A method for optimizing the throughput of a trommel screen

Abstract

The present application relates to a kind of drum screen screen rate optimization method, belong to screen rate optimization technical field, solve the low accuracy of screen rate prediction method in prior art, high computing cost and difficult to obtain optimal screen rate by adjusting operating parameter problem.Based on the discrete element model of drum screen and particle plant space model, a screen rate sample dataset is constructed; a target machine learning model is trained based on the screen rate sample dataset to obtain a screen rate prediction model; obtain a sample of parameters to be tested, input the sample of parameters to be tested into the screen rate prediction model to obtain the screen rate of different particle sizes corresponding to the sample of parameters to be tested; if the screen rate meets the expected target, the sample of parameters to be tested is used as the control parameter of the drum screen; if it does not meet the expected target, the particle swarm optimization algorithm is used to optimize the sample of parameters to be tested, and the optimized parameters are used as the control parameter of the drum screen, thereby improving the screen rate of the test vibration screen. An accurate and efficient screen rate prediction and optimization method is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of screen passing rate optimization, and in particular to a drum screen screen passing rate optimization method. BACKGROUND

[0002] The screen passing rate (i.e. the rate of screening out particles smaller than the screen hole) is an important indicator for evaluating the effect of drum screening on granular materials, and the screen passing rate is affected by many factors such as drum speed, drum installation angle, residence time of the material in the drum screen, length-to-diameter ratio of the drum, and material properties, and there is a non-linear relationship between each factor.

[0003] In the prior art, the prediction of the screening behavior of the drum screen mainly includes the traditional probability screen passing model, the dynamic screen passing model, and the discrete element numerical simulation, but the traditional probability screen passing model is based on idealized assumptions, ignores the complex behaviors such as collision, accumulation and rolling between particles, cannot capture the motion trajectory of the particles, and does not consider the influence of particle shape and particle size distribution, making it difficult to describe the dynamic evolution of the screening process; the dynamic screen passing model analyzes the force and motion of the particles based on the principles of physical mechanics, but the description of the interaction between particles is simplified, and the calculation complexity is high, making it difficult to simulate large-scale particle systems, resulting in insufficient description of the non-linear dynamic characteristics of the particles; the discrete element numerical simulation (DEM) as a fine modeling method can better describe the motion behavior of particles in the drum screen, but the calculation cost is huge, and the universality and accuracy of the simulation results still need to be further verified and improved.

[0004] The neural network model has strong non-linear mapping ability and self-learning ability, can extract useful features from a large amount of data, and can establish a complex relationship between input and output. Therefore, it is necessary to combine the advantages of the neural network model and the DEM simulation to provide an accurate and efficient screen passing rate prediction and optimization method, so as to more comprehensively reveal the screening mechanism of the drum screen. SUMMARY

[0005] In view of the above analysis, the embodiments of the present application aim to provide a drum screen screen passing rate optimization method to solve the problems of low accuracy, high calculation cost and difficulty in adjusting the working condition parameters to obtain the optimal screen passing rate of the existing screen passing rate prediction method.

[0006] The embodiments of the present application provide a drum screen screen passing rate optimization method, which comprises:

[0007] constructing a discrete element model of the drum screen and a particle plant space model;

[0008] constructing a screen passing rate sample dataset based on the drum screen discrete element model and the particle plant space model, the screen passing rate sample dataset comprising: particle size, feeding rate of each particle, drum screen inclination angle, drum screen screen mesh rotating speed, drum screen built-in component rotating speed, screen passing rate;

[0009] training a target machine learning model based on the screen passing rate sample dataset to obtain a screen passing rate prediction model;

[0010] obtaining a to-be-tested parameter sample, and inputting the to-be-tested parameter sample into the screen passing rate prediction model to obtain screen passing rates of different particle sizes corresponding to the to-be-tested parameter sample;

[0011] if the screen passing rate meets an expected target, taking the to-be-tested parameter sample as a control parameter of the drum screen, if the screen passing rate does not meet the expected target, optimizing the to-be-tested parameter sample by using a particle swarm algorithm, and taking the optimized parameter as the control parameter of the drum screen, so as to improve the screen passing rate of the to-be-tested vibrating screen;

[0012] In the optimization process, the input parameter represented by each particle is input into the screen passing rate prediction model to obtain a screen passing rate corresponding to the particle, and a fitness function is calculated based on the screen passing rate of the particle and the drum screen inclination angle, the drum screen screen mesh rotating speed and the drum screen built-in component rotating speed corresponding to the particle.

[0013] Further improvement based on the above method, the construction of the drum screen discrete element model and the particle plant space model comprises: constructing a drum screen discrete element model, the drum screen discrete element model comprising: a cylindrical screen mesh and a built-in component, and dividing the cylindrical screen mesh discrete element model to obtain a plurality of sub-cylindrical screen mesh discrete element models with the same screen length; constructing a particle plant space model, the particle plant space model being placed at the feeding end of the drum screen discrete element model and being located inside the cylindrical screen mesh discrete element model, and setting the particle generation mode and particle parameters of the particle plant space model based on business requirements.

[0014] Further improvement based on the above method, the division of the cylindrical screen mesh discrete element model to obtain a plurality of sub-cylindrical screen mesh discrete element models with the same screen length comprises: obtaining the drum screen length of the cylindrical screen mesh discrete element model and the particle plant space model length of the particle plant space model along the drum screen length direction, obtaining the available drum screen length based on the drum screen length and the particle plant space model length along the drum screen length direction; dividing the cylindrical screen mesh discrete element model based on the available drum screen length and a preset length, so as to obtain a plurality of sub-cylindrical screen mesh discrete element models with the same screen length.

[0015] Based on the further improvement of the above method, the cylinder screen discrete element model is divided based on the available drum screen length and the preset length, thereby obtaining a plurality of sub-cylinder screen discrete element models with the same screen length, comprising:

[0016] A1: set a division starting point, the division starting point is located on the axis of the cylinder screen discrete element model, and the shortest distance from the division starting point to the discharge end of the cylinder screen discrete element model is equal to the available drum screen length;

[0017] A2: taking the division starting point as the division starting point, the preset length is the step length, a plurality of division points are divided on the axis, and the cylinder screen discrete element model is divided by a plane passing through the division point and perpendicular to the axis to obtain a plurality of sub-cylinder screen discrete element models with the same screen length;

[0018] A3: offset the division starting point to the discharge end by a preset distance, take the offset point as a new division starting point, judge whether the current offset number exceeds the preset moving number, if not, return to A2; if yes, take the result of each division in step A2 as a sub-cylinder screen discrete element model.

[0019] Based on the further improvement of the above method, the drum screen discrete element model and the particle factory space model are used to construct a screen permeation rate sample data set, comprising:

[0020] B1: based on the drum screen discrete element model and the particle factory space model, realize the simulation of the drum screen screening process, and obtain the simulation results, the simulation results include the feeding rate of various particles, the spatial position of each particle, the inclination angle of the drum screen, the screen mesh rotating speed of the drum screen, the built-in component rotating speed of the drum screen corresponding to each time step in the simulation process;

[0021] B2: based on the spatial position of each particle, calculate the screen permeation rate of different particle sizes corresponding to each sub-cylinder screen discrete element model;

[0022] B3: taking the feeding rate of various particles, the inclination angle of the drum screen, the screen mesh rotating speed of the drum screen, the built-in component rotating speed of the drum screen and the screen permeation rate of various particles corresponding to each time step of each sub-cylinder screen discrete element model as a sample in the screen permeation rate sample data set corresponding to the screen length;

[0023] B4: judge whether the amount of sample data generated at present meets the preset quantity requirement, if yes, stop generating; if not, adjust the current particle generation mode and particle parameters based on the business demand, and return to B1.

[0024] Further improvement based on the above method, the screen rate prediction model is a neural network model, and the number of neurons, Dropout coefficient and batch size of the neural network model are optimized by Harris Hawk optimization algorithm, and the optimization results are used as the initial parameters of the neural network model.

[0025] Further improvement based on the above method, the particle swarm optimization algorithm is used to optimize the parameter sample to be tested, and the optimized parameters are used as the control parameters of the drum screen.

[0026] C1: initialize the particle swarm, select the sample data meeting the expected target from the screen rate sample data set as the first sample data set, use the feed rate, drum screen inclination angle, drum screen screen speed and drum screen built-in component speed of each data in the first sample data set as a particle, use the feed rate, drum screen inclination angle, drum screen screen speed and drum screen built-in component speed of each particle in the parameter sample to be tested as a particle, and judge whether the total number of current particles meets the preset number, if yes, execute C3, if not, execute C2;

[0027] C2: statistics of the range of each parameter in the first sample data set, random value generation algorithm is used to generate multiple particles in the range of each parameter, so that the total number of particles meets the preset number requirement;

[0028] C3: calculate the fitness function of each particle to update the individual optimal and global optimal value, judge whether the end condition is met, if yes, execute C5, if not, execute C4;

[0029] C4: update the speed and position of each particle, and return to C3;

[0030] C5: output the optimal position, and the particle group parameters corresponding to the optimal position are the control parameters of the drum screen.

[0031] Further improvement based on the above method, the fitness function is calculated based on the screen rate of the particle and the drum screen inclination angle, drum screen screen speed and drum screen built-in component speed corresponding to the particle.

[0032]

[0033] Wherein, y' is the predicted screen rate corresponding to the current particle, is the expected target, α', ω1', ω2' is the drum screen inclination angle, drum screen screen speed and drum screen built-in component speed corresponding to the current particle, α, ω1, ω2 is the drum screen inclination angle, drum screen screen speed and drum screen built-in component speed of the parameter sample to be tested, a, b, c, d is a constant.

[0034] Further improvement based on the above method, the screen passing rate of different particle sizes corresponding to each sub-cylindrical screen discrete element model is calculated based on the spatial position of each particle, including: for the first sub-cylindrical screen discrete element model close to the feed end, the screen passing rate of different particle sizes is calculated based on the feed rate of various particles generated in the particle factory spatial model, the inclination angle of the trommel screen, the screen speed of the trommel screen, the built-in component speed of the trommel screen, and the mass of various sizes of particles under the first sub-cylindrical screen discrete element model; for other sub-cylindrical screen discrete element models, the particle size at the discharge end of the previous sub-cylindrical screen discrete element model, the feed rate of various particles, the inclination angle of the trommel screen, the screen speed of the trommel screen, the built-in component speed of the trommel screen are taken as the feed parameters of the current sub-cylindrical screen discrete element model, and the screen passing rate of different particle sizes is calculated based on the feed parameters of the sub-cylindrical screen discrete element model and the mass of various sizes of particles under the sub-cylindrical screen discrete element model.

[0035] Further improvement based on the above method, the simulation result refers to the particle size of each sub-cylindrical screen discrete element model, the feed rate of various particles, the spatial position of each particle, the inclination angle of the trommel screen, the screen speed of the trommel screen, and the built-in component speed of the trommel screen in each time step after the screening process reaches a steady state; the steady state refers to the sum of the mass of particles passing through the sub-cylindrical screen discrete element model and the discharge rate of particles at the discharge end per unit time being equal to the particle feed rate.

[0036] Compared with the prior art, the present application can at least achieve one of the following beneficial effects:

[0037] 1. The present application provides a trommel screen screen passing rate optimization method, which can guide users to design working condition parameters that meet the expected screen passing rate, save process cost and improve design efficiency after obtaining the screen passing rate of different particle sizes according to the current working condition parameters and judging based on the expected target.

[0038] 2. The present application provides a trommel screen screen passing rate optimization method, which can more flexibly meet business needs, and compared with the traditional probability screen passing model, dynamic screen passing model and discrete element numerical simulation in the prior art, the method proposed in the present application has higher prediction accuracy and calculation efficiency, strong real-time performance and can adapt to various complex working conditions.

[0039] 3. The application provides a trommel screen screening rate optimization method, based on a preset length, a plurality of length-identical sub-screen discrete element models are obtained by dividing the trommel screen discrete element model, for each working condition simulation of the particle plant space model, after steady state, each time step can generate a plurality of groups of simulation data corresponding to the working condition and the preset length screen, compared with the prior art, each time step can only generate a group of simulation data corresponding to the working condition and the preset length, the efficiency of the screening efficiency sample data set constructed by the application is higher; in addition, the various sub-screen discrete element models are continuously arranged, the discharge of the previous sub-screen discrete element model is used as the feed of the next sub-screen discrete element model, compared with the prior art in which sample data is only obtained for a fixed length screen model, the screening sample data obtained by the simulation of the various sub-screen discrete element models in the application is closer to the actual vibrating screen screening process.

[0040] In the application, the above technical solutions can be combined with each other to realize more preferred combination solutions. Other features and advantages of the application will be described in the subsequent description, and some advantages will become apparent from the description, or will be understood by implementing the application. The purposes and other advantages of the application can be realized and obtained from the contents specifically pointed out in the description and the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0041] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and serve to explain the principles of the application, but are not intended to limit the scope of the application.

[0042] Figure 1 An example diagram of a trommel screen screening rate optimization method in an embodiment of the application;

[0043] Figure 2 An example diagram of a trommel screen discrete element model in an embodiment of the application. DETAILED DESCRIPTION

[0044] The preferred embodiments of the application will be specifically described below with reference to the accompanying drawings, wherein the drawings constitute a part of this application and are used to illustrate the principles of the embodiments of the application and are not intended to limit the scope of the application.

[0045] One specific embodiment of the application discloses a trommel screen screening rate optimization method, as shown in Figure 1 , comprising:

[0046] S1: constructing a trommel screen discrete element model and a particle plant space model, comprising:

[0047] S11: Construct a drum screen discrete element model, the drum screen discrete element model comprising a cylindrical screen and an internal component, and divide the cylindrical screen discrete element model to obtain a plurality of sub-cylindrical screen discrete element models with the same screen length.

[0048] Exemplarily, any modeling software can be used to establish the drum screen discrete element model. Figure 2 Further, considering the calculation efficiency, resource consumption and other factors, the unnecessary details and complex structures in the drum screen discrete element model are removed, and the drum screen discrete element model is simplified to the model shown in the figure based on the business requirements.

[0049] In the present application, the left side of the drum screen discrete element model is the feeding end.

[0050] After obtaining the drum screen discrete element model, the screen discrete element model can be divided to obtain a plurality of sub-screen discrete element models with the same screen length based on the preset screen length specified by the business requirements.

[0051] The dividing of the cylindrical screen discrete element model to obtain a plurality of sub-cylindrical screen discrete element models with the same screen length comprises:

[0052] Obtaining the drum screen length of the cylindrical screen discrete element model and the particle plant space model length along the drum screen length direction of the particle plant space model, and obtaining the available drum screen length based on the drum screen length and the particle plant space model length along the drum screen length direction;

[0053] Dividing the cylindrical screen discrete element model based on the available drum screen length and the preset length to obtain a plurality of sub-cylindrical screen discrete element models with the same screen length, comprising:

[0054] A1: Setting a division starting point, the division starting point being located on the axis of the cylindrical screen discrete element model, and the shortest distance from the division starting point to the discharge end of the cylindrical screen discrete element model being equal to the available drum screen length;

[0055] A2: Taking the division starting point as the division starting point, the preset length being the step length, dividing a plurality of division points on the axis to divide the cylindrical screen discrete element model to obtain a plurality of sub-cylindrical screen discrete element models with the same screen length by a plane passing through the division points and perpendicular to the axis;

[0056] A3: offsetting the segmentation starting point to the direction of the discharge end by a preset distance, taking the offset point as a new segmentation starting point, judging whether the current offset times exceeds the preset moving times, if not, returning to A2; if yes, taking the result of step A2 as a sub-cylindrical screen discrete element model.

[0057] Initially, the segmentation starting point should be set at the right side of the particle factory space model, therefore, a distance should be reserved between the initial segmentation starting point and the particle factory space model, then the segmentation point is determined according to the segmentation starting point and the preset length, and a plurality of sub-cylindrical screen discrete element models with the same screen length are obtained by sectioning. Then the segmentation starting point is moved based on the preset distance, and the process of dividing the sub-cylindrical screen discrete element model is repeated, and a plurality of sub-cylindrical screen discrete element models are obtained. Finally, a plurality of cylindrical screen discrete element models are obtained, each of which is divided into different numbers of sub-cylindrical screen discrete element models.

[0058] It is worth noting that the preset distance should be set according to the length of the actual screen discrete element model and the preset length. If the preset distance is too close, it may cause the difference between the two sample data obtained by simulation to be too small, so that the training data obtained is relatively close, which leads to the training effect of the sample on the screen efficiency prediction model being small.

[0059] Exemplarily, the preset distance is set to 50mm. The preset moving times can be set according to actual needs, or can be calculated according to the segmentation starting point, the available drum screen length, the preset moving times and the preset distance. Assuming that the length of the drum screen discrete element model is 600mm, the length of the particle factory space model is 15mm, the preset screen length is 100mm, and the preset moving times is 1, the available drum screen length is 585mm, the segmentation starting point is set at the right side of the inlet end face by 15mm, and the division is performed, a total of 5 sub-cylindrical screen discrete element models are obtained, and finally the remaining 85mm is not considered. Then the segmentation starting point is offset to the direction of the discharge end by 50mm, and the segmentation starting point is located at the right side of the inlet end face by 65mm, and the division is performed, a total of 5 sub-cylindrical screen discrete element models are obtained, and finally the remaining 35mm is not considered.

[0060] S12: constructing a particle factory space model, the particle factory space model is placed at the inlet end of the drum screen discrete element model and located inside the cylindrical screen discrete element model, and the particle generation mode and particle parameters of the particle factory space model are set based on business requirements.

[0061] Figure 2 The position of the particle factory space model is not shown in the figure, but it is generally placed on the left side of the drum screen discrete element model so that the particles generated by it can fall into the screen area.

[0062] The present application does not limit the software for constructing the discrete element model, and can adopt discrete element software such as EDEM, YADE, LIGGGHTS, DEMms, SDEM. Exemplarily, EDEM can be adopted to define the particle factory space, including:

[0063] Particle definition: the user can define the physical properties of the particle, such as shape (spherical, block, fiber, etc.), size, mass, density, etc.;

[0064] Particle generation: through the API source code, the generation process of the particle in the simulation space can be realized, including the position, speed, rotation angle, etc. of the particle generation and the particle generation mode, the user can write code to generate a single particle, or can batch generate a particle group with different properties (for example, generate a particle group with different particle sizes according to the different particle sizes defined in the particle definition step);

[0065] Particle behavior control: the particle factory API provides methods to control the motion and interaction behavior of the particle, and the user can specify the collision, friction, combination between particles, and the interaction between the particle and the simulation device, etc.;

[0066] Customized simulation: by modifying and extending the API source file, the user can create a customized simulation scene to realize specific particle flow patterns, particle loading and unloading processes, particle and device interactions, etc.;

[0067] Integrate external data: the API supports importing initial conditions and attribute parameters of particles from external data sources, such as importing particle shapes from CAD models, or obtaining physical properties of particles from experimental data;

[0068] Performance optimization: by writing API code, the user can optimize the simulation performance, for example, by algorithm optimization to reduce the simulation calculation time, or adjust the particle management strategy in the simulation process;

[0069] Visual display: during the simulation running process, the API also provides visual output of the particle state, and the user can control the output of the particle position, speed, etc. through programming to analyze and display the results;

[0070] User interaction and automation: the API allows the user to interactively control the simulation process, such as starting, pausing, stopping the simulation, and dynamically adjusting the particle parameters during the simulation, at the same time, the user can automate a series of simulation tasks through the API to improve work efficiency.

[0071] The above is the common operation of the particle factory space, and the purpose of the particle factory space in the present application is to generate particles that meet business requirements. The specific operation of generating particles, such as the generation method of particles and the particle size, is not limited in the present application, and can be determined according to the business requirements. Exemplarily, the particle generation method can adopt a dynamic method, and the particle size can be 3 mm, 3 mm-4.5 mm, 4.5 mm-6 mm, 6 mm-9 mm, 9 mm-13 mm, or 13 mm-25 mm.

[0072] It can be understood that in the simulation experiment, the proportion of each particle needs to be set. In the process of generating particles, the proportion of each particle is kept unchanged. For example, for particles in the range of 3-4.5 mm, the proportions of particles with sizes of 3.5 mm, 4 mm, and 4.4 mm can be set to 30%, 35%, and 35%, respectively.

[0073] S2: Construct a screen passing rate sample data set based on the drum screen discrete element model and the particle factory space model, wherein the screen passing rate sample data set includes particle size, feeding rate of each particle, drum screen inclination angle, drum screen screen mesh rotation speed, drum screen built-in component rotation speed, and screen passing rate.

[0074] The screen passing rate sample data set is constructed based on the drum screen discrete element model and the particle factory space model, and includes:

[0075] B1: Realize drum screen screening process simulation based on the drum screen discrete element model and the particle factory space model, and obtain simulation results, wherein the simulation results include feeding rate of each particle, spatial position of each particle, drum screen inclination angle, drum screen screen mesh rotation speed, and drum screen built-in component rotation speed corresponding to each time step in the simulation process.

[0076] It can be understood that before the simulation process, simulation parameters need to be set, including: screen body simulation material (mass, density, Young's modulus, sliding friction coefficient, rolling friction coefficient, etc.), setting the motion mode of the cylindrical screen mesh and the built-in rotating component, adding a particle model (for example, setting the radius of a perfect spherical particle), setting the particle factory (location, size, shape, and feeding rate of the particle factory), setting the time step and simulation time, setting the simulation parameter group (here, the rotation speed of the cylindrical screen mesh, the rotation speed of the built-in component, the feeding rate of each particle, and the inclination angle of the drum screen), and setting each simulation parameter according to business requirements. Finally, carry out discrete element simulation. Exemplarily, the screen surface motion mode can be simple harmonic vibration.

[0077] The simulation result refers to the result after the screening process reaches a steady state. The steady state refers to the fact that the sum of the mass of the particles passing through the sub-cylindrical screen mesh discrete element model and the particle discharge rate at the discharge end per unit time is equal to the particle feeding rate.

[0078] After setting the simulation parameters, iterative simulation can be performed until the screening process enters a steady state, and discrete element iteration solving is continued, and at least 10s of steady state simulation is performed, and discrete element simulation data is saved.

[0079] The discrete element simulation data mainly includes position information (three-dimensional coordinates) of particles and geometric bodies (screen mesh), motion information (force, speed, acceleration, angular velocity, torque, etc.), and attribute information (mass, density, Young's modulus, sliding friction coefficient, rolling friction coefficient, etc.). Taking EDEM as an example, the information of particles and geometric models at each time step is stored in an h5 format file, and the naming method is "time step +.h5".

[0080] In the present application, according to the business requirements, the data required by the present application is the feeding rate of each time step corresponding to various particles, the spatial position of each particle, the inclination angle of the drum screen, the screen mesh rotating speed of the drum screen, and the rotating speed of the built-in component of the drum screen.

[0081] The present application divides the discrete element model of the vibrating screen into a plurality of sub-screen discrete element models of the same length based on a preset length, the sub-screen discrete element models are arranged continuously, and the discharge of the previous sub-screen discrete element model is used as the feed of the next sub-screen discrete element model. As can be seen from step A3, by continuously shifting the segmentation starting point, a plurality of cylindrical screen discrete element models are actually obtained. These cylindrical screen discrete element models are controlled by a program to realize simulation under the same working condition parameters. The input parameters of each cylindrical screen discrete element model are the same working condition parameters. Since the number of sub-cylindrical screen discrete element models corresponding to each cylindrical screen discrete element model is different, a plurality of training sample data of sub-cylindrical screen discrete element models of the same screen length based on one working condition parameter are obtained.

[0082] B2: calculating the screen passing rate of different particle sizes corresponding to each sub-cylindrical screen discrete element model based on the spatial position of each particle.

[0083] For the first sub-cylindrical screen discrete element model close to the feed end, the screen passing rate of each particle size is calculated based on the particle size generated in the particle factory spatial model, the feeding rate of various particles, the inclination angle of the drum screen, the screen mesh rotating speed of the drum screen, the rotating speed of the built-in component of the drum screen, and the mass of various sizes of particles below the first sub-cylindrical screen discrete element model.

[0084] For other sub-cylindrical screen discrete element models, the particle size, the feeding rate of various particles, the drum screen inclination, the drum screen screen rotation speed, and the drum screen built-in component rotation speed of the previous sub-cylindrical screen discrete element model are taken as the particle size, the feeding rate of various particles, the drum screen inclination, the drum screen screen rotation speed, and the drum screen built-in component rotation speed of the sub-cylindrical screen discrete element model, and the mass of various sizes of particles below the sub-cylindrical screen discrete element model is calculated based on the particle size, the feeding rate of various particles, the drum screen inclination, the drum screen screen rotation speed, and the drum screen built-in component rotation speed of the sub-cylindrical screen discrete element model to calculate the screening rate of each particle size.

[0085] For each particle size in each sub-cylindrical screen discrete element model, the screening rate is calculated by the following method:

[0086] B1: The number Q of the particle size located below the screen is counted.

[0087] B2: The feeding rate L of the particle size is calculated.

[0088] B3: The screening rate S = Q*M / L, wherein M is the mass of the particle.

[0089] The feeding rate of various particles, the drum screen inclination, the drum screen screen rotation speed, the drum screen built-in component rotation speed, and the screening rate of various particles of each time step corresponding to each sub-cylindrical screen discrete element model are taken as one sample in the screening rate sample data set corresponding to the screen length.

[0090] For each working condition simulation of the particle factory space model, after the steady state, a plurality of groups of simulation data corresponding to the working condition and the preset length screen can be generated at each time step. For example, assuming that the length of the drum screen discrete element model is 600 mm, the length of the particle factory space model is 15 mm, the preset screen length is 100 mm, and the preset moving number is 1, the division starting point is set at the right side of the feeding end face by 20 mm, the division is performed, and a total of 5 sub-cylindrical screen discrete element models are obtained. Then, the division starting point is offset by 50 mm in the direction of the discharge end, the division is performed, and a total of 5 sub-cylindrical screen discrete element models are obtained. Therefore, a total of 10 sub-cylindrical screen discrete element models are obtained. Thus, 10 samples can be obtained at each time step for one working condition simulation, and 10*N samples can be obtained after the working condition simulation is completed, wherein N is the number of time steps.

[0091] B4: It is judged whether the amount of the generated sample data meets the preset quantity requirement. If yes, the generation is stopped; and if no, the current particle generation mode and particle parameters are adjusted based on the business requirement, and B1 is returned.

[0092] The machine learning model is used as the screen passing rate prediction model, and the number of training samples is an important indicator affecting the prediction accuracy of the machine learning model. Therefore, the number of training sample data needs to be set to make the prediction effect of the machine learning model meet the requirements.

[0093] Exemplarily, the sample data amount can be set to 5000. If the current generated sample data amount does not meet the requirements, the sample can be continuously generated by adjusting the particle generation mode and particle parameters.

[0094] S3: training a target machine learning model based on the screen passing rate sample data set to obtain a screen passing rate prediction model.

[0095] Understandably, the screen passing rate sample data set should be preprocessed first to meet the input requirements of the machine learning model. The data preprocessing operation includes: missing value processing: the samples or features containing missing values can be selected to be deleted, or the missing values can be filled by interpolation, mean, median, mode, etc.; abnormal value processing: the abnormal values can be processed by statistical methods (such as Z-Score) or model-based methods (such as IQR); data standardization: scaling the data to a specific range (such as [0, 1]) to make different features have the same scale; data normalization: converting the data to a value between 0 and 1 to eliminate the influence of the dimension between features. Feature selection: selecting features according to importance and correlation to improve the performance of the model.

[0096] Exemplarily, the machine learning model can be a support vector machine, a random forest, a Gaussian process regression, a BP neural network regression, a gradient boosting regression, etc., which is not limited in the present application. A BP neural network model with a Dropout layer can be used as the screen passing rate prediction model, the input sample data includes the feeding rate of each particle, the inclination angle of the drum screen, the screen mesh speed of the drum screen, and the built-in component speed of the drum screen, and the output data is the screen passing rate of various particles. For example, the screen passing rate of 3mm, the screen passing rate of 3mm-4.5mm, the screen passing rate of 4.5mm-6mm, etc. Understandably, the training process of the machine learning model determines the output result of the screen passing rate prediction model. When using the method proposed in the present application, the input and output parameters of the machine learning model can be set according to the needs, so as to obtain the screen passing rate of different particles corresponding to the working condition based on a group of working condition parameters.

[0097] Preferably, the Harris Hawk optimization algorithm is used to optimize the number of neurons, the Dropout coefficient and the batch size of the neural network model, and the optimization results are used as the initial parameters of the neural network model. The specific implementation method of optimizing the BP neural network model by the Harris Hawk optimization algorithm is not limited in the present application.

[0098] S4: obtaining a to-be-tested parameter sample, and inputting the to-be-tested parameter sample into the screening rate prediction model to obtain the screening rate of different particle sizes corresponding to the to-be-tested parameter sample.

[0099] The to-be-tested parameter sample includes the feeding rate of each particle, the drum screen inclination angle, the drum screen screen rotation speed, and the drum screen built-in component rotation speed, which are input into the trained screening rate prediction model to obtain the screening rate of different particle sizes.

[0100] S5: if the screening rate meets the expected target, the to-be-tested parameter sample is used as the control parameter of the drum screen, if not, the particle swarm algorithm is used to optimize the to-be-tested parameter sample, and the optimized parameter is used as the control parameter of the drum screen, so as to improve the screening rate of the to-be-tested vibration screen.

[0101] In the optimization process, the input parameters represented by each particle are input into the screening rate prediction model to obtain the screening rate corresponding to the particle, and the fitness function is calculated based on the screening rate of the particle and the drum screen inclination angle, the drum screen screen rotation speed, and the drum screen built-in component rotation speed corresponding to the particle.

[0102] After obtaining the screening rate of different particle sizes corresponding to the to-be-tested parameter sample in step S4, the screening rate of a certain particle size may not meet the expected target (i.e. the expected value of the screening rate of a certain particle), therefore, each parameter in the to-be-tested parameter sample needs to be adjusted to make the corresponding screening rate meet the expected target.

[0103] The fitness function is calculated based on the screening rate of the particle and the drum screen inclination angle, the drum screen screen rotation speed, and the drum screen built-in component rotation speed corresponding to the particle, including:

[0104]

[0105] wherein y' is the predicted screening rate corresponding to the current particle, is the expected target, a', ω1', ω2' are the drum screen inclination angle, the drum screen screen rotation speed, and the drum screen built-in component rotation speed corresponding to the current particle, a, ω1, ω2 are the drum screen inclination angle, the drum screen screen rotation speed, and the drum screen built-in component rotation speed of the to-be-tested parameter sample, and a, b, c, d are constants which can be set according to trial and error method or experience.

[0106] In the fitness function, the screen rate and the effects of the inclination of the drum screen, the screen mesh speed of the drum screen and the speed of the built-in component of the drum screen in the working condition parameters are comprehensively considered. For example, for the current application scene, the placement position of the drum screen is fixed in advance, and the inclination change range of the drum screen is limited, and the inclination needs to be adjusted in the limited change range to make the screen rate meet the requirements, so that the inclination change can be used as a consideration standard of the optimal parameters when the particles in the particle group are screened.

[0107] The particle swarm algorithm is used to optimize the parameter sample to be tested, and the optimized parameter is used as the control parameter of the drum screen.

[0108] C1: initialize the particle group, screen the sample data meeting the expected target from the screen rate sample data set as the first sample data set, use the feeding rate of various particles, the inclination of the drum screen, the screen mesh speed of the drum screen and the speed of the built-in component of the drum screen in the first sample data set as a particle, use the feeding rate of various particles, the inclination of the drum screen, the screen mesh speed of the drum screen and the speed of the built-in component of the drum screen in the parameter sample to be tested as a particle, and judge whether the total number of the current particles meets the preset number, if yes, execute C3, if not, execute C2.

[0109] Exemplarily, the preset number can be 30.

[0110] C2: statistics the range of each parameter in the first sample data set, randomly take values in the range of each parameter based on the random generation algorithm, thereby generating a plurality of particles, so that the total number of the particles meets the preset number requirement.

[0111] C3: calculate the fitness function of each particle to update the individual optimal value and the global optimal value, judge whether the end condition is met, if yes, execute C5, if not, execute C4.

[0112] C4: update the speed and position of each particle, and return to C3.

[0113] C5: output the optimal position, and the particle group parameter corresponding to the optimal position is the control parameter of the drum screen.

[0114] It can be understood that the updating mode of the position and the speed in the particle swarm algorithm, the updating mode of the individual optimal value and the global optimal value, the adjustment of the inertia weight and the adjustment of the acceleration constant are all known in the art, and the present application does not limit them.

[0115] Compared with the prior art, the drum screen screen-through rate optimization method provided by the embodiment can obtain the screen-through rates of particles of different sizes according to current working condition parameters, judge based on an expected target, and further optimize the current working condition parameters by using a particle swarm algorithm to obtain working condition parameters meeting the expected target if the current working condition parameters do not meet the target requirement, so as to guide a user to design working condition parameters meeting the expected screen-through rate, save process cost, and improve design efficiency; the screen-through efficiency prediction model is constructed based on a machine learning model, can more flexibly meet business requirements, has higher prediction accuracy and calculation efficiency, is strong in real-time performance, and can be adapted to various complex working conditions compared with the traditional probability screen-through model, the kinetic screen-through model, and the discrete element numerical simulation in the prior art; the drum screen discrete element model is divided based on a preset length to obtain a plurality of length-identical sub-screen discrete element models, for each working condition simulation of a particle plant space model, after a steady state, each time step can generate a plurality of sets of simulation data corresponding to the working condition and the preset length screen, compared with the prior art in which each time step can only generate a set of simulation data corresponding to the working condition and the preset length, the efficiency of constructing the screen efficiency sample data set is higher; in addition, the sub-screen discrete element models are continuously arranged, the discharge of a previous sub-screen discrete element model is used as the feed of a next sub-screen discrete element model, compared with the prior art in which sample data is obtained only for a fixed-length screen model, the screen separation sample data obtained by the simulation of the sub-screen discrete element models is closer to the actual vibrating screen separation process.

[0116] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiments can be completed by a computer program instructing related hardware, and the program can be stored in a computer readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0117] The above description is only the preferred embodiment of the present application, but the protection scope of the present application is not limited to this. Any changes or replacements within the technical scope disclosed by the present application can be easily thought of by those skilled in the art, and should be covered within the protection scope of the present application.

Claims

1. A trommel screen throughput rate optimization method characterized by, The method comprises the following steps: constructing a drum screen discrete element model and a particle plant space model; constructing a screen passing rate sample data set based on the drum screen discrete element model and the particle plant space model, the screen passing rate sample data set comprising particle size, feeding rate of each particle, drum screen inclination angle, drum screen screen mesh rotating speed, drum screen built-in component rotating speed, and screen passing rate; performing missing value processing, abnormal value processing, data standardization, and data normalization on the screen passing rate sample data set, and training a target machine learning model based on the screen passing rate sample data set to obtain a screen passing rate prediction model; obtaining a to-be-tested parameter sample, inputting the to-be-tested parameter sample into the screen passing rate prediction model to obtain screen passing rates of different particle sizes corresponding to the to-be-tested parameter sample; if the screen passing rate meets an expected target, taking the to-be-tested parameter sample as a control parameter of the drum screen, if not, optimizing the to-be-tested parameter sample by using a particle swarm algorithm, and taking the optimized parameter as the control parameter of the drum screen, so as to improve the screen passing rate of the to-be-tested vibrating screen; in the optimization process, inputting the input parameter represented by each particle into the screen passing rate prediction model to obtain a screen passing rate corresponding to the particle, and calculating a fitness function based on the screen passing rate of the particle and the drum screen inclination angle, the drum screen screen mesh rotating speed, and the drum screen built-in component rotating speed corresponding to the particle; the method comprises the following steps: constructing a drum screen discrete element model, the drum screen discrete element model comprising a cylindrical screen mesh and a built-in component, and dividing the cylindrical screen mesh discrete element model to obtain a plurality of sub-cylindrical screen mesh discrete element models with the same screen length; constructing a particle plant space model, the particle plant space model being placed at a feeding end of the drum screen discrete element model and being located inside the cylindrical screen mesh discrete element model, and setting a particle generation mode and particle parameters of the particle plant space model based on business requirements; the method comprises the following steps: ， wherein, is a predicted screen pass rate corresponding to the current particle, is an expected target, , is a drum screen inclination, a drum screen screen mesh rotation speed, a drum screen built-in component rotation speed corresponding to the current particle, , , is a drum screen inclination, a drum screen screen mesh rotation speed, a drum screen built-in component rotation speed of the to-be-measured parameter sample, and a, b, c, and d are constants.

2. The method according to claim 1, wherein the method comprises the following steps: obtaining a drum screen length of the cylindrical screen mesh discrete element model and a particle plant space model length of the particle plant space model in the direction of the drum screen length, obtaining an available drum screen length based on the drum screen length and the particle plant space model length in the direction of the drum screen length; dividing the cylindrical screen mesh discrete element model based on the available drum screen length and a preset length, so as to obtain a plurality of sub-cylindrical screen mesh discrete element models with the same screen length.

3. A trommel screen throughput rate optimisation method according to claim 2, characterised in that, the method comprises the following steps: A1: set a segmentation starting point, the segmentation starting point is located on the axis of the cylindrical screen discrete element model, and the shortest distance from the segmentation starting point to the discharge end of the cylindrical screen discrete element model is equal to the available length of the drum screen; A2: take the segmentation starting point as the division starting point, the preset length is the step size, and a plurality of segmentation points are divided on the axis to divide the cylindrical screen discrete element model to obtain a plurality of sub-cylindrical screen discrete element models with the same screen length by a plane passing through the segmentation points and perpendicular to the axis; A3: offset the segmentation starting point by a preset distance in the direction of the discharge end, take the offset point as a new segmentation starting point, and determine whether the current offset number exceeds a preset movement number, if not, return to A2; if yes, take the result of each division in step A2 as a sub-cylindrical screen discrete element model.

4. A trommel screen throughput rate optimisation method according to claim 3, characterised in that, The construction of the screen passing rate sample data set based on the drum screen discrete element model and the particle plant space model comprises: B1: based on the drum screen discrete element model and the particle plant space model, realize the simulation of the drum screen screening process, and obtain a simulation result, the simulation result comprising the feeding rate of various particles, the spatial position of each particle, the inclination angle of the drum screen, the screen mesh rotating speed of the drum screen, and the built-in component rotating speed of the drum screen corresponding to each time step in the simulation process; B2: based on the spatial position of each particle, calculate the screen passing rate of different particle sizes corresponding to each sub-cylindrical screen discrete element model; B3: take the feeding rate of various particles, the inclination angle of the drum screen, the screen mesh rotating speed of the drum screen, the built-in component rotating speed of the drum screen, and the screen passing rate of various particles corresponding to each time step of each sub-cylindrical screen discrete element model as a sample in the screen passing rate sample data set corresponding to the screen length; B4: determine whether the amount of the generated sample data meets a preset quantity requirement, if yes, stop generating; if not, adjust the current particle generation mode and particle parameters based on the business requirement, and return to B1.

5. A trommel screen throughput rate optimisation method according to claim 4, wherein, The screen passing rate prediction model is a neural network model, and the number of neurons, the Dropout coefficient and the batch size of the neural network model are optimized by using the Harris hawk optimization algorithm, and the optimization results are taken as the initial parameters of the neural network model. The particle swarm algorithm is used to optimize the to-be-tested parameter sample, and the optimized parameters are taken as the control parameters of the drum screen, comprising:

6. A trommel screen throughput rate optimisation method according to claim 5, wherein, C1: initialize the particle swarm, select sample data meeting the expected target from the screen passing rate sample data set as a first sample data set, take the feeding rate of various particles, the inclination angle of the drum screen, the screen mesh rotating speed of the drum screen, and the built-in component rotating speed of the drum screen of each data in the first sample data set as a particle, take the feeding rate of various particles, the inclination angle of the drum screen, the screen mesh rotating speed of the drum screen, and the built-in component rotating speed of the drum screen in the to-be-tested parameter sample as a particle, and determine whether the total number of the current particles meets a preset number, if yes, execute C3, if not, execute C2; ​ C2: statistics of the first sample data set of each parameter range, based on random generation algorithm in each parameter range random value, thereby generating a plurality of particles, so that the total number of particles to meet the pre-set quantity requirements; C3: calculate the fitness function of each particle, to update the individual optimal and global optimal value, to determine whether to meet the end condition, if satisfied, then perform C5, if not satisfied, then perform C4; C4: update the speed and position of each particle, return to C3; C5: output the optimal position, the particle group parameters corresponding to the optimal position are the control parameters of the drum screen.

7. A trommel screen throughput rate optimization method according to claim 1, wherein, The screen passing rate of different particle sizes corresponding to each sub-cylindrical screen discrete element model is calculated based on the spatial position of each particle, including: For the first sub-cylindrical screen discrete element model close to the feed end, the screen passing rate of various particle sizes is calculated based on the feed rate of various particles generated in the particle factory spatial model, the inclination angle of the drum screen, the screen speed of the drum screen, the rotation speed of the built-in component of the drum screen, and the mass of various particle sizes below the first sub-cylindrical screen discrete element model; For other sub-cylindrical screen discrete element models, the particle size at the discharge end of the previous sub-cylindrical screen discrete element model, the feed rate of various particles, the inclination angle of the drum screen, the screen speed of the drum screen, the rotation speed of the built-in component of the drum screen are taken as the feed parameters of the current sub-cylindrical screen discrete element model, and the screen passing rate of various particle sizes is calculated based on the feed parameters of the sub-cylindrical screen discrete element model and the mass of various particle sizes below the sub-cylindrical screen discrete element model.

8. A trommel screen throughput rate optimisation method according to claim 7, characterised in that, Including: The simulation results refer to the particle size of each sub-cylindrical screen discrete element model, the feed rate of various particles, the spatial position of each particle, the inclination angle of the drum screen, the screen speed of the drum screen, and the rotation speed of the built-in component of the drum screen in each time step after the screening process reaches a steady state; The steady state refers to the sum of the screen passing particle mass of the sub-cylindrical screen discrete element model and the discharge rate of the particles at the discharge end per unit time being equal to the particle feed rate.

Images