 A flow chart of a molecular simulation method for suppressing the microscopic reaction of a gas-coal dust composite explosion using a powder explosion suppressant provided by an embodiment of the present invention;

[0059] Figure 2 Three intercalation-modified kaolin crystal models provided in the embodiments of the present invention;

[0060] Figure 3 A gas / solid mixed reaction system model provided in an embodiment of the present invention;

[0061] Figure 4 Iterative fitting of the loss function during the machine learning training reaction force field provided by the embodiment of the present invention;

[0062] Figure 5 The long flame coal macromolecular pyrolysis reaction path provided by the embodiment of the present invention;

[0063] Figure 6 Statistics on the types and quantities of molecular products provided by the embodiments of the present invention;

[0064] Figure 7 The CO gas formation mechanism provided by the embodiment of the present invention;

[0065] Figure 8 The H2 gas formation mechanism provided by the embodiment of the present invention;

[0066] Figure 9 The formation mechanism of H2O molecules provided by the embodiment of the present invention;

[0067] Figure 10 The charge distribution of long flame coal macromolecules provided by the embodiment of the present invention;

[0068] Figure 11 This paper provides the key elementary reactions and free radical sensitivity analysis of the embodiments of the present invention. DETAILED DESCRIPTION

[0069] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0070] See also Figures 1 to 11 The embodiment of the present invention provides a molecular simulation method for suppressing the microscopic reaction of gas-coal dust composite explosion by a powder explosion suppressant, which is a molecular dynamics simulation and data post-processing and calculation method based on machine learning force field.

[0071] like Figure 1As shown, the molecular simulation method specifically includes the following steps:

[0072] S1: Establish the powder explosion suppressant crystal or cluster model, CH4 / O2 / N2 molecular model and coal macromolecular model, and calculate the volume parameters of the periodic boundary condition simulation box.

[0073] Specifically, the molecular models were established using Materials Studio (MS) software, namely, the modified kaolin powder explosion suppressant crystal model, the gas / air mixture molecular model, and the long flame coal macromolecular model, and the volume parameters of the simulation box with periodic boundary conditions were calculated;

[0074] Kaolin unit cell parameters and atomic positions are obtained from the Materials Project database. In this example, the molecular formula of kaolin is Al4[Si4O 10 ](OH)8 Konlin 721216 type kaolin, the specific parameters of which are shown in Table 1 below.

[0075] Table 1 Konlin 721216 kaolin unit cell parameters

[0076]

[0077]

[0078] The Konlin 721216 kaolin unit cell model was imported into the BuildCrystal module of Materials Studio (MS) software to construct a single kaolin unit cell model. Then, the single kaolin unit cell structure was expanded by 3 units in the X direction and 2 units in the Y direction and orthogonalized using the Supercell module in Build Symmetry to obtain a supercell kaolin crystal model.

[0079] The supercellular kaolin crystal model was modified by intercalation to obtain three modified kaolin crystal models. The hydrogen bond forces between the three intercalated particles and the kaolin sheets were calculated using the Hydrogen bond calculation in the Build module of Materials Studio (MS) software. The spacing between H and O atoms was adjusted to The crystal models of three modified kaolin powder explosion suppressants, namely dimethyl sulfoxide (DMSO), potassium acetate (KAc), and ammonium sulfamate (AS), were constructed. Figure 2 shown.

[0080] The type and structure of the gas / air mixture can be obtained by testing with a gas analyzer model LRGA-3100. The main components are CH4, O2, and N2 molecules. In this example, Materials Studio (MS) software is used to construct a CH4 concentration of 9.5 vol% and an O2 concentration of 21 vol% to simulate the gas / air mixture.

[0081] The macromolecular structure and chemical characteristics of long flame coal are obtained through experimental analysis and testing, through industrial element analysis, CRD microcrystalline structure analysis, FTIR functional group analysis, 13 C NMR carbon spectrum analysis and XPS element distribution analysis determined its molecular formula (C 123 H 79 O 21 The long flame coal molecular model was established using Materials Studio (MS) software based on the characteristic information of N1) and molecular structure.

[0082] The simulation box with periodic boundary conditions was created using Crystals in the Build module of Materials Studio (MS) software, with an initial density of 0.1 g / cm 3 , the calculation formula of the simulation box volume parameter is:

[0083]

[0084] Among them, V A represents the volume parameter of the simulation box under periodic boundary conditions; Mr represents the relative molecular mass of the gas / solid mixed reaction system in the system; V H represents the volume of the pyrolysis chamber; m represents the mass of the gas / solid mixture in the system; N A Represents Avogadro's constant. Specific parameters are shown in Table 2 below:

[0085] Table 2 Simulation working condition settings

[0086]

[0087]

[0088] Note: Mixed Coal is a mixed system of CH4 / O2 / N2 / Coal; Mixed K(1) is a mixed system of CH4 / O2 / N2 / Coal and kaolin crystals; Mixed MK(1) is a mixed system of CH4 / O2 / N2 / Coal and three modified kaolin crystals (DMSO, KAc, AS).

[0089] S2: Perform geometric optimization and simulated annealing calculations on the molecular model to obtain the molecular structure after preliminary optimization of atomic spacing, bond angles and dihedral angles, and establish the initial model of the gas / solid mixed reaction system of powder explosion suppressant and CH4 / O2 / N2 molecules and coal macromolecules.

[0090] Specifically, the geometry optimization is a preliminary optimization of the modified kaolin crystal model and the long flame coal macromolecular model using the Forcite module of Materials Studio (MS) software, wherein the task objective is GeometryOptimization, the accuracy is Fine, the force field selects the Dreiding force field with stronger universality in organic molecules, the algorithm is Smart, the maximum number of iterations is 50,000, the charge distribution is QEq, and the electrostatic interaction and van der Waals interaction use Atombased to obtain the modified kaolin crystal model and the long flame coal macromolecular model with stable and reasonable atomic spacing, bond angles and dihedral angles.

[0091] Simulated annealing is a global energy minimum optimization of long flame coal macromolecules using the Anneal module based on geometric optimization. The number of annealing cycles is set to 30, the initial temperature is 300K, the intermediate cycle temperature is 800K, the step size is 0.1fs, the temperature control program is Nose, and the dynamic calculation of the NVT ensemble is performed 100 times in each step. Finally, the long flame coal macromolecular model with the minimum energy and the most stable structure is obtained.

[0092] The initial model of the gas / solid mixed reaction system is a gas / solid system model that is a mixture of a periodic boundary condition simulation box, a modified kaolin crystal model, a gas / air mixed gas model, and a long flame coal macromolecule model established using the Amorphous Cell module of Materials Studio (MS) software. Figure 3 shown.

[0093] S3: For the initial model of the gas / solid mixed reaction system, a DFT training set is generated using the ADF / BAND and Quantum Espresso methods or experimental data to obtain a machine learning force field that can reasonably describe the interactions between atoms in the reaction model.

[0094] Specifically, the machine learning force field is based on the ADF or BAND&Quantum Espresso method or input experimental parameters to perform parallel calculations, cross-validation, and iteratively expand the training set to generate a DFT training set, thereby obtaining a machine learning force field that can reasonably describe the interactions between C / H / O / N / S / Si / Al / P atoms in the reaction model;

[0095] The machine learning force field has minimal differences in energy, bond length, bond angle, and other parameters from DFT or experiments. The specific steps include:

[0096] (1) Let the number of iterations k be 1, and the original training set be the DFT training set generated by the kth iteration.

[0097] (2) Randomly select K mutually exclusive and equal-sized training subsets from the original training set and perform K rounds of training and cross-validation.

[0098] (3) The various structural parameters of the gas / solid mixed reaction system in each training subset, such as bond lengths and bond angles, are imported into the multi-layer neural network model of the corresponding training subset as reference data to obtain the system energy, atomic charge and force conditions of the molecular model of each gas / solid mixed reaction system.

[0099] (4) According to the system energy, atomic charge and force conditions of each gas / solid mixed reaction system molecular model, the loss function value of the multi-layer neural network model corresponding to each training subset is calculated, that is, the value describing the difference between the force field calculation system and the DFT calculation system after considering the weights of each sample.

[0100] (5) When the loss function value is greater than the preset loss function threshold, return to step (3), update the model parameters corresponding to each training subset and further optimize until the loss function value is less than the preset loss function threshold, and the loss function is considered to be minimized, that is, a so-called optimal force field is obtained through training.

[0101] The calculation formula of the loss function value is:

[0102]

[0103] Among them, W(E i ,E j ) represents the loss function value; N1, N2, and N3 are the atomic numbers of powder explosion suppressant, gas, and coal molecules in the gas / solid mixed reaction system, respectively; Δx k is the root mean square error of the force on the kth atom in the molecular configuration of the gas / solid mixed reaction system for each training subset; ΔI is the root mean square error of the system energy corresponding to each training subset; α and β are the correction coefficients of the atomic force and the system energy, respectively; E i and E j They represent the time-related pre-exponential factors of the force term and energy term respectively; C is a constant.

[0104] (6) Output the minimum force field of the loss function corresponding to each training subset as the machine learning force field of the kth successful iteration corresponding to each training subset, such as Figure 4 shown.

[0105] (7) According to the atomic charges and forces of the gas / solid mixed reaction system molecular model in the training subset, the model deviation of each atom and the maximum model deviation of the gas / solid mixed reaction system molecular model are calculated in turn.

[0106] The formula for calculating model bias is:

[0107]

[0108] Among them, S k is the model deviation value of the kth atom calculated based on the machine learning optimized force field; is the average force on the kth atom; is the average charge intensity of the kth atom; σ is the maximum standard deviation of the force acting on the atom; γ is the deviation correction coefficient; M is a constant.

[0109] (8) The atomic charge and force conditions of the gas / solid mixed reaction system molecular model corresponding to the maximum model deviation are selected as the expanded training set for k+1 iterations, and the training is considered successful when the maximum model deviation of the gas / solid mixed reaction system molecular model is less than the preset deviation threshold.

[0110] S4: The initial model is optimized based on the machine learning force field to obtain a balanced and uniform reaction model, and ab initio molecular dynamics simulations are performed under NVT or NPT ensemble and specific temperature and pressure conditions.

[0111] Specifically, based on the machine learning force field, the initial model of the gas / solid mixed reaction system was fully relaxed at a temperature of 300K and a time of 10ps under the NVT ensemble to obtain a molecular dynamics model that is uniformly balanced, has the lowest energy, and is closer to reality.

[0112] Based on the machine learning force field, ab initio molecular dynamics simulations were performed on the optimized molecular dynamics model in the NVT ensemble with a temperature of 3000-4000 K, a time of 100 ps, ​​and a constant pressure of 0.8 MPa. The temperature control method was the Berendsen method, which is stable and has a fast equilibrium point. The heating rate was 10 K / ps, and the charge balance convergence accuracy was 1×10 -6 C.

[0113] Ab initio molecular dynamics simulation is a method based on bond order theory to describe the dynamic evolution of chemical bonds between atoms in complex systems. The continuity equation for bond order and interatomic distance is shown below:

[0114]

[0115] Among them, BO, γ ij , γ o 、P boThe equations are the bond order, interatomic distance, equilibrium bond length, and empirical parameters between atoms. The equations generate differentiable potential energy surfaces by using characteristic transitions between σ, π, and ππ bonds to calculate the interatomic forces.

[0116] The total energy of the reaction system is the sum of all energies, as shown in the following formula:

[0117] E system =E bond +E over +E under +E lp +E val +E tor +E vdWaals +E Coulomb

[0118] Among them, E system 、E bond 、E over 、E under 、E lp 、E val 、E tor 、E vdWaals and E Coulomb They represent total energy, bond energy, overcoordination penalty, undercoordination stability, lone pair energy, valence angle energy, torsion angle energy, van der Waals energy and Coulomb energy respectively.

[0119] S5: Perform data post-processing and calculation on the molecular dynamics simulation results, analyze the pyrolysis evolution process, types and structural information through the ChemTraYzer method, construct a reaction trajectory network and extract key elementary reactions and free radicals, perform sensitivity analysis through the Chemkin post-processor, calculate the weak interactions and electronic structure and potential distribution in the molecular dynamics evolution process through density functional theory (DFT) combined with transition state theory (TST), and reveal the inhibition mechanism of the selected powder explosion suppressant and its evolution law of inhibiting gas / coal dust explosion.

[0120] Specifically, the data post-processing and calculation of the reaction molecular dynamics simulation results include:

[0121] The ChemTraYzer method is used to track the motion trajectories of molecules and atoms as they change with molecular dynamics simulation compensation. The pyrolysis evolution process, types and structural information of reactants such as modified kaolin powder explosion suppressant, gas / air mixture, long flame coal macromolecules, as well as intermediate and final products are analyzed. A detailed molecular reaction path is constructed and the reaction rate constant is calculated. The key elementary reactions and free radicals are extracted. The coupling reaction law of the selected powder explosion suppressant and gas / coal dust mixture system is analyzed. The formation mechanism of the main products such as CO, H2, H2O in this embodiment is as follows: Figures 5 to 9 As shown;

[0122] The weak interactions, electronic structure, and potential distribution during the molecular dynamics evolution process are calculated by density functional theory (DFT) combined with transition state theory (TST), and the Gibbs free energy of reactants and products is calculated to obtain the activation energy. The molecular reaction path is further optimized and improved, and the inhibitory characteristics and evolution process of the selected powder explosion suppressant on gas / coal dust explosion are revealed. The charge distribution of the long flame coal macromolecules in this embodiment is as follows: Figure 10 As shown;

[0123] The sensitivity analysis was performed by the Chemkin post-processor to reveal the influence of the selected powder explosion suppressant on the key elementary reactions and free radicals in the gas / coal dust composite explosion process, such as Figure 11 shown.

[0124] The molecular simulation method for the microscopic reactions of powdered explosion suppressants in suppressing gas-coal dust combined explosions belongs to the field of gas / coal dust combined explosion suppression simulation technology and is used to study the mechanism and evolution of powdered explosion suppressants in suppressing gas / coal dust combined explosions. A machine learning training method based on parallel calculations, cross-validation, and iterative generation of a DFT training set by expanding the training set using ADF / BAND and Quantum Espresso methods or by inputting experimental parameters can fill the gap in molecular dynamics simulations where traditional force fields cannot accurately and reasonably describe the interactions between atoms within the system. Through data post-processing and analytical calculations combining the ChemTraYzer method, density functional theory (DFT) combined with transition state theory (TST) and the Chemkin post-processor, a detailed reaction trajectory network and key elementary reactions, as well as information on the types and structures of free radicals, were established. This further reveals the coupled reaction patterns of the gas / solid mixed reaction system and the explosion suppression characteristics and mechanism of action of the powdered explosion suppressant. The present invention has important scientific significance for enriching and developing the "multiphase explosion theory" and the theoretical model of gas / coal dust explosion suppression and disaster reduction technology, and ultimately achieves the purpose of ensuring coal mine production safety and the safety of people's lives and property.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A molecular simulation method for suppressing the microscopic reaction of gas-coal dust composite explosion by a powder explosion suppressant, characterized in that: The method specifically comprises the following steps: S1: Obtain the unit cell or molecular information of the powder explosion suppressant, the type and structure of gas, the molecular structure and chemical characteristics of coal, establish the crystal or cluster model of the powder explosion suppressant, the CH4 / O2 / N2 molecular model and the coal macromolecular model, and calculate the volume parameters of the simulation box with periodic boundary conditions; S2: Perform geometric optimization and simulated annealing calculations on the molecular model to obtain the molecular structure after preliminary optimization of atomic spacing, bond angles and dihedral angles, and establish the initial model of the gas / solid mixed reaction system of powder explosion suppressant, CH4 / O2 / N2 molecules and coal macromolecules; S3: For the initial model of the gas / solid mixed reaction system, generate a DFT training set based on the ADF or BAND & Quantum Espresso method or experimental data, and train to obtain a machine learning force field that can reasonably describe the interactions between atoms in the reaction model. Where DFT represents density functional theory, ADF represents DFT calculation software, and BAND represents periodic system calculation program; The machine learning force field is based on the ADF or BAND & Quantum Espresso method or input experimental parameters to perform parallel calculations, cross-validation and iterative expansion of the training set to generate a DFT training set, thereby obtaining a machine learning force field that can reasonably describe the interactions between atoms in the reaction model; specifically, the following steps are included: S31: Let the number of iterations k be 1, and the original training set be the DFT training set generated by the kth iteration; S32: Randomly select K mutually exclusive and equal-sized training subsets from the original training set and perform K rounds of training and cross-validation; S33: Importing various bond lengths and bond angles of the gas / solid mixed reaction system in each training subset as reference data into the multi-layer neural network model of the corresponding training subset, and obtaining the system energy, atomic charge and force conditions of the molecular model of each gas / solid mixed reaction system; S34: Calculate the loss function value of the multi-layer neural network model corresponding to each training subset based on the system energy, atomic charge and force conditions of each gas / solid mixed reaction system molecular model, that is, the value describing the difference between the force field calculation system and the DFT calculation system after considering the weights of each sample; S35: When the loss function value is greater than the preset loss function threshold, return to step S33, update the model parameters corresponding to each training subset and further optimize until the loss function value is less than the preset loss function threshold, and the loss function is considered to be minimized, that is, the training obtains a so-called optimal force field; The calculation formula of the loss function value is: in, Represents the loss function value; 、 、 are the atomic numbers of powder explosion suppressant, gas, and coal molecules in the gas / solid mixed reaction system respectively; is the root mean square error of the force on the kth atom of the molecular configuration in the gas / solid mixed reaction system of each training subset, m 、 n Respectively m Hedi n atoms; is the root mean square error of the system energy corresponding to each training subset; 、 are the correction coefficients for atomic forces and system energy, respectively; and They represent the time-dependent pre-exponential factors of the force term and energy term respectively; is a constant; S36: Output the minimum force field of the loss function corresponding to each training subset as the machine learning force field of the kth successful iteration corresponding to each training subset; S37: Calculate the model deviation of each atom and the maximum model deviation of the gas / solid mixed reaction system molecular model in sequence according to the atomic charge and force conditions of the gas / solid mixed reaction system molecular model in the training subset; The calculation formula of the model deviation is: in, is the model deviation value of the kth atom calculated based on the machine learning optimized force field; is the average force on the kth atom; is the average charge intensity of the kth atom; is the maximum standard deviation of the forces acting on the atoms; is the deviation correction factor; is a constant; S38: Select the atomic charge and force conditions of the gas / solid mixed reaction system molecular model corresponding to the maximum model deviation as the expanded training set and perform k+1 iterations, returning to step S32 until the maximum model deviation of the gas / solid mixed reaction system molecular model is less than the preset deviation threshold, and the training is considered successful; S4: Optimize the initial model based on the machine learning force field to obtain a balanced reaction model, and perform ab initio molecular dynamics simulations under specific temperature and pressure conditions in the NVT or NPT ensemble; NVT represents the canonical ensemble and NPT represents the isothermal and isobaric ensemble; S5: Perform data post-processing and calculation on the molecular dynamics simulation results. Use the ChemTraYzer method to analyze the pyrolysis evolution process, product types and structural information, construct a reaction trajectory network and extract key elementary reactions and free radicals. Perform sensitivity analysis using the Chemkin post-processor. Use DFT combined with TST theory to calculate the weak interactions, electronic structure and potential distribution during the molecular dynamics evolution process. This reveals the inhibition mechanism of the selected powder explosion suppressant and its evolution law in inhibiting gas / coal dust explosions. TST represents the transition state.

2. The molecular simulation method for suppressing the microscopic reaction of gas-coal dust composite explosion by powder explosion suppressant according to claim 1 is characterized in that: In step S1, the powder explosion suppressant crystal or cluster model, CH4 / O2 / N2 molecular model and coal macromolecular model are respectively established using Materials Studio software; The calculation formula for the volume parameter of the periodic boundary condition simulation box is: in, represents the volume parameter of the simulation box with periodic boundary conditions; Indicates the relative molecular mass of the gas / solid mixed reaction system in the system; represents the volume of the pyrolysis chamber; Indicates the mass of the gas / solid mixture in the system; represents Avogadro's constant.

3. The molecular simulation method for suppressing the microscopic reaction of gas-coal dust composite explosion by powder explosion suppressant according to claim 1 is characterized in that: In step S2, the geometry optimization and simulated annealing are to perform preliminary optimization on the powder explosion suppressant crystal or cluster model, CH4 / O2 / N2 molecular model and coal macromolecular model using the Dreiding force field of the Forcite module of Materials Studio software to obtain a molecular model with stable and reasonable atomic spacing, bond angles and dihedral angles; The initial model of the gas / solid mixed reaction system is a gas / solid system model of a periodic boundary condition simulation box and a mixture of powder explosion suppressant crystals or clusters, CH4 / O2 / N2 molecules and coal molecules established using the Amorphous Cell module of Materials Studio software.

4. The molecular simulation method for suppressing the microscopic reaction of gas-coal dust composite explosion by powder explosion suppressant according to claim 1, characterized in that: Step S4 specifically includes: performing NVT or NPT ensemble relaxation or energy minimization on the initial model based on the machine learning force field to obtain a balanced and uniform molecular dynamics model; The NVT ensemble is a canonical ensemble, which means it has a certain number of particles, volume and temperature, and only sets the Thermostat but not the Barostat; The NPT ensemble is an isothermal and isobaric ensemble, which means it has a certain number of particles, pressure and temperature, and the Thermostat and Barostat are set at the same time; The relaxation or energy minimization is to place the atoms and molecules in the reaction system in a balanced and uniform position under a suitable force field, thereby obtaining a molecular dynamics model close to reality.

5. The molecular simulation method for suppressing the microscopic reaction of gas-coal dust composite explosion by powder explosion suppressant according to claim 4 is characterized in that: In step S4, ab initio molecular dynamics simulation is a method based on bond order theory to describe the dynamic evolution of chemical bonds between atoms in a complex system. The continuity equation of bond order and interatomic distance is shown as follows: in, BO 、 ij 、 o 、 P bo are the bond order, interatomic distance, equilibrium bond length and empirical parameters between atoms respectively; this equation generates a differentiable potential energy surface through the characteristic transitions between σ, π and ππ bonds, thereby calculating the interatomic forces.

6. The molecular simulation method for suppressing the microscopic reaction of gas-coal dust composite explosion by powder explosion suppressant according to claim 4, characterized in that: In step S4, the total energy of the reaction system is the sum of all energy items, and the expression is as follows: in, 、 、 、 、 、 、 、 and They represent total energy, bond energy, overcoordination penalty, undercoordination stability, lone pair energy, valence angle energy, torsion angle energy, van der Waals energy and Coulomb energy respectively.

7. The molecular simulation method for suppressing the microscopic reaction of gas-coal dust composite explosion by powder explosion suppressant according to claim 1, characterized in that: In step S5, data post-processing and calculation are performed on the molecular dynamics simulation results, specifically including: The ChemTraYzer method is used to track the motion trajectories of molecules and atoms as they change during molecular dynamics simulations. The pyrolysis evolution, types, and structures of reactants, intermediates, and final products are analyzed. A detailed molecular reaction pathway is constructed, and reaction rate constants are calculated. Key elementary reactions and free radicals are extracted, and the coupling reaction patterns of the selected powdered explosion suppressant and gas / coal dust mixture are analyzed. By combining DFT with TST theory to calculate the weak interactions, electronic structure, and potential distribution during the molecular dynamics evolution process, the Gibbs free energy of reactants and products was accurately calculated and the activation energy was obtained. This allowed the molecular reaction pathway to be further optimized and refined, revealing the suppression characteristics and evolution process of the selected powder explosion suppressant on gas / coal dust explosions. Sensitivity analysis was performed using the Chemkin post-processor to reveal the influence of the selected powder explosion suppressant on the key elementary reactions and free radicals in the gas / coal dust composite explosion process.

Images