A model for inferring ore-forming fluid main channel based on particle filtering

Abstract

The application discloses a metallogenic fluid main channel inference model based on particle filtering. The model comprises the following steps: collecting relevant data of a target deposit, establishing an inference database, and defining a state sequence of spatial discrete elements; performing statistics on ore-bearing elements known in all exploration information, taking the known ore-bearing elements as an initial particle group of particle filtering, and generating a spatial channel path of each particle; establishing a metallogenic fluid ion state transfer model coupled with probability and velocity according to the spatial position of the particle and fluid dynamics; establishing a particle weight observation model based on a particle state transfer likelihood function, and dynamically updating the weight of the fluid particle; performing spatial sampling on the particle through a resampling algorithm, updating the weight particle, and outputting a maximum posterior probability path through particle number threshold judgment. The model combines multivariate data constraint and particle filtering algorithm to dynamically estimate the main channel path of the metallogenic fluid in a three-dimensional geological space, and fundamentally solves the problems of high cost and difficult sampling sample acquisition of the traditional method.

Description

TECHNICAL FIELD

[0001] The present application relates to a metallogenic fluid main channel inference model, in particular to a metallogenic fluid main channel inference model based on particle filtering, and belongs to the technical field of three-dimensional metallogenic research. BACKGROUND

[0002] The metallogenic fluid channel is closely related to the key path of metal element migration and mineralization enrichment and deposition, and is the bridge connecting "geological process" and "resource distribution". Inference of the metallogenic fluid main channel can reflect the migration direction of the fluid in the three-dimensional geological space, deepen the cognition of the time and space structure of the metallogenic system, and has important significance in the aspects of mineral resource prediction, revealing of metallogenic dynamics mechanism, and subsequent exploration engineering design and disaster prevention and control. However, the current traditional geological method for revealing the fluid channel needs high cost, and it is difficult to trace back the ancient fluid path and further quantify the dynamic behavior of the fluid, resulting in low success rate of subsequent mineral exploration, and it is difficult to realize high-precision prediction of concealed ore bodies. SUMMARY

[0003] In view of the problems existing in the prior art, a first purpose of the present application is to provide a metallogenic fluid main channel inference model based on particle filtering. The model combines multi-element data constraints, and uses a particle filtering algorithm to dynamically estimate the main channel path of the metallogenic fluid in the three-dimensional geological space, fundamentally solving the problems of high cost and difficult to obtain sampling samples of the traditional method, and maximizing the fusion of multi-type data constraints, realizing dynamic and low-cost simulation of the main channel of the metallogenic fluid in the three-dimensional geological space.

[0004] In order to realize the above technical purpose, the present application provides a metallogenic fluid main channel inference model based on particle filtering, characterized in that:

[0005] Step S1, collecting related mineralization data of the target ore deposit, establishing a metallogenic fluid main channel inference database, and defining a state sequence for the spatial discrete elements in the three-dimensional geological space;

[0006] Step S2, counting all known ore body elements in the three-dimensional geological space, as the initial particle group of the particle filtering, and generating the spatial channel path of each particle in the group;

[0007] Step S3, establishing a probability-velocity coupled metallogenic fluid ion state transfer model according to the spatial position of the particle in step S2 and the fluid dynamics;

[0008] Step S4, establishing a particle weight observation model based on the particle state transfer likelihood function, and dynamically updating the fluid particle weight according to the stress observation likelihood function;

[0009] Step S5, the particles are spatially sampled by a resampling algorithm, high-weight particles are copied in the process of updating the fluid particle weight, low-weight particles are removed, the particle number threshold is judged, the above particle weight updating is stopped, and the maximum a posteriori probability path is output, that is, the maximum a posteriori probability path is obtained.

[0010] The present application aims at the technical problems of high cost, low precision and difficulty in fully revealing the spatial distribution of the ancient fluid channel in the prior art metallogenic fluid main channel method, such as tracer test based on geological sampling samples, petrographic analysis and other methods, and adopts a particle filtering algorithm combined with multivariate data constraints, which can not only greatly reduce the inference cost of the metallogenic main channel and the ancient fluid channel, but also greatly improve the rapidity and accuracy of the inference result.

[0011] As a preferred scheme, the related mineralization data of the target deposit include mineralization distribution, drilling data, geological profile and plan.

[0012] As a preferred scheme, the process of defining the state sequence is that in the three-dimensional geological space, the fluid main channel is defined as a series of three-dimensional discrete unit coordinate points, and the three-dimensional discrete unit coordinates of the sequence are represented as:

[0013] Formula 1: ;

[0014] In formula 1, is a three-dimensional space discrete unit coordinate, representing the position of the fluid at time .

[0015] As a preferred scheme, the generation process of the space channel path of each particle is that the initial particle group contains N particles, and the flow channel path of the metallogenic fluid represented by each particle is:

[0016] Formula 2: ;

[0017] In formula 2, represents the initial flow path probability of the corresponding particle, represents the possible initial flow position of the corresponding particle generated according to prior knowledge, characterizes the initial fluid flow position weight, and the initial weight is subject to a uniform distribution, that is, .

[0018] As a preferred scheme, the establishment process of the probability-velocity coupled metallogenic fluid ion state transition model is:

[0019] Step S3-1, a fluid particle group state transition model is constructed according to the structural form factor and the permeability gradient direction factor related to fluid migration;

[0020] Step S3-2, taking the spatial position of each particle in the initial particle group in step S2 as the end point of each channel path, measuring the current fluid migration position of the initial particle group according to the fluid particle group state transition model of the previous time fluid migration position , and the current fluid migration position is obtained.

[0021] For each particle , there is a three-dimensional geological space fluid migration channel path Since fluid migrates from the deep to the shallow of the geological space, the initial particle group can be regarded as the end point of each channel path, and thus the previous time fluid migration path position of the current initial particle group can be predicted by the fluid dynamics model.

[0022] As a preferred scheme, the expression of the fluid particle group state transition model is:

[0023] Formula 3: ;

[0024] The current fluid migration position and the previous time fluid migration position are related by:

[0025] Formula 4: ;

[0026] In formula 3 and formula 4, represents the permeability gradient direction, and its physical meaning indicates that the fluid migration direction is always from the low-permeability area to the high-permeability area; represents the fluid migration factor related to the tectonic form, which is subject to a probability distribution; represents a random disturbance term, which is subject to a Gaussian distribution; , and respectively represent the weight coefficients of the corresponding constraint terms; represents the time interval of fluid transfer from the previous time position to the next time position; represents the noise effect in the fluid migration process.

[0027] In the present application, the fluid migration factor related to the tectonic form includes the characteristics of the tectonic structural surface dip angle and fluctuation, and the physical meaning of this constraint indicates that the fluid migration speed is closely related to the tectonic form, for example, the fluid migration speed is small in the area with a small tectonic structural surface dip angle, and the fluid migration speed is large in the area with a large tectonic structural surface dip angle, and the fluctuation of the structural surface can also affect the fluid flow speed; , and can be determined according to the geological conditions of the target research area, and can be constants input by the user.

[0028] As a preferred scheme, the particle state transition likelihood function is constructed by combining the tectonic stress characteristic data as direct observation data of fluid migration position , and its expression is

[0029] Formula 5: ;

[0030] In formula 5, represents the current state observation data, characterizes the stress value of the corresponding voxel, is the maximum value of the regional stress value, is the variance of the likelihood function distribution, represents a probability normalization constant.

[0031] During the tectonic evolution process, fractures are generated in the three-dimensional geological space under the action of stress, which directly provides favorable space for fluid migration and affects the spatial position of fluid migration. Therefore, when constructing the particle weight observation model, the tectonic stress characteristic data should be combined as direct observation data of fluid migration position.

[0032] As a preferred scheme, the expression of the particle weight observation model is

[0033] Formula 6: ;

[0034] In formula 6, represents the particle weight observation value of the corresponding position.

[0035] Since the above calculation particle group includes all ore-containing voxels in the three-dimensional geological space, the fluid channel path represented thereby may be distributed throughout the three-dimensional geological space, making it difficult to observe the main channel position distribution rule of the fluid. Therefore, the particle group needs to be resampled, and particles with low weight influence are discarded to observe the overall characteristic trend of fluid migration in the space.

[0036] As a preferred scheme, before the particle is spatially sampled, the number of effective particles needs to be calculated, and the process is as follows:

[0037] Formula 7: ;

[0038] In formula 7, is the number of particles in the initial particle group, i.e. the number of all ore-containing voxels in the three-dimensional space.

[0039] As a preferred scheme, the process of updating the fluid particle weight is that after the high-weight particles are reserved, the above fluid path inference process is repeated for the next particle path inference, and the weight is reset to

[0040] Formula 8: ;

[0041] The judgment process of the high-weight particles and the low-weight particles is as follows: a threshold value is set When , the high-weight particles are reserved according to the weight distribution rule.

[0042] The above threshold value is set for a user, when the value is large, the obtained fluid main channel distribution will highlight the local path characteristics, and when the value is set small, the output main channel highlights the overall characteristics of the ore-forming fluid main channel.

[0043] As a preferred scheme, the output process of the maximum posterior probability path is as follows: the particle with the largest weight is selected for output, and the high-weight particles are connected in the order of state transition inference, so as to form the main channel path of the ore-forming fluid flowing in the three-dimensional geological space.

[0044] Compared with the prior art, the technical scheme of the present application has the beneficial technical effects that:

[0045] 1) The ore-forming fluid main channel inference model provided by the present application combines multiple data constraints such as structural morphology, permeability gradient direction and stress data, adopts a particle filtering algorithm, and performs three-dimensional geological space fluid migration main channel inference. This method avoids the high-cost inference of traditional methods based on geological sampling samples using tracers and petrographic fluid channel analysis, and solves the limitation that the traditional method is difficult to trace back the ancient fluid path.

[0046] 2) In the technical scheme provided by the present application, based on the fusion of multiple data constraints in the model, dynamic and low-cost high-precision inference of the ore-forming fluid main channel in the three-dimensional geological space is realized, especially for the accurate prediction of the spatio-temporal structure of the ore-forming system, which provides a reliable model basis for subsequent exploration engineering and mineral resource quantity prediction. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The fluid main channel spatial distribution characteristics of the inference model provided in Embodiment 1 of the present application are shown in the following table:

[0048] Figure 2 The fluid main channel statistical characteristics of the inference model provided in Embodiment 1 of the present application are shown in the following table:

[0049] Wherein, Figure 2 (a) is a relationship diagram between the body element through which the fluid main channel passes and the corresponding fracture surface model normal, Figure 2 (b) is a cumulative frequency relationship diagram between the body element through which the fluid main channel passes and the corresponding fracture surface model normal. DETAILED DESCRIPTION

[0050] The technical solutions of the present application will be further described in detail below in combination with specific embodiments and drawings. In order to facilitate understanding of the present application, the present application will be described more fully and in detail in combination with the drawings and preferred embodiments of the present application. It should be noted that the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present application.

[0051] Embodiment 1

[0052] The present embodiment provides a particle filtering-based ore-forming fluid main channel inference model, and the modeling process is as follows:

[0053] Step S1, collect and organize the related mineralization data of the target deposit, such as mineralization distribution, drilling data, geological profile, plan view and other data, form an ore-forming fluid main channel inference database, and realize state space modeling in three-dimensional geological space, and define the state sequence of spatial discrete elements;

[0054] Collect and organize the related mineralization data of the target deposit, such as mineralization distribution, drilling data, geological profile, plan view and other data, form an ore-forming fluid main channel inference database. In addition, the coordinates of three-dimensional geological space discrete units are determined, and the fluid flow state space is established, that is, in the three-dimensional geological space, the fluid main channel is defined as a series of three-dimensional discrete unit coordinate point sequences, which can be represented as a series of three-dimensional discrete unit coordinates:

[0055] Formula 1: ;

[0056] In formula 1, is the three-dimensional space discrete unit coordinate, which represents the position of the fluid at time .

[0057] Step S2, count all ore-bearing elements in the three-dimensional geological space, and use them as the initial particle group of the particle filtering, and generate the spatial position of each particle, and store it separately, so as to generate the spatial channel path of each particle. In the main channel inference process, each particle group represents a possible fluid migration channel path, which can be represented by coupling the ore-forming fluid state transition model and the particle weight observation model;

[0058] Based on the fluid main channel inference database, count all ore-bearing elements and their spatial position coordinates in the three-dimensional geological space, and use them as the initial particle group of the particle filtering algorithm. The number of particles contained in the initial particle group is , and each particle can represent a flow channel path of the ore-forming fluid, that is, it can be represented as:

[0059] Formula 2: ;

[0060] In formula 2, represents the initial flow path probability of the corresponding particle, represents the initial flow position of the corresponding particle according to prior knowledge, characterizes the initial fluid flow position weight, and the initial weight is subject to a uniform distribution, that is, .

[0061] Step S3, based on the spatial position of the particle group, a probability-velocity coupled metallogenic fluid ion state transition model is established in combination with a fluid dynamics model. The model mainly considers the influence of factors such as tectonic form and permeability gradient direction on the fluid migration velocity, so in the state transition model, in addition to the spatial position term related to the fluid transition state, a velocity model term is also included to consider the fluid migration influencing factors. The state transition probability is fused into the state transition velocity model in the form of state transition probability, so as to deduce the possible position of the fluid particle in space at the previous moment;

[0062] For each particle above, there is a three-dimensional geological space fluid migration channel path. Since the fluid migrates from the deep to the shallow of the geological space, the initial particle group can be regarded as the end point of each channel path, so the fluid migration path position of the current initial particle group at the previous moment can be predicted by the fluid dynamics model, and the current fluid migration position is related to the fluid migration position at the previous moment, and the relationship can be expressed as:

[0063] Formula 3: ;

[0064] wherein, represents the fluid particle group state transition velocity model, represents the time interval of fluid transition from the position at the previous moment to the position at the next moment. In the inference process, it is assumed that the time interval is small, so as to completely simulate the spatial position change process of fluid migration, and therefore it can be regarded as a unit time interval, represents the noise influence in the fluid migration process, so as to simulate the uncertainty of fluid state transition in the three-dimensional geological space;

[0065] For the fluid particle group state transition velocity model, the tectonic form factor and the permeability gradient direction factor closely related to fluid migration can be expressed as:

[0066] Formula 4: ;

[0067] In formula 4, represents the permeability gradient direction, and its physical meaning indicates that the fluid migration direction is always from the low-permeability area to the high-permeability area. represents the factor related to the structure morphology, which includes the characteristics of the structure surface dip angle, relief, etc. It is represented in the form of probability distribution. The physical meaning of this constraint indicates that the fluid migration velocity is closely related to the structure morphology. For example, the fluid migration velocity is small in the area with small structure surface dip angle, and the fluid migration velocity is large in the area with large structure surface dip angle. The relief of the structure surface can also affect the fluid migration velocity; represents the random disturbance term, which considers the random factors in the fluid migration process. This term is subject to Gaussian distribution; 、 、 respectively represent the weight coefficients of the corresponding constraint terms. These terms need to be determined according to the geological conditions of the target research area and can be constants input by the user;

[0068] permeability gradient direction It is usually related to the geological prior knowledge and its value is closely related to the sampling in the corresponding area by previous researchers. The factor related to the structure morphology affecting fluid migration Usually includes the characteristics such as structure morphology. However, in the implementation process, due to the nonlinear relationship between the structure morphology characteristics and the fluid flow velocity, it is difficult to directly give the nonlinear analytical expression of the two. Therefore, it is assumed that the value of the fluid flow velocity remains unchanged in a very small time interval Therefore, its velocity value can be given by the Fokker-Planck equation, that is:

[0069] Equation 5: ;

[0070] In equation 5, , represents a small time interval, and represent the spatial positions of the fluid at the current time and , respectively; represents the diffusion coefficient related to the random fluctuation;

[0071] The probability distribution of the fluid flow position in the above formula can be represented by the fluid transfer probability, that is:

[0072] Equation 6: ;

[0073] In equation 6, represents the domain point position of point ;

[0074] Subsequently, the factor related to the structure morphology affecting fluid migration can be represented by the transfer probability in the form of probability distribution. For example, for the structure surface dip angle, its probability distribution can be represented as:

[0075] Equation 7: ;

[0076] In Equation 7, is the variance of the probability distribution, is the maximum dip value of the structure surface, represents the dip value of the corresponding voxel, is a normalization constant. This term of the probability distribution expression ensures that when the dip of the corresponding voxel is large, the fluid transfer probability is greater, and the fluid flow velocity will be greater;

[0077] Step S4, a particle state transition likelihood function is established to represent the particle weight observation model. In this regard, the stress feature data related to the structure is considered to be expressed in the form of a probability likelihood function, and the fluid particle weight is dynamically updated according to the stress observation likelihood function;

[0078] In the process of geological evolution, the structure produces fractures in the three-dimensional geological space under the action of stress, which directly provides a favorable space for fluid migration and affects the spatial position of fluid migration. Therefore, when building the particle weight observation model, the stress feature data of the structure should be combined as the direct observation data of the fluid migration position, and the likelihood function is built here , which represents the difference between the fluid position and the observed stress data in the fluid main channel inference process, that is, it can be expressed as:

[0079] Equation 8: ;

[0080] wherein, represents the observation data of the current state (the position of the fluid path inference), characterizes the stress value of the corresponding voxel, is the maximum value of the regional stress value, is the variance of the likelihood function distribution, represents a probability normalization constant;

[0081] The difference between the current state and the observed stress data represented by the likelihood function can be further used to build the particle weight observation model:

[0082] Equation 9: ;

[0083] Equation 9 represents that the particle weight observation is a recursive function. Each time the fluid state position is dynamically updated, the particle weight needs to be dynamically updated to solve the particle weight observation at the next time;

[0084] In the inference process, the stress size of the discrete body element can be difficult to use quantitative numerical characterization, but the overall morphological characteristics of the structural surface can be used to infer the tensile and compressive stress changes during the geological activity process. Therefore, in the specific implementation process, the stress value of the discrete body element is represented by coupling the indexes of the strike and the dip two geological elements, and the strike expansion and the dip expansion indexes are provided to characterize the stress change of the body element. The two indexes are solved by the third derivative of the local implicit function of the structural surface, and finally the stress of the discrete body element can be represented as the coupling form of the strike expansion and the dip expansion indexes, that is ;

[0085] Step S5, the particles are spatially sampled by a resampling algorithm, and high-weight particles are copied according to the updating process of the particle weight, and low-weight particles are discarded. The particle weight needs to be dynamically updated and reset after each sampling;

[0086] Since the above-mentioned particle group includes all ore-bearing body elements in the three-dimensional geological space, the fluid channel path represented thereby can be distributed throughout the three-dimensional geological space, and it is difficult to observe the main channel position distribution rule of the fluid. Therefore, the particle group needs to be resampled, and the particles with low influence weight are discarded to observe the overall characteristic trend of the fluid migration in the space.

[0087] Before resampling, the number of effective particles is calculated, that is:

[0088] Formula 10: ;

[0089] In formula 10, is the number of particles in the initial particle group, that is, the number of all ore-bearing body elements in the three-dimensional space, represents the particle weight observation value at the corresponding position. Then, according to the particle number threshold, when , the high-weight particles are retained according to the weight distribution rule, and the low-weight particles are discarded, wherein is a threshold value set by the user. When the value is large, the main channel distribution of the fluid obtained will highlight the local path characteristics, and when the value is set to be small, the main channel output will highlight the overall characteristics of the main channel of the ore-forming fluid;

[0090] After retaining the high-weight particles, the above-mentioned fluid path inference process is repeated to infer the next particle path, and the weight is reset to:

[0091] Formula 11: ;

[0092] In this way, the real-time updating of the particle weight observation model is realized;

[0093] Step S6, stopping the particle weight update by the particle number threshold judgment, and outputting the maximum posterior probability path as the main channel of the ore-forming fluid in the three-dimensional geological space;

[0094] When the fluid path inference process of all particles, i.e. the resampling process, is completed, the maximum posterior probability path is output, i.e. the particle with the maximum weight is selected and output, and the high-weight particles are connected in the order of state transition inference, so as to form the main channel path of the ore-forming fluid flowing in the three-dimensional geological space.

[0095] In order to verify the performance of the above ore-forming fluid main channel inference model, the present application takes a certain mining area as an example to simulate the spatial distribution position of the deep fluid main channel, i.e. the distribution characteristics.

[0096] In this process, due to the ore-controlling characteristics of the structure of the mining area, the ore body of the mining area is closely related to the structure, therefore, in the database construction process, the structural surface related data of the fault zone is collected for model construction, and the morphological feature index is extracted, in addition, the permeability gradient direction data and the like are collected, the particle swarm is initialized by using the particle filtering algorithm for the ore body containing element in the database, then the fluid particle state transition model and the particle weight observation model are constructed, the spatial channel path position of the fluid migration is inferred, and the main channel of the fluid migration is extracted by the resampling algorithm. According to the inference of the above model, the spatial distribution of the fluid main channel of the mining area is as shown in Figure 1 , and the statistical relationship between the path and the depth of the ore-forming fluid migrating along the main channel to the shallow part is as shown in Figure 2 .

[0097] Figure 1 characterize the distribution characteristics of the fluid main channel in the three-dimensional space of the inference model, from the figure, it can be seen that the method of the present application can effectively combine the morphological characteristics of the fracture surface and the observed mineralization distribution, and infer the reliable position of the fluid main channel by the particle filtering algorithm, clearly visualize the flow characteristics of the fluid in the three-dimensional space, and can clearly observe the branch channel of the fluid in space, so as to further infer the flux size of the fluid channel. Figure 2 show the statistical characteristics of the fluid main channel of the inference model provided by the present embodiment, wherein, Figure 2 (a) is the relationship between the body element passed by the fluid main channel and the corresponding fracture surface model normal, which shows the influence of the morphological characteristics of the fracture surface on the fluid migration, Figure 2 (b) is the cumulative frequency relationship between the body element passed by the fluid main channel and the corresponding fracture surface model normal, which shows that the fluid main channel inferred by the present method is beneficial to subsequent ore prospecting application.

[0098] It can be known from the above description that the ore-forming fluid main channel inference model provided by the application effectively overcomes the limitations of high cost and difficulty in inferring the main fluid channel of the traditional method, can not only fuse the constraints of multiple data, but also can realize dynamic and low-cost fluid main channel simulation, and ensures high-precision inference of the fluid main channel.

[0099] The above is the preferred embodiment of the application. It should be noted that those skilled in the art can make some improvements and refinements without departing from the principles of the application, and these improvements and refinements should also be considered as the protection scope of the application.

Claims

1. A method for inferring ore-forming fluid main channel based on particle filtering, characterized in that: Step S1, collecting relevant mineralization data of the target deposit, establishing an ore-forming fluid main channel inference database, and defining a state sequence for spatial discrete elements in a three-dimensional geological space; Step S2, counting all known ore-bearing elements in the three-dimensional geological space, as the initial particle group of particle filtering, and generating the spatial channel path of each particle in the group; Step S3, establishing a probability-velocity coupled ore-forming fluid particle state transition model according to the spatial position of the particle in step S2 and fluid dynamics; Step S4, establishing a particle weight observation model based on the particle state transition likelihood function, and dynamically updating the fluid particle weight according to the stress observation likelihood function; Step S5, spatially sampling the particles through a resampling algorithm, copying high-weight particles and removing low-weight particles in the process of updating the fluid particle weight, and stopping the above particle weight updating through a particle number threshold judgment, and outputting the maximum a posteriori probability path. 2.The method of claim 1, wherein the relevant mineralization data of the target deposit includes mineralization distribution, drilling data, geological profile, and plan view; and the process of defining a state sequence is that in the three-dimensional geological space, the fluid main channel is defined as a series of three-dimensional discrete element coordinate points, and the three-dimensional discrete element coordinates of the series are represented as: Formula 1: ; In formula 1, is a three-dimensional spatial discrete unit coordinate, representing the position of the fluid at time . 3.The method of claim 1, wherein the process of generating the spatial channel path of each particle is that the initial particle group contains N particles, and the flow channel path of the ore-forming fluid represented by each particle is: Formula 2: ; In formula 2, denotes the initial flow path probability of the corresponding particle, denotes the initial flow position of the corresponding particle possible according to prior knowledge, characterizes the initial fluid flow position weight, which is subject to a uniform distribution, i.e. . 4.The method of claim 1, wherein the process of establishing the probability-velocity coupled ore-forming fluid particle state transition model is: Step S3-1, constructing a fluid particle group state transition model according to the structural form factor related to fluid migration and the permeability gradient direction factor; Step S3-2, taking the spatial position of each particle in the initial particle group in step S2 as the end point of each channel path, measuring the current fluid migration position of the initial particle group according to the fluid particle group state transition model described above the fluid migration position of the previous time . 5.The method of claim 4, wherein the expression of the fluid particle group state transition model is: Formula 3: ; the current fluid migration position the relationship with the fluid migration position of the previous time instant is given by the equation: Formula 4: ; in formula 3 and formula 4, represents the permeability gradient direction, and its physical meaning indicates that the fluid migration direction is always from the low permeability zone to the high permeability zone; represents a factor affecting fluid migration related to the structural form, and is subject to a probability distribution; represents a random disturbance term, which is subject to a Gaussian distribution; , and respectively represent the weight coefficients of the corresponding constraint terms; represents the time interval of fluid transfer from the position at the previous time to the position at the next time; represents the noise influence in the fluid migration process.

6. The method for inferring the main channel of ore-forming fluid based on particle filtering according to claim 1, characterized in that: the construction process of the particle state transition likelihood function is as follows: combining tectonic stress characteristic data as direct observation data of fluid migration position, a likelihood function is constructed. Its expression is: Formula 5: ; in formula 5, denotes the current state observation data, denotes the stress value of the corresponding voxel, is the maximum value of the regional stress value, is the variance of the likelihood function distribution, denotes a probability normalization constant. 7.The method of claim 1, wherein the expression of the particle weight observation model is: Formula 6: ; In formula 6, represents the particle weight observation value at the corresponding position. 8.The method of claim 1, wherein before the spatial sampling of the particles, the number of effective particles is calculated, and the process is: Formula 7: ; In formula 7, is the number of particles in the initial particle group, that is, the number of all ore-containing element groups in the three-dimensional space. 9.The method of claim 8, wherein after the high-weight particles are reserved, the resampling algorithm is repeated to spatially sample the particles for the next particle path inference, and the weight is reset to: Formula 8: ; The judgment process of the high-weight particles and the low-weight particles is: setting a threshold value When The high-weight particles are reserved according to the weight distribution rule.

10. The method of claim 1, wherein the output process of the maximum a posteriori probability path is that the particle with the largest weight is outputted, and the particles with high weights are connected in the order of state transition inference to form the main channel path of the ore-forming fluid flowing in the three-dimensional geological space.

Images