A method and system for sorting and controlling finished petroleum proppant products based on machine learning

Abstract

The present application discloses a petroleum proppant finished product sorting control method and system based on machine learning, which relates to the technical field of petroleum proppant production. The method and system extract shape characteristic parameters; construct a three-dimensional geometric model of an airflow classifier to generate three-dimensional trajectories of spherical particle samples and non-spherical particle samples; obtain the shape characteristic parameters of the proppant particles to be classified, find the petroleum proppant particle samples with the highest shape similarity to the proppant particles, and calculate the speed deviation; calculate the comprehensive deviation index based on the shape characteristic parameters and speed deviation of the proppant particles, and if the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles are directly classified, otherwise they are processed twice; use an intelligent correction model to predict the best correction parameters, correct the proppant particles, and send the corrected proppant particles into the airflow classifier; construct an optimization objective function, and search for the optimal process parameters; and achieve quality control of the petroleum proppant finished product sorting.

Description

Technical Field

[0001] The present application relates to the technical field of petroleum proppant production, and in particular to a petroleum proppant finished product sorting control method and system based on machine learning. Background Art

[0002] Oil proppants are a key material used in oil extraction, mainly used in the fracturing process. They are high-strength, high-sphericity ceramic microspheres that can support fractures and maintain fracture conductivity, thereby improving recovery. Common proppant materials include alumina, mullite, and silica sand.

[0003] Chinese patent publication number CN114289158A discloses a petroleum fracturing proppant and its preparation process, the preparation process steps are: (1) conveying ore to a crusher, crushing the ore particles, then lifting it to a temporary storage bin, and then entering a double-roll crusher; (2) the ore is secondary crushed in the double-roll crusher, and after crushing, it is screened. Larger particles continue to enter the double-roll crusher for crushing, and smaller particles are conveyed to a shaping sand making machine for fine crushing, strength optimization, and rounding to obtain quartz sand particles; (3) the quartz sand particles after fine crushing and shaping are preliminarily classified; (4) quartz sand particles with a particle size of 1-0.106 mm enter a swing screen for fine screening.

[0004] In the existing petroleum proppant finished product sorting process, the irregular shape of the proppant particles deviates from the ideal spherical shape, resulting in low classification accuracy and stability, inaccurate particle size distribution, and low quality of the finished proppant particles. Summary of the invention

[0005] The present application aims to solve at least one of the technical problems in the related art to a certain extent. To this end, one purpose of the present application is to propose a method and system for controlling the sorting of petroleum proppant finished products based on machine learning, so as to achieve quality control of the sorting of petroleum proppant finished products.

[0006] One aspect of the present application provides a method for controlling the sorting of petroleum proppant products based on machine learning, comprising:

[0007] Step S100: obtaining a petroleum proppant particle sample, and extracting shape characteristic parameters of the petroleum proppant particle sample, wherein the petroleum proppant particle sample includes a spherical particle sample and a non-spherical particle sample, and the shape characteristic parameters include roundness, sphericity and equivalent diameter;

[0008] Step S200: constructing a three-dimensional geometric model of the airflow classifier to generate three-dimensional trajectories of spherical particle samples and non-spherical particle samples during the classification process;

[0009] Step S300: for the proppant particles to be classified, obtain their shape characteristic parameters, find the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified, and calculate the velocity deviation between the proppant particles to be classified and the spherical particle sample based on the three-dimensional trajectory;

[0010] Step S400: Based on the shape characteristic parameters and velocity deviation of the proppant particles, the comprehensive deviation index is calculated, and a comprehensive deviation index threshold is preset. If the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification, otherwise, they are subjected to secondary processing;

[0011] Step S500: using the trained intelligent correction model to predict the best correction parameters, correcting the proppant particles undergoing secondary processing until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and sending the corrected proppant particles to an airflow classifier for classification;

[0012] Step S600: Real-time detection of the actual particle size distribution of the classification result of the airflow classifier, calculation of its deviation from the target particle size distribution, calculation of the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, construction of the optimization objective function, and search for the optimal process parameters.

[0013] The specific method of obtaining the petroleum proppant particle sample and extracting the shape characteristic parameters of the petroleum proppant particle sample is:

[0014] Step S110: collecting spherical particle samples and non-spherical particle samples of petroleum proppant particles to obtain petroleum proppant particle samples, and scanning a two-dimensional contour image of the petroleum proppant particle samples;

[0015] Step S120: Obtaining the maximum inscribed circle diameter of the petroleum proppant particle sample according to the two-dimensional contour image and the minimum circumscribed circle diameter , calculate the sphericity of petroleum proppant particle samples ;

[0016] Step S130: Find the corresponding horizontal coordinate position on the Krumbien / Sloss plate according to the sphericity, compare the two-dimensional contour image with standard contour lines of different roundness levels, and select the most similar standard contour line to obtain the roundness C of the petroleum proppant particle sample;

[0017] Step S140: Determine the actual volume V of the petroleum proppant particle sample and calculate the equivalent diameter of the petroleum proppant particle sample ;

[0018] Step S150: The roundness, sphericity and equivalent diameter constitute the shape characteristic parameters of the petroleum proppant particle sample.

[0019] The specific method for constructing the three-dimensional geometric model of the airflow classifier is:

[0020] Step S210: Obtaining the structural parameters and process parameters of the airflow classifier, the structural parameters include the feed port diameter , Exit height , the process parameters include feed rate, air flow velocity, and temperature T;

[0021] Step S220: determining the computational domain of the gas phase-particle phase numerical model, including the feed port, classification cavity and outlet of the gas flow classifier; designing a gas phase control equation to solve the gas phase velocity vector at any position and time in the computational domain; designing a particle phase control equation to solve the particle velocity of the petroleum proppant particle sample; the gas phase control equation and the particle phase control equation constitute the gas phase-particle phase numerical model;

[0022] Step S230: Determine the gas phase density according to the ideal gas state equation, the actual temperature and pressure p , according to the dynamic viscosity of the air flow, determine the gas phase dynamic viscosity ; Based on the particle density of petroleum proppant particle samples and equivalent diameter Calculate particle mass , the aerodynamic drag is calculated based on the relative velocity between the petroleum proppant particle sample and the airflow , calculate the gravity based on the particle mass and gravitational acceleration ;

[0023] Step S240: constructing a three-dimensional geometric model of the airflow classifier, dividing the feed inlet, classification cavity and outlet of the airflow classifier into different areas and generating a grid.

[0024] The specific method for generating the three-dimensional trajectories of the spherical particle samples and the non-spherical particle samples during the classification process is:

[0025] Step S250: Simulate the classification process of the petroleum proppant particle sample in the airflow classifier, solve the gas phase-particle phase numerical model, and obtain the gas phase velocity vector at any position and time in the calculation domain , obtain the particle velocity of the petroleum proppant particle sample at any position and time;

[0026] Step S260: obtaining the particle position of the petroleum proppant particle sample at each moment according to the particle velocity, and drawing the three-dimensional trajectory of the petroleum proppant particle sample during the classification process;

[0027] Step S270: Generate three-dimensional trajectories of the spherical particle samples and the non-spherical particle samples during the classification process and store them in a database.

[0028] The specific method for obtaining the shape characteristic parameters of the proppant particles to be classified and finding the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified is:

[0029] Step S310: For the proppant particles to be classified, obtain their two-dimensional contour images and define their shape characteristic parameters as For the petroleum proppant particle samples in the database, the shape characteristic parameters are defined as , calculate the weighted Euclidean distance between the shape characteristic parameters of the proppant particles to be classified and the shape characteristic parameters of the petroleum proppant particle samples in the database , calculate the inverse of the weighted Euclidean distance to get the shape similarity ;in, are the sphericity, roundness and equivalent diameter of the proppant particles to be classified, are the sphericity, roundness and equivalent diameter of the jth petroleum proppant particle sample in the database, respectively;

[0030] Step S320: taking the three-dimensional trajectory corresponding to the petroleum proppant particle sample having the highest shape similarity to the proppant particle to be classified as the most similar trajectory of the proppant particle to be classified.

[0031] The specific method for calculating the velocity deviation between the proppant particles to be classified and the spherical particle samples based on the three-dimensional trajectory is:

[0032] Step S330: Find a spherical particle sample in the database that is closest to the equivalent diameter of the proppant particle to be classified, and use the three-dimensional trajectory of the spherical particle sample as the reference three-dimensional trajectory;

[0033] Step S340: A three-dimensional coordinate system (x, y, z) is established based on the grid of the computational domain, wherein the z-axis direction is the axial direction of the airflow classifier, and the computational domain is equally divided into N grids in the z-axis direction, and the size of each grid is ;

[0034] Step S350: define the z-axis coordinate of the sampling point as the midpoint of each grid, that is, the expression of the z-axis coordinate of the ith sampling point is , where i=1,2,...,N;

[0035] Step S360: For the most similar trajectory and the reference three-dimensional trajectory, obtain the particle velocity at each sampling point respectively, and obtain the particle velocity at the i-th sampling point of the most similar trajectory , get the particle velocity of the benchmark three-dimensional trajectory at the i-th sampling point ;in, are the component velocities of the particle velocity in the x, y, and z directions of the most similar trajectory at the i-th sampling point, are the component velocities of the particle velocity in the x, y, and z axes of the reference three-dimensional trajectory at the i-th sampling point;

[0036] Step S370: Calculate the speed deviation between the most similar trajectory and the reference three-dimensional trajectory at each sampling point ;

[0037] Step S380: Calculate the average speed deviation between the most similar trajectory and the reference three-dimensional trajectory over the entire trajectory , the average velocity deviation is taken as the velocity deviation between the proppant particle and the spherical particle sample .

[0038] The method of calculating the comprehensive deviation index based on the shape characteristic parameters and velocity deviation of the proppant particles and presetting the comprehensive deviation index threshold is as follows: if the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification; otherwise, the proppant particles are subjected to secondary treatment as follows:

[0039] Step S410: Calculate the comprehensive deviation index by combining the shape characteristic parameters of the proppant particles and the velocity deviation between the proppant particles and the spherical particle samples. ;

[0040] Step S420: Preset comprehensive deviation index threshold , compare the size of the comprehensive deviation index with the comprehensive deviation index threshold. If the comprehensive deviation index of the proppant particle is less than or equal to the comprehensive deviation index threshold, the proppant particle directly enters the air flow classifier for classification; if the comprehensive deviation index of the proppant particle is greater than the comprehensive deviation index threshold, the proppant particle undergoes secondary processing.

[0041] The specific method of using the trained intelligent correction model to predict the optimal correction parameters, correcting the proppant particles undergoing secondary processing until the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and sending the corrected proppant particles to the airflow classifier for classification is:

[0042] Step S510: obtaining shape characteristic parameters and corresponding optimal correction parameters of non-spherical particles in historical data, taking the shape characteristic parameters of non-spherical particles as input data and the corresponding optimal correction parameters as output data to form training samples, and constructing a training set;

[0043] Step S520: using a deep neural network as the basic framework for training an intelligent correction model, the intelligent correction model includes an input layer, a hidden layer, and an output layer, using mean square error as a loss function, and using the Adam optimization algorithm to optimize model parameters;

[0044] Step S530: Input the training set into the deep neural network, perform forward propagation, calculate the output data and loss function of each group of training samples, calculate the gradient of the loss function to the model parameters through the back propagation algorithm, and use the Adam optimization algorithm to update the model parameters to minimize the loss function;

[0045] Step S540: When the loss function between the predicted optimal correction parameter and the actual optimal correction parameter reaches convergence, the training is completed and a trained intelligent correction model is obtained;

[0046] Step S550: For the proppant particles to be subjected to secondary treatment, the trained intelligent correction model is used to predict the optimal correction parameters of the proppant particles, the proppant particles are corrected based on the optimal correction parameters, and the comprehensive deviation index of the corrected proppant particles is calculated. If the comprehensive deviation index is still greater than the comprehensive deviation index threshold, the proppant particles are repeatedly corrected until the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, thereby meeting the classification requirements of the airflow classifier;

[0047] Step S560: Send the corrected proppant particles to an air flow classifier for classification.

[0048] The specific method of detecting the actual particle size distribution of the classification result of the airflow classifier in real time, calculating the deviation value between the actual particle size distribution and the target particle size distribution, calculating the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, constructing the optimization objective function, and searching for the optimal process parameters is as follows:

[0049] Step S610: Real-time monitoring of the actual particle size distribution of the classification results of the airflow classifier , calculate the actual particle size distribution and the target particle size distribution The deviation between ;

[0050] Step S620: Real-time statistics of shape characteristic parameters and comprehensive deviation indexes of proppant particles entering the airflow classifier, and calculation of the average value of the shape characteristic parameters And the average value of comprehensive deviation index ;

[0051] Step S630: define the deviation value, shape feature parameter average value and comprehensive deviation index average value The optimization objective function , process parameter k is the optimization variable, defining the process parameter range constraint;

[0052] Step S640: using a genetic algorithm to solve the optimization objective function and obtain the optimal process parameters.

[0053] One aspect of the present application provides a petroleum proppant finished product sorting control system based on machine learning, comprising:

[0054] A feature data acquisition module is used to acquire a petroleum proppant particle sample and extract shape feature parameters of the petroleum proppant particle sample, wherein the petroleum proppant particle sample includes a spherical particle sample and a non-spherical particle sample, and the shape feature parameters include roundness, sphericity and equivalent diameter;

[0055] A three-dimensional trajectory generation module is used to construct a three-dimensional geometric model of the airflow classifier and generate three-dimensional trajectories of spherical particle samples and non-spherical particle samples during the classification process;

[0056] The velocity deviation calculation module is used to obtain the shape characteristic parameters of the proppant particles to be classified, find the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified, and calculate the velocity deviation between the proppant particles to be classified and the spherical particle sample based on the three-dimensional trajectory;

[0057] A particle classification processing module is used to calculate the comprehensive deviation index of the proppant particles based on their shape characteristic parameters and velocity deviation, and preset a comprehensive deviation index threshold. If the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification, otherwise they are subjected to secondary processing;

[0058] The secondary correction classification module is used to use the trained intelligent correction model to predict the best correction parameters, correct the proppant particles undergoing secondary processing until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and send the corrected proppant particles to the airflow classifier for classification;

[0059] The process parameter optimization module is used to detect the actual particle size distribution of the classification results of the airflow classifier in real time, calculate its deviation value from the target particle size distribution, calculate the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, construct the optimization objective function, and search for the optimal process parameters.

[0060] One aspect of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in a method for controlling the sorting of finished petroleum proppant products based on machine learning are implemented.

[0061] One aspect of the present application provides a readable storage medium storing a computer program, wherein the computer program is suitable for loading by a processor to execute steps in a method for controlling the sorting of finished petroleum proppant products based on machine learning.

[0062] Compared with the prior art, the petroleum proppant finished product sorting control method and system proposed in this application based on machine learning have the following advantages:

[0063] This application comprehensively considers various shape characteristic parameters of proppant particles, including sphericity, roundness and equivalent diameter, to provide a more comprehensive characterization and quantitative evaluation of particle shape.

[0064] This application introduces a gas phase-particle phase numerical model, which simulates the movement trajectory of proppant particles in an airflow classifier, predicts the classification behavior of particles with different shape characteristics, and provides theoretical support for optimizing classification.

[0065] This application proposes the concept of comprehensive deviation index, which combines shape characteristic parameters and velocity deviation to comprehensively evaluate proppant particles, and uses this as a basis to decide whether secondary treatment is needed to improve classification accuracy.

[0066] This application uses an intelligent correction model to correct particles that require secondary processing, and predicts the optimal correction parameters through machine learning methods to achieve targeted improvement of non-spherical particles.

[0067] The present application detects the classification results in real time and uses an optimization algorithm to search for the optimal process parameters, thereby achieving dynamic optimization and control of the classification process and ensuring that the classification results meet the target requirements.

[0068] The present application makes targeted improvements to the grading process and finished product quality of non-spherical proppant particles through shape characteristic parameter analysis based on sphericity, roundness and equivalent diameter, particle trajectory simulation during grading, intelligent correction and real-time optimization, thereby improving the accuracy and stability of the grading results and the quality of the finished petroleum proppant products. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A method flow chart of a petroleum proppant finished product sorting control method based on machine learning provided in this application;

[0070] Figure 2 A flow chart of the method for calculating trajectory deviation provided for this application;

[0071] Figure 3 A schematic diagram of the determination process of secondary processing of proppant particles provided in this application;

[0072] Figure 4 A functional module diagram of a petroleum proppant finished product sorting control system based on machine learning provided in this application;

[0073] Figure 5 A schematic diagram of the Krumbien / Sloss plate. DETAILED DESCRIPTION

[0074] In order to better understand the present application, various aspects of the present application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are only descriptions of exemplary embodiments of the present application and do not limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0075] In the accompanying drawings, the size, dimensions and shapes of the elements have been slightly adjusted for ease of illustration. The drawings are for illustration only and are not strictly drawn to scale. As used herein, the terms "substantially", "approximately" and similar terms are used as terms of approximation, not as terms of degree, and are intended to illustrate the inherent deviations in measurements or calculations that will be recognized by those of ordinary skill in the art. In addition, in this application, the order in which the steps are described does not necessarily represent the order in which these processes occur in actual operation, unless otherwise specified or can be derived from the context.

[0076] It should also be understood that expressions such as "including", "comprising", "having", "containing" and / or "comprising" are open rather than closed expressions in this specification, which indicate the presence of the stated features, elements and / or components, but do not exclude the presence of one or more other features, elements, components and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present application, "may" is used to mean "one or more embodiments of the present application". And, the term "exemplary" is intended to refer to an example or illustration.

[0077] Unless otherwise defined, all words (including engineering terms and scientific and technological terms) used in this article have the same meaning as those commonly understood by ordinary technicians in the field to which this application belongs. It should also be understood that unless there is a clear explanation in this application, words defined in commonly used dictionaries should be interpreted as having the same meaning as their meaning in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0078] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0079] Example 1

[0080] like Figure 1 As shown, a petroleum proppant finished product sorting control method based on machine learning provided in this application includes:

[0081] Step S100: obtaining a petroleum proppant particle sample, and extracting shape characteristic parameters of the petroleum proppant particle sample, wherein the petroleum proppant particle sample includes a spherical particle sample and a non-spherical particle sample, and the shape characteristic parameters include roundness, sphericity and equivalent diameter;

[0082] The specific method of obtaining the petroleum proppant particle sample and extracting the shape characteristic parameters of the petroleum proppant particle sample is:

[0083] Step S110: collecting spherical particle samples and non-spherical particle samples of petroleum proppant particles to obtain petroleum proppant particle samples, and scanning a two-dimensional contour image of the petroleum proppant particle samples;

[0084] The method for acquiring the two-dimensional contour image is to take a high-definition photo of the petroleum proppant particle sample by photographic technology, and use image analysis software to obtain the two-dimensional contour image of the petroleum proppant particle sample.

[0085] Step S120: Obtaining the maximum inscribed circle diameter of the petroleum proppant particle sample according to the two-dimensional contour image and the minimum circumscribed circle diameter , calculate the sphericity of petroleum proppant particle samples ;

[0086] The calculation formula of the sphericity is: ;

[0087] The sphericity ranges from 0 to 1, and the closer it is to 1, the closer the particle shape is to a sphere.

[0088] Step S130: Find the corresponding horizontal coordinate position on the Krumbien / Sloss plate according to the sphericity, and professionals compare the two-dimensional contour image with standard contour lines of different roundness levels, and select the most similar standard contour line to obtain the roundness C of the petroleum proppant particle sample;

[0089] The roundness ranges from 0.1 to 0.9, and the larger the value, the smoother the edges and corners of the particles;

[0090] like Figure 5 As shown, it is a schematic diagram of the Krumbien / Sloss plate, where X is the abscissa, the numbers on the abscissa represent the roundness, and Y is the ordinate, the numbers on the ordinate represent the sphericity;

[0091] The Krumbien / Sloss chart is a graphical tool used to illustrate geological sedimentary layers, rock types, and stratigraphic evolution.

[0092] Step S140: Determine the actual volume V of the petroleum proppant particle sample and calculate the equivalent diameter of the petroleum proppant particle sample ;

[0093] The calculation formula of the equivalent diameter is: ;

[0094] Step S150: The roundness, sphericity and equivalent diameter constitute the shape characteristic parameters of the petroleum proppant particle sample.

[0095] Step S100 obtains petroleum proppant particle samples, extracts their shape characteristic parameters, and comprehensively describes the shape characteristics of the proppant particles, laying a data foundation for subsequent classification.

[0096] Step S200: constructing a three-dimensional geometric model of the airflow classifier to generate three-dimensional trajectories of spherical particle samples and non-spherical particle samples during the classification process;

[0097] The specific method of constructing the three-dimensional geometric model of the airflow classifier and generating the three-dimensional trajectories of the spherical particle samples and the non-spherical particle samples during the classification process is:

[0098] Step S210: Obtaining the structural parameters and process parameters of the airflow classifier, the structural parameters include the feed port diameter , Exit height , the process parameters include feed rate, air flow velocity, and temperature T;

[0099] Step S220: determining the computational domain of the gas phase-particle phase numerical model, including the feed port, classification cavity and outlet of the gas flow classifier; designing a gas phase control equation to solve the gas phase velocity vector at any position and time in the computational domain; designing a particle phase control equation to solve the particle velocity of the petroleum proppant particle sample; the gas phase control equation and the particle phase control equation constitute the gas phase-particle phase numerical model;

[0100] The gas phase control equation is: , ,in, is the gas phase velocity vector, is the gas phase density, For pressure, is the kinematic viscosity of the gas phase, is the gravitational acceleration, t is the time variable, represents the gradient operation;

[0101] Among them, the gas phase control equation is used to solve the gas phase velocity vector at any position and time in the calculation domain;

[0102] The particle phase governing equation is: ,in, is the particle mass, is the particle velocity, is the aerodynamic drag, is gravity;

[0103] Among them, the particle phase control equation is used to solve the particle velocity of each petroleum proppant particle sample at any time during the classification process, and then obtain the motion trajectory of the petroleum proppant particle sample.

[0104] Step S230: Determine the gas phase density according to the ideal gas state equation, the actual temperature and pressure p , according to the dynamic viscosity of the air flow, determine the gas phase dynamic viscosity ; Based on the particle density of petroleum proppant particle samples and equivalent diameter Calculate particle mass , the aerodynamic drag is calculated based on the relative velocity between the petroleum proppant particle sample and the airflow , calculate the gravity based on the particle mass and gravitational acceleration ;

[0105] The calculation formula of the gas phase density is: ,in, is the gas constant;

[0106] The gas constant is related to the gas type and is determined by referring to relevant literature according to the specific gas type;

[0107] The calculation formula of the gas phase dynamic viscosity is: ,in, is the reference temperature The dynamic viscosity of the airflow under , S is the Sutherland constant;

[0108] Preferably, the reference temperature Take 273.15K or 298.15K;

[0109] The Sutherland constant is related to the type of gas and is determined by consulting relevant literature;

[0110] The particle mass is calculated as follows: ;

[0111] The calculation formula of the aerodynamic resistance is: ,in, is the drag coefficient;

[0112] The drag coefficient is set by those skilled in the art based on experience;

[0113] The calculation formula of gravity is: .

[0114] It should be noted that there are lift and virtual mass forces when the particles are classified. Compared with aerodynamic drag and gravity, lift and virtual mass forces are usually smaller and can be ignored in some cases. However, in the case of obvious particle acceleration or rotating flow fields, the influence of lift and virtual mass forces may become important, and it is necessary to choose whether to consider these two forces according to the actual situation.

[0115] According to the process parameters of the airflow classifier, the initial inlet velocity and initial outlet pressure of the calculation domain are set. For the wall boundary, the gas phase velocity is zero on the wall.

[0116] Step S240: constructing a three-dimensional geometric model of the airflow classifier, dividing the feed inlet, classification cavity and outlet of the airflow classifier into different areas and generating a grid;

[0117] The three-dimensional geometric model of the air flow classifier is constructed in CFD software;

[0118] The size of the grid is set by those skilled in the art based on experience. Need to meet: ,in, is the median diameter of the petroleum proppant particle sample; the median diameter of the petroleum proppant particle sample represents the particle diameter corresponding to the cumulative mass fraction of 50% on the particle size distribution curve of the petroleum proppant particle sample;

[0119] Step S250: Simulate the classification process of the petroleum proppant particle sample in the airflow classifier, solve the gas phase-particle phase numerical model, and obtain the gas phase velocity vector at any position and time in the calculation domain , obtain the particle velocity of the petroleum proppant particle sample at any position and time;

[0120] By simulating the classification process in the airflow classifier to obtain relevant experimental data, the gas phase-particle phase numerical model is numerically solved to obtain the gas phase velocity vector and particle velocity at different positions and times;

[0121] Step S260: obtaining the particle position of the petroleum proppant particle sample at each moment according to the particle velocity, and drawing a three-dimensional trajectory of the petroleum proppant particle sample during the classification process;

[0122] Use CFD software to process gas-solid two-phase flow and track the three-dimensional trajectory of petroleum proppant particle samples;

[0123] Step S270: Generate three-dimensional trajectories of the spherical particle samples and the non-spherical particle samples during the classification process and store them in a database.

[0124] Step S200 uses the gas phase-particle phase numerical model to simulate the three-dimensional trajectory of the proppant particles during the classification process, which can predict the movement behavior of particles with different shape characteristics in the air flow classifier and provide theoretical support for optimizing the classification.

[0125] Step S300: for the proppant particles to be classified, obtain their shape characteristic parameters, find the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified, and calculate the velocity deviation between the proppant particles to be classified and the spherical particle sample based on the three-dimensional trajectory;

[0126] The specific method of obtaining the shape characteristic parameters of the proppant particles to be classified, finding the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified, and calculating the velocity deviation between the proppant particles to be classified and the spherical particle sample based on the three-dimensional trajectory is:

[0127] Step S310: For the proppant particles to be classified, obtain their two-dimensional contour images and define their shape characteristic parameters as For the petroleum proppant particle samples in the database, the shape characteristic parameters are defined as , calculate the weighted Euclidean distance between the shape characteristic parameters of the proppant particles to be classified and the shape characteristic parameters of the petroleum proppant particle samples in the database , calculate the inverse of the weighted Euclidean distance to get the shape similarity ;in, are the sphericity, roundness and equivalent diameter of the proppant particles to be classified, are the sphericity, roundness and equivalent diameter of the jth petroleum proppant particle sample in the database, respectively;

[0128] The calculation formula of the weighted Euclidean distance is: ,in, , , are the weight coefficients of sphericity, roundness and equivalent diameter respectively, represents the weighted Euclidean distance between the shape characteristic parameters of the proppant particle to be classified and the j-th petroleum proppant particle sample in the database;

[0129] The weight coefficients of the sphericity, roundness and equivalent diameter items are set by those skilled in the art according to actual needs, and the sum of their weight coefficients is 1;

[0130] The calculation formula of the shape similarity is: ;

[0131] The greater the shape similarity, the more similar the shape of the proppant particle to be classified is to the jth petroleum proppant particle sample in the database.

[0132] Step S320: taking the three-dimensional trajectory corresponding to the petroleum proppant particle sample having the highest shape similarity to the proppant particle to be classified as the most similar trajectory of the proppant particle to be classified;

[0133] Step S330: Find a spherical particle sample in the database that is closest to the equivalent diameter of the proppant particle to be classified, and use the three-dimensional trajectory of the spherical particle sample as the reference three-dimensional trajectory;

[0134] Step S340: A three-dimensional coordinate system (x, y, z) is established based on the grid of the computational domain, wherein the z-axis direction is the axial direction of the airflow classifier, and the computational domain is equally divided into N grids in the z-axis direction, and the size of each grid is ;

[0135] The z-axis is the main direction of motion of the proppant particles;

[0136] Step S350: define the z-axis coordinate of the sampling point as the midpoint of each grid, that is, the expression of the z-axis coordinate of the ith sampling point is , where i=1,2,...,N;

[0137] Step S360: For the most similar trajectory and the reference three-dimensional trajectory, obtain the particle velocity at each sampling point respectively, and obtain the particle velocity at the i-th sampling point of the most similar trajectory , get the particle velocity of the benchmark three-dimensional trajectory at the i-th sampling point ;in, are the component velocities of the particle velocity in the x, y, and z directions of the most similar trajectory at the i-th sampling point, are the component velocities of the particle velocity in the x, y, and z axes of the reference three-dimensional trajectory at the i-th sampling point;

[0138] Step S370: Calculate the speed deviation between the most similar trajectory and the reference three-dimensional trajectory at each sampling point ;

[0139] The calculation formula of the speed deviation is: ;

[0140] Step S380: Calculate the average speed deviation between the most similar trajectory and the reference three-dimensional trajectory over the entire trajectory , the average velocity deviation is taken as the velocity deviation between the proppant particle and the spherical particle sample ;

[0141] The calculation formula of the average speed deviation is: .

[0142] Figure 2Flowchart of the method for calculating trajectory deviation provided in this application.

[0143] In step S300, for the proppant particles to be classified, by finding the most similar trajectory and calculating the velocity deviation between it and the reference three-dimensional trajectory, the difference in aerodynamic forces between the irregular particles and the spherical particles during the airflow classification process can be quantified, providing a basis for correction.

[0144] Step S400: Based on the shape characteristic parameters and velocity deviation of the proppant particles, the comprehensive deviation index is calculated, and a comprehensive deviation index threshold is preset. If the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification, otherwise, they are subjected to secondary processing;

[0145] The method of calculating the comprehensive deviation index based on the shape characteristic parameters and velocity deviation of the proppant particles and presetting the comprehensive deviation index threshold is as follows: if the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification; otherwise, the proppant particles are subjected to secondary treatment as follows:

[0146] Step S410: Calculate the comprehensive deviation index by combining the shape characteristic parameters of the proppant particles and the velocity deviation between the proppant particles and the spherical particle samples. ;

[0147] The calculation formula of the comprehensive deviation index is: ,in, is the reference diameter, is the air flow velocity, , , , are the weight coefficients of velocity deviation, sphericity, roundness, and equivalent diameter, respectively;

[0148] The weight coefficients of speed deviation, sphericity, roundness and equivalent diameter are set by those skilled in the art according to actual needs. , , , The sum of the weight coefficients is 1;

[0149] The reference diameter is the median of the range of equivalent diameters of the proppant particles to be classified, i.e. , , Respectively represent the maximum and minimum values ​​of the equivalent diameter;

[0150] The comprehensive deviation index refers to the comprehensive deviation degree of the proppant particles to be classified relative to the standard spherical particles in terms of shape and movement trajectory.

[0151] Step S420: Preset comprehensive deviation index threshold , comparing the size of the comprehensive deviation index with the comprehensive deviation index threshold, if the comprehensive deviation index of the proppant particle is less than or equal to the comprehensive deviation index threshold, the proppant particle directly enters the airflow classifier for classification; if the comprehensive deviation index of the proppant particle is greater than the comprehensive deviation index threshold, the proppant particle is subjected to secondary processing;

[0152] The comprehensive deviation index threshold is set by those skilled in the art based on experience.

[0153] Figure 3 A schematic diagram of the determination process for secondary processing of proppant particles provided in this application.

[0154] Step S400 introduces a comprehensive deviation index, which combines the shape characteristic parameters and the velocity deviation to comprehensively evaluate the deviation of the proppant particles from the ideal spherical shape, and uses this as a basis to determine whether secondary processing is required to improve classification accuracy.

[0155] Step S500: using the trained intelligent correction model to predict the best correction parameters, correcting the proppant particles undergoing secondary processing until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and sending the corrected proppant particles to an airflow classifier for classification;

[0156] The specific method of using the trained intelligent correction model to predict the optimal correction parameters, correcting the proppant particles undergoing secondary processing until the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and sending the corrected proppant particles to the airflow classifier for classification is:

[0157] Step S510: obtaining shape characteristic parameters and corresponding optimal correction parameters of non-spherical particles in historical data, taking the shape characteristic parameters of non-spherical particles as input data and the corresponding optimal correction parameters as output data to form training samples, and constructing a training set;

[0158] Step S520: using a deep neural network as the basic framework for training an intelligent correction model, the intelligent correction model includes an input layer, a hidden layer, and an output layer, using mean square error as a loss function, and using the Adam optimization algorithm to optimize model parameters;

[0159] The input layer is used to receive shape characteristic parameters of non-spherical particles, the hidden layer is composed of multiple fully connected layers, each layer uses a ReLU activation function, the number of hidden layers and the number of neurons in each layer are optimized through hyperparameter search, and the output layer outputs the predicted optimal correction parameters. The number of neurons in the output layer is the same as the dimension of the optimal correction parameters, and a linear activation function is used.

[0160] Step S530: Input the training set into the deep neural network, perform forward propagation, calculate the output data and loss function of each group of training samples, calculate the gradient of the loss function to the model parameters through the back propagation algorithm, and use the Adam optimization algorithm to update the model parameters to minimize the loss function;

[0161] Step S540: When the loss function between the predicted optimal correction parameter and the actual optimal correction parameter reaches convergence, the training is completed and a trained intelligent correction model is obtained;

[0162] Step S550: For proppant particles that need to be processed twice, the trained intelligent correction model is used to predict the optimal correction parameters of the proppant particles, the proppant particles are corrected based on the optimal correction parameters, and the comprehensive deviation index of the corrected proppant particles is calculated. If the comprehensive deviation index is still greater than the comprehensive deviation index threshold, the proppant particles are repeatedly corrected until the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, thereby meeting the classification requirements of the airflow classifier;

[0163] Step S560: Send the corrected proppant particles to an air flow classifier for classification.

[0164] Step S500 uses the intelligent correction model to correct the particles that need secondary treatment. The corrected proppant particles have their shape characteristics improved in a targeted manner to meet the classification requirements and improve the quality of the finished proppant particles.

[0165] Step S600: Real-time detection of the actual particle size distribution of the classification result of the airflow classifier, calculation of the deviation value between the actual particle size distribution and the target particle size distribution, calculation of the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, construction of the optimization objective function, and search for the optimal process parameters;

[0166] The specific method of detecting the actual particle size distribution of the classification result of the airflow classifier in real time, calculating the deviation value between the actual particle size distribution and the target particle size distribution, calculating the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, constructing the optimization objective function, and searching for the optimal process parameters is as follows:

[0167] Step S610: Real-time monitoring of the actual particle size distribution of the classification results of the airflow classifier , calculate the actual particle size distribution and the target particle size distribution The deviation between ;

[0168] The calculation formula of the deviation value is: ,in, is the actual particle size distribution, is the target particle size distribution, , Respectively represent the maximum and minimum values ​​of the equivalent diameter;

[0169] The actual particle size distribution detection method is: scanning the three-dimensional geometric model of the proppant particles, calculating the equivalent diameter thereof, obtaining a frequency distribution histogram of the equivalent diameter of the proppant particles, and obtaining the actual particle size distribution by statistics;

[0170] The target particle size distribution is set according to product quality standards, customer requirements or production experience.

[0171] Step S620: Real-time statistics of shape characteristic parameters and comprehensive deviation indexes of proppant particles entering the airflow classifier, and calculation of the average value of the shape characteristic parameters And the average value of comprehensive deviation index ;

[0172] Step S630: define the deviation value, shape feature parameter average value and comprehensive deviation index average value The optimization objective function , process parameter k is the optimization variable, defining the process parameter range constraint;

[0173] The optimization objective function The expression of (k) is: ,in, , , is the weight coefficient of the deviation value, the average value of the shape characteristic parameter and the average value of the comprehensive deviation index, which is set by those skilled in the art based on experience;

[0174] The expression of the process parameter range constraint is: ,in, , are the nth process parameters The minimum and maximum value range of n is 1, 2, 3;

[0175] Step S640: using a genetic algorithm to solve the optimization objective function and obtain the optimal process parameters.

[0176] The optimal process parameters are used to adjust and control the airflow classifier in real time.

[0177] Step S600 can dynamically adjust the classification process by real-time detection of classification results and optimization of process parameters to ensure that the classification results meet the target particle size distribution and improve classification stability and accuracy.

[0178] Example 2

[0179] like Figure 4 As shown, a petroleum proppant finished product sorting control system based on machine learning provided by this application includes:

[0180] A feature data acquisition module is used to acquire a petroleum proppant particle sample and extract shape feature parameters of the petroleum proppant particle sample, wherein the petroleum proppant particle sample includes a spherical particle sample and a non-spherical particle sample, and the shape feature parameters include roundness, sphericity and equivalent diameter;

[0181] A three-dimensional trajectory generation module is used to construct a three-dimensional geometric model of the airflow classifier and generate three-dimensional trajectories of spherical particle samples and non-spherical particle samples during the classification process;

[0182] The velocity deviation calculation module is used to obtain the shape characteristic parameters of the proppant particles to be classified, find the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified, and calculate the velocity deviation between the proppant particles to be classified and the spherical particle sample based on the three-dimensional trajectory;

[0183] A particle classification processing module is used to calculate the comprehensive deviation index of the proppant particles based on their shape characteristic parameters and velocity deviation, and preset a comprehensive deviation index threshold. If the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification, otherwise they are subjected to secondary processing;

[0184] The secondary correction classification module is used to use the trained intelligent correction model to predict the best correction parameters, correct the proppant particles undergoing secondary processing until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and send the corrected proppant particles to the airflow classifier for classification;

[0185] The process parameter optimization module is used to detect the actual particle size distribution of the classification results of the airflow classifier in real time, calculate its deviation value from the target particle size distribution, calculate the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, construct the optimization objective function, and search for the optimal process parameters.

[0186] Example 3

[0187] According to another aspect of the present application, an electronic device is provided. The electronic device may include one or more processors and one or more memories. The memories store computer-readable codes, and when the computer-readable codes are run by one or more processors, the method for controlling the sorting of petroleum proppant products based on machine learning as described above may be executed.

[0188] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. The storage device in the electronic device, such as a ROM or a hard disk, can store a petroleum proppant finished product sorting control method based on machine learning provided in the present application. A petroleum proppant finished product sorting control method based on machine learning may, for example, include: obtaining a petroleum proppant particle sample, extracting shape characteristic parameters of the petroleum proppant particle sample, the petroleum proppant particle sample includes a spherical particle sample and a non-spherical particle sample, and the shape characteristic parameters include roundness, sphericity and equivalent diameter; constructing a three-dimensional geometric model of the airflow classifier, generating a three-dimensional trajectory of the spherical particle sample and the non-spherical particle sample during the classification process; for the proppant particles to be classified, obtaining their shape characteristic parameters, finding the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified, and calculating the velocity deviation between the proppant particles to be classified and the spherical particle sample based on the three-dimensional trajectory; calculating the velocity deviation between the proppant particles to be classified and the spherical particle sample based on the shape characteristic parameters and the velocity of the proppant particles ... Degree deviation, calculate its comprehensive deviation index, preset the comprehensive deviation index threshold, if the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification, otherwise they are subjected to secondary processing; use the trained intelligent correction model to predict the best correction parameters, correct the proppant particles for secondary processing until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and send the corrected proppant particles to the airflow classifier for classification; real-time detection of the actual particle size distribution of the classification result of the airflow classifier, calculate its deviation value from the target particle size distribution, calculate the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, construct an optimization objective function, and search for the optimal process parameters. Furthermore, the electronic device may also include a user interface. Of course, when implementing different devices, one or more components in the electronic device may be omitted according to actual needs.

[0189] Example 4

[0190] According to one embodiment of the present application, a readable storage medium is also provided. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are executed by the processor, a machine-learning-based petroleum proppant finished product sorting control method according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0191] In addition, according to the implementation mode of the present application, the process described in the above reference flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: obtaining a petroleum proppant particle sample, extracting shape characteristic parameters of the petroleum proppant particle sample, the petroleum proppant particle sample includes a spherical particle sample and a non-spherical particle sample, the shape characteristic parameters include roundness, sphericity and equivalent diameter; constructing a three-dimensional geometric model of the airflow classifier, generating a three-dimensional trajectory of the spherical particle sample and the non-spherical particle sample during the classification process; for the proppant particles to be classified, obtaining their shape characteristic parameters, finding the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified, and calculating the shape similarity between the proppant particles to be classified and the proppant particles to be classified based on the three-dimensional trajectory. The speed deviation between spherical particle samples; based on the shape characteristic parameters and speed deviation of the proppant particles, the comprehensive deviation index is calculated, and the comprehensive deviation index threshold is preset. If the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification, otherwise they are subjected to secondary processing; the trained intelligent correction model is used to predict the best correction parameters, and the proppant particles subjected to secondary processing are corrected until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and the corrected proppant particles are sent to the airflow classifier for classification; the actual particle size distribution of the classification result of the airflow classifier is detected in real time, and its deviation value from the target particle size distribution is calculated, the shape characteristic parameter average value and the comprehensive deviation index average value of the proppant particles entering the airflow classifier are calculated, and the optimization objective function is constructed to search for the optimal process parameters. When the computer program is executed by the central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0192] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0193] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0194] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for sorting and controlling finished petroleum proppant products based on machine learning, characterized in that: include: Acquire a petroleum proppant particle sample, and extract shape characteristic parameters of the petroleum proppant particle sample, wherein the petroleum proppant particle sample includes a spherical particle sample and a non-spherical particle sample, and the shape characteristic parameters include roundness, sphericity, and equivalent diameter; Construct a three-dimensional geometric model of the airflow classifier and generate three-dimensional trajectories of spherical and non-spherical particle samples during the classification process; For the proppant particles to be classified, the shape characteristic parameters are obtained, and the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified is found, and the velocity deviation between the proppant particles to be classified and the spherical particle sample is calculated based on the three-dimensional trajectory; Based on the shape characteristic parameters and velocity deviation of the proppant particles, the comprehensive deviation index is calculated, and a comprehensive deviation index threshold is preset. If the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification, otherwise they are subjected to secondary processing; The trained intelligent correction model is used to predict the optimal correction parameters, and the proppant particles undergoing secondary processing are corrected until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and the corrected proppant particles are sent to the airflow classifier for classification; The actual particle size distribution of the classification results of the airflow classifier is detected in real time, and its deviation from the target particle size distribution is calculated. The average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier are calculated, and the optimization objective function is constructed to search for the optimal process parameters.

2. A method for controlling the sorting of petroleum proppant products based on machine learning as claimed in claim 1, characterized in that: The specific method of obtaining the petroleum proppant particle sample and extracting the shape characteristic parameters of the petroleum proppant particle sample is: collecting spherical particle samples and non-spherical particle samples of petroleum proppant particles to obtain petroleum proppant particle samples, and scanning two-dimensional contour images of the petroleum proppant particle samples; Obtaining the maximum inscribed circle diameter of a petroleum proppant particle sample based on a two-dimensional contour image and the minimum circumscribed circle diameter , calculate the sphericity of petroleum proppant particle samples ; According to the sphericity, the corresponding horizontal coordinate position is found on the Krumbien / Sloss plate, the two-dimensional contour image is compared with the standard contour lines of different roundness levels, and the most similar standard contour line is selected to obtain the roundness C of the petroleum proppant particle sample; Determine the actual volume V of the petroleum proppant particle sample and calculate the equivalent diameter of the petroleum proppant particle sample ; The shape characteristic parameters of petroleum proppant particle samples are composed of roundness, sphericity and equivalent diameter.

3. A method for controlling the sorting of petroleum proppant products based on machine learning as claimed in claim 2, characterized in that: The specific method for constructing the three-dimensional geometric model of the airflow classifier is: Obtain the structural parameters and process parameters of the airflow classifier, including the feed port diameter , Exit height , the process parameters include feed rate, air flow velocity, and temperature T; Determine the computational domain of the gas phase-particle phase numerical model, including the feed port, classification cavity and outlet of the gas flow classifier; design the gas phase control equation to solve the gas phase velocity vector at any position and time in the computational domain; design the particle phase control equation to solve the particle velocity of the petroleum proppant particle sample; the gas phase control equation and the particle phase control equation constitute the gas phase-particle phase numerical model; Determine the gas phase density based on the ideal gas state equation, actual temperature and pressure p , according to the dynamic viscosity of the air flow, determine the gas phase dynamic viscosity ; Based on the particle density of petroleum proppant particle samples and equivalent diameter Calculate particle mass , the aerodynamic drag is calculated based on the relative velocity between the petroleum proppant particle sample and the airflow , calculate the gravity based on the particle mass and gravitational acceleration ; A three-dimensional geometric model of the airflow classifier is constructed, and the feed inlet, classification cavity and outlet of the airflow classifier are divided into different areas and meshes are generated.

4. A method for controlling the sorting of petroleum proppant products based on machine learning as claimed in claim 3, characterized in that: The specific method for generating the three-dimensional trajectories of the spherical particle samples and the non-spherical particle samples during the classification process is: Simulate the classification process of petroleum proppant particle samples in the airflow classifier, solve the gas phase-particle phase numerical model, and obtain the gas phase velocity vector at any position and time in the calculation domain , obtain the particle velocity of the petroleum proppant particle sample at any position and time; The particle position of the petroleum proppant particle sample at each moment is obtained according to the particle velocity, and the three-dimensional trajectory of the petroleum proppant particle sample in the classification process is drawn; The three-dimensional trajectories of spherical particle samples and non-spherical particle samples during the classification process are generated and stored in the database.

5. The method for controlling the sorting of petroleum proppant products based on machine learning as claimed in claim 4, characterized in that: The specific method for obtaining the shape characteristic parameters of the proppant particles to be classified and finding the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified is: For the proppant particles to be classified, obtain their two-dimensional contour images and define their shape characteristic parameters as For the petroleum proppant particle samples in the database, the shape characteristic parameters are defined as , calculate the weighted Euclidean distance between the shape characteristic parameters of the proppant particles to be classified and the shape characteristic parameters of the petroleum proppant particle samples in the database , calculate the inverse of the weighted Euclidean distance to get the shape similarity ;in, are the sphericity, roundness and equivalent diameter of the proppant particles to be classified, are the sphericity, roundness and equivalent diameter of the jth petroleum proppant particle sample in the database, respectively; The three-dimensional trajectory corresponding to the petroleum proppant particle sample having the highest shape similarity to the proppant particle to be classified is taken as the most similar trajectory of the proppant particle to be classified.

6. A method for controlling the sorting of petroleum proppant products based on machine learning as claimed in claim 5, characterized in that: The specific method for calculating the velocity deviation between the proppant particles to be classified and the spherical particle samples based on the three-dimensional trajectory is: Find a spherical particle sample in the database that is closest to the equivalent diameter of the proppant particle to be classified, and use the three-dimensional trajectory of the spherical particle sample as the reference three-dimensional trajectory; A three-dimensional coordinate system (x, y, z) is established based on the grid of the computational domain, where the z-axis direction is the axial direction of the airflow classifier. The computational domain is divided into N grids at equal intervals in the z-axis direction, and the size of each grid is ; The z-axis coordinate of the sampling point is defined as the midpoint of each grid, that is, the expression of the z-axis coordinate of the i-th sampling point is , where i=1,2,...,N; For the most similar trajectory and the reference three-dimensional trajectory, the particle velocity at each sampling point is obtained respectively, and the particle velocity at the i-th sampling point of the most similar trajectory is obtained. , get the particle velocity of the benchmark three-dimensional trajectory at the i-th sampling point ;in, are the component velocities of the particle velocity in the x, y, and z directions of the most similar trajectory at the i-th sampling point, are the component velocities of the particle velocity in the x, y, and z axes of the reference three-dimensional trajectory at the i-th sampling point; Calculate the velocity deviation between the most similar trajectory and the reference 3D trajectory at each sampling point ; Calculate the average speed deviation between the most similar trajectory and the reference 3D trajectory over the entire trajectory , the average velocity deviation is taken as the velocity deviation between the proppant particle and the spherical particle sample .

7. A method for controlling the sorting of petroleum proppant products based on machine learning as claimed in claim 6, characterized in that: The method of calculating the comprehensive deviation index based on the shape characteristic parameters and velocity deviation of the proppant particles and presetting the comprehensive deviation index threshold is as follows: if the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification; otherwise, the proppant particles are subjected to secondary treatment as follows: Combine the shape characteristic parameters of the proppant particles and the velocity deviation between the proppant particles and the spherical particle samples to calculate the comprehensive deviation index ; Preset comprehensive deviation indicator threshold , compare the size of the comprehensive deviation index with the comprehensive deviation index threshold. If the comprehensive deviation index of the proppant particle is less than or equal to the comprehensive deviation index threshold, the proppant particle directly enters the air flow classifier for classification; if the comprehensive deviation index of the proppant particle is greater than the comprehensive deviation index threshold, the proppant particle undergoes secondary processing.

8. A method for controlling the sorting of petroleum proppant products based on machine learning as claimed in claim 7, characterized in that: The specific method of using the trained intelligent correction model to predict the optimal correction parameters, correcting the proppant particles undergoing secondary processing until the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and sending the corrected proppant particles to the airflow classifier for classification is: Obtaining shape characteristic parameters and corresponding optimal correction parameters of non-spherical particles in historical data, using the shape characteristic parameters of non-spherical particles as input data and the corresponding optimal correction parameters as output data to form training samples, and constructing a training set; A deep neural network is used as the basic framework for training an intelligent correction model, which includes an input layer, a hidden layer, and an output layer. The mean square error is used as a loss function, and the Adam optimization algorithm is used to optimize the model parameters. The training set is input into the deep neural network for forward propagation, the output data and loss function of each group of training samples are calculated, the gradient of the loss function to the model parameters is calculated through the back-propagation algorithm, and the Adam optimization algorithm is used to update the model parameters and minimize the loss function; When the loss function between the predicted optimal correction parameters and the actual optimal correction parameters reaches convergence, the training is completed and a trained intelligent correction model is obtained; For the proppant particles undergoing secondary processing, the trained intelligent correction model is used to predict the optimal correction parameters of the proppant particles, the proppant particles are corrected based on the optimal correction parameters, and the comprehensive deviation index of the corrected proppant particles is calculated. If the comprehensive deviation index is still greater than the comprehensive deviation index threshold, the proppant particles are repeatedly corrected until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, thereby meeting the classification requirements of the airflow classifier; The rectified proppant particles are sent to the air flow classifier for classification.

9. A method for controlling the sorting of petroleum proppant products based on machine learning as claimed in claim 8, characterized in that: The specific method of detecting the actual particle size distribution of the classification result of the airflow classifier in real time, calculating the deviation value between the actual particle size distribution and the target particle size distribution, calculating the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, constructing the optimization objective function, and searching for the optimal process parameters is as follows: Real-time monitoring of the actual particle size distribution of the classification results of the airflow classifier , calculate the actual particle size distribution and the target particle size distribution The deviation between ; Real-time statistics of the shape characteristic parameters and comprehensive deviation index of the proppant particles entering the airflow classifier, and calculation of the average value of the shape characteristic parameters And the average value of comprehensive deviation index ; The definition includes deviation value, shape characteristic parameter average value and comprehensive deviation index average value The optimization objective function , process parameter k is the optimization variable, defining the process parameter range constraint; Genetic algorithm is used to solve the optimization objective function and obtain the optimal process parameters.

10. A petroleum proppant finished product sorting control system based on machine learning, which is used to implement a petroleum proppant finished product sorting control method based on machine learning as claimed in any one of claims 1 to 9, characterized in that: include: A feature data acquisition module is used to acquire a petroleum proppant particle sample and extract shape feature parameters of the petroleum proppant particle sample, wherein the petroleum proppant particle sample includes a spherical particle sample and a non-spherical particle sample, and the shape feature parameters include roundness, sphericity and equivalent diameter; A three-dimensional trajectory generation module is used to construct a three-dimensional geometric model of the airflow classifier and generate three-dimensional trajectories of spherical particle samples and non-spherical particle samples during the classification process; The velocity deviation calculation module is used to obtain the shape characteristic parameters of the proppant particles to be classified, find the petroleum proppant particle sample with the highest shape similarity to the proppant particles to be classified, and calculate the velocity deviation between the proppant particles to be classified and the spherical particle sample based on the three-dimensional trajectory; A particle classification processing module is used to calculate the comprehensive deviation index of the proppant particles based on their shape characteristic parameters and velocity deviation, and preset a comprehensive deviation index threshold. If the comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, the proppant particles directly enter the airflow classifier for classification, otherwise they are subjected to secondary processing; The secondary correction classification module is used to use the trained intelligent correction model to predict the best correction parameters, correct the proppant particles undergoing secondary processing until their comprehensive deviation index is less than or equal to the comprehensive deviation index threshold, and send the corrected proppant particles to the airflow classifier for classification; The process parameter optimization module is used to detect the actual particle size distribution of the classification results of the airflow classifier in real time, calculate its deviation value from the target particle size distribution, calculate the average value of the shape characteristic parameters and the average value of the comprehensive deviation index of the proppant particles entering the airflow classifier, construct the optimization objective function, and search for the optimal process parameters.

Images