Method for calculating optimal wind speed of large-particle pneumatic conveying pipeline based on digital twinning and machine learning

Abstract

The invention discloses a large-particle pneumatic conveying pipeline optimal wind speed calculation method based on digital twinning and machine learning, and the method comprises the steps: S1, carrying out the digital twinning modeling of a gas fluid region and large-volume particles, including an overall pipeline fluid model, fluid region grid division, a large-particle model, and a large-particle transportation volume fraction algorithm; s2, a gas-solid coupling numerical simulation experiment is carried out to calculate simulation data, and the optimal conveying wind speed for conveying spherical particles with different volumes and masses is obtained; and S3, fitting simulation data by adopting multiple linear regression analysis in machine learning, establishing a pipeline optimal wind speed calculation model, and determining a high-order term order of an independent variable. According to the method, digital twinborn modeling is conducted on the pneumatic conveying pipeline and the large-size particles, the optimal conveying wind speed for conveying the spherical particles of different sizes and masses is obtained through a gas-solid coupling numerical simulation experiment, and it is avoided that data are obtained through complex, tedious and even difficult-to-execute actual experiments.

Description

Technical Field

[0001] The present invention relates to the technical field of computer numerical analysis, and more specifically, to the comprehensive application of digital twin model simulation and machine learning algorithm, and relates to a method for calculating the optimal wind speed of a large particle pneumatic conveying pipeline based on digital twin and machine learning. Background Art

[0002] Pipeline material transportation is an important means of logistics, and the earliest technology used is pneumatic conveying. The working principle of pneumatic conveying is to put the conveyed material into a closed pipeline, control the airflow in the pipeline through a fan, and drive the material to a specified position. It is a specific application of fluidization technology and is usually used to convey tiny particles such as powders. With the continuous development and innovation of technology, pneumatic conveying technology has begun to be successfully applied to the collection and transportation of large-volume materials.

[0003] As a typical gas-solid two-phase flow, pneumatic conveying has always been a research focus in the field of multiphase flow. In the process of pneumatic conveying of large particles, the mass and volume of the particles determine their flow characteristics in the pipeline. The flow characteristics of the gas-solid two-phase flow will affect the effect of pneumatic conveying, thereby affecting the energy consumption of the system. How to determine the appropriate conveying wind speed based on the mass and volume of large-volume particulate materials, properly adjust the fan frequency in the system, ensure smooth transportation, avoid redundancy of fan output power, and reduce system energy consumption, is a practical problem in the current application of large-particle pneumatic systems.

[0004] The patent of this invention starts with a simulation experiment on the gas-solid two-phase flow characteristics of large-particle pneumatic conveying, establishes a large-particle material model and a pipeline fluid domain grid model, simulates the transportation process of large-particle materials of different masses and volumes, analyzes the factors affecting material flow, and obtains the corresponding simulation data of pipeline conveying wind speed. Subsequently, based on the multivariate linear regression algorithm of machine learning, the simulation data is fitted to establish a pipeline optimal wind speed calculation model for calculating the reasonable operating frequency of the fan.

[0005] After searching, the following existing patent technologies have been discovered:

[0006] A method for detecting the movement speed of tobacco shreds during the pneumatic conveying of dense-phase tobacco shreds in a horizontal tube (Announcement No. CN110940826B), characterized in that: a continuous image of tobacco shreds moving in a horizontal tube is obtained by using an image acquisition device, and the movement distance of tobacco shreds is calculated by registering tobacco shreds particles between images, thereby calculating the movement speed of tobacco shreds; the image acquisition device comprises a transparent square pipe connected to the horizontal tube dense-phase tobacco shreds pneumatic conveying pipeline, a light-filling device, an image shooting device, and a host computer control system. The beneficial effects of the present invention are: 1. It can realize the quantitative and accurate measurement of the speed of tobacco shreds during pneumatic conveying; 2. It can quantitatively analyze the flow field of tobacco shreds in the pipeline during pipeline pneumatic conveying; 3. It can be used to quantitatively analyze the relationship between the degree of tobacco shreds breaking and wind speed during pneumatic conveying; 4. It can be used to quantitatively detect and guide the airflow speed during pneumatic conveying at the production site; 5. The present invention provides a new direction for multiphase flow speed detection.

[0007] A real-time detection device for coal powder concentration and phase distribution in a pneumatic conveying pipeline (Announcement No. CN101430269B) discloses a real-time detection device for coal powder concentration and phase distribution in a pneumatic conveying pipeline in the field of gas-solid two-phase flow detection technology. The sensor composite pipe section of the detection device is connected to the measured pipeline through a flange and a sealing gasket, and the γ source is arranged on the axis of the sensor composite pipe section. The γ-ray emission angle of the γ source through the collimator is 360° plane emission, and the emission plane is perpendicular to the axis of the sensor composite pipe section. The γ-ray detector is composed of n γ-ray detection ionization chambers of electrometers to form an annular ionization chamber array coaxial with the sensor measurement pipe section. The device can reduce the influence of the gas-solid two-phase flow pattern and phase distribution on the volume concentration measurement to a negligible level, and under the premise of the same measurement accuracy requirement, the radioactivity of the required γ-ray source is reduced by multiples compared with the existing method, and the space occupied by the measurement device is significantly reduced; the cost is significantly reduced, the radiation protection range is significantly reduced, and the requirements and difficulty of γ-ray protection are greatly reduced.

[0008] After comparison, it is found that the methods adopted in this patent application and the comparative documents are very different, and the technical problems to be solved are also different. The technology difference between the comparative documents and this patent application is quite large. Summary of the invention

[0009] For pneumatic conveying systems, there is currently no effective method for calculating the optimal conveying wind speed of the pipeline based on the mass and volume of large-volume granular materials. Based on the shortcomings of the prior art, the purpose of the present invention is to provide a method for calculating the optimal wind speed of large-particle pneumatic conveying pipelines based on digital twins and machine learning, so as to properly adjust the fan frequency in the system to ensure smooth transportation while reducing system energy consumption.

[0010] The present invention solves the technical problem by adopting the following technical solutions:

[0011] A method for calculating the optimal wind speed of a large particle pneumatic conveying pipeline based on digital twins and machine learning, characterized in that the steps include:

[0012] S1: Digital twin modeling of gas fluid regions and large volume particles, including overall pipeline fluid model, fluid region meshing, large particle model, and large particle transport volume fraction algorithm;

[0013] S2: CFD-DEM is used to perform gas-solid coupling numerical simulation experiments on the pneumatic conveying system to calculate simulation data and obtain the optimal conveying wind speed for transporting spherical particles of different volumes and masses;

[0014] S3: The multivariate linear regression analysis in machine learning is used to fit the simulation data and establish a calculation model for the optimal wind speed in the pipeline. The volume and mass are used as independent variables, and the optimal conveying wind speed is used as the dependent variable.

[0015] By analyzing the residual distribution characteristics and the adjusted R 2 The significance probability p-value of the fitting coefficient is used to determine the order of the higher-order terms of the independent variables, and the model is fitted and corrected. The corrected model output is the optimal pipeline wind speed calculation model.

[0016] Furthermore, in step 1,

[0017] The overall pipeline fluid model is obtained by using SpaceClaim software to perform proportional modeling of the gas fluid area; then, the overall pipeline fluid model is meshed using Gambit software and then imported into Fluent to set the fluid parameters;

[0018] The large particle model is created using a round sphere in EDEM;

[0019] The large particle transport volume fraction algorithm uses the virtual grid method to calculate the volume fraction of the particles on the virtual coarse grid, and add up the volumes of all small grids in the area as the volume of the virtual coarse grid unit. At this time, the volume fraction is calculated as the particle volume divided by the volume of the virtual coarse grid unit, and then the calculated result is evenly distributed to each small grid.

[0020] Moreover, the step 2 specifically comprises:

[0021] The gas phase and solid phase are regarded as continuous phase and discrete phase respectively. Based on the solid phase control equation of Newton's second law and the gas phase control equation based on conservation of mass, conservation of momentum and conservation of energy, the CFD-DEM method is used to carry out gas-solid coupling numerical simulation experiments on the large particle pneumatic conveying system. The transportation data of large particles of different volumes and masses are simulated. Through simulation experiments, the optimal conveying wind speed for transporting large particles of each different volume and mass combination is found.

[0022] Furthermore, in step 3:

[0023] Multiple regression analysis was used to fit the simulation data obtained in step 2, and a regression model was established with volume and mass as independent variables and the optimal conveying wind speed as the dependent variable. The multiple linear regression equation is shown below:

[0024] y=a0+a1x1+a2x2+…+a n x n (1)

[0025] Where a0, a1, …, a n is called the regression coefficient, y is called the dependent variable, x1, x2, …, x n is called the independent variable.

[0026] Moreover, in step 3, by analyzing the residual distribution characteristics and the adjusted R 2 The specific steps to determine the order of the higher-order terms of the independent variables include:

[0027] First, the binary linear regression equation was used to fit the data to obtain the residual distribution diagram, R 2 and p-value;

[0028] Then, observe the residual distribution graph. The residuals should be randomly distributed around the zero point. If the residuals are in a curved shape, it indicates that there is a nonlinear relationship or variance drift in the model. For the independent variables with curved residuals, introduce higher-order terms and refit.

[0029] After that, the order of higher-order terms is gradually increased, and the model is fitted and corrected until R 2 After reaching the maximum value, it starts to decrease, or until the p value of a certain independent variable is greater than 0.05, the order at this time is used as the high-order term finally selected as the independent variable of the model; finally, the corrected model output is the optimal pipeline wind speed calculation model.

[0030] Moreover, in step 3, after adjustment, R 2 The p-value is a measure of the significance of the multiple regression equation. 2 is the original R 2It is a modification of the test statistic that takes into account the number of independent variables in the model and is used to evaluate the model's fit under different combinations of independent variables and which model is better at explaining the variation in the dependent variable. The p-value is the probability value corresponding to the test statistic and is used to determine whether the effect of the independent variable on the dependent variable is statistically significant. When evaluating a multivariate linear regression model, the p-value is usually compared with the adjusted R 2 Use together, R 2 The closer the value is to 1, the more significant the multiple linear regression equation is. When p<α=0.05, the regression model is established.

[0031] Moreover, the significance probability p is between 0.01-0.05, and the smaller the p value, the better.

[0032] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute any of the above-mentioned methods for calculating the optimal wind speed of a large particle pneumatic conveying pipeline based on digital twins and machine learning.

[0033] A computer-readable storage medium stores instructions which, when executed by one or more processors, enable an electronic device to execute any of the above-mentioned methods for calculating the optimal wind speed of a large particle pneumatic conveying pipeline based on digital twins and machine learning.

[0034] The advantages and positive effects of the present invention are:

[0035] (1) The present invention performs digital twin modeling of pneumatic conveying pipelines and large-volume particles, and obtains the optimal conveying wind speed for transporting spherical particles of different volumes and masses through gas-solid coupling numerical simulation experiments, thereby avoiding the need to obtain data through complex, tedious, and even difficult-to-perform actual experiments.

[0036] (2) The present invention uses the multivariate linear regression analysis in machine learning to establish a calculation model for the optimal wind speed in the pipeline, and can flexibly adjust the calculation model and its fitting coefficients according to simulation data. This method can be extended to be used in pneumatic conveying systems with different pipeline structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flow chart for calculating the optimal wind speed of a large particle pneumatic conveying pipeline in an embodiment of the present invention;

[0038] Figure 2 is a virtual grid schematic diagram of a large particle transport volume fraction algorithm in an embodiment of the present invention;

[0039] Figure 3 is a distribution diagram of the independent variable residuals of the binary linear regression fitting in step 3 of the embodiment of the present invention;

[0040] (a) The residual of volume has no obvious distribution pattern; (b) The residual of mass has a relatively obvious curve shape. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application clearer and easier to understand, the present invention is further described below in conjunction with the accompanying drawings and specific implementation methods. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0042] The present invention provides a method for calculating the optimal wind speed of a large particle pneumatic conveying pipeline based on digital twin and machine learning. The overall calibration process is as follows: Figure 1 The overall process of this method is as follows:

[0043] Step 1: Digital twin modeling of pneumatic conveying pipelines and large-volume particles, including overall pipeline fluid model, pipeline meshing, large particle model, large particle transport volume fraction algorithm, etc.

[0044] First, the gas fluid region was modeled in proportion using SpaceClaim software. Then, the model was meshed using Gambit software and imported into Fluent to set the fluid parameters. Then, a large particle model was created using a round sphere in EDEM. Afterwards, since the volume fraction algorithm in the existing Fluent and EDEM coupling interface is only applicable to the coupling calculation of tiny particles, the virtual grid method was used to calculate the volume fraction of the particles on a virtual coarse grid, as shown in the following example. Figure 2 As shown in the figure, the volume of the small grids in the black area is added together as the volume of the virtual coarse grid unit. At this time, the volume fraction is calculated by dividing the particle volume by the volume of the virtual coarse grid unit, and then the calculated result is evenly distributed to each small grid. Through the establishment of the above digital twin model, the coupled calculation of the particle size larger than the fluid domain grid size can be realized. In this embodiment, the pipe diameter is 50cm, and the diameter of the maximum simulated particle can reach 25cm.

[0045] Step 2: Use CFD-DEM to conduct gas-solid coupling numerical simulation experiments on the pneumatic conveying system to obtain the optimal conveying wind speed for transporting spherical particles of different volumes and masses.

[0046] The gas phase and solid phase are regarded as continuous phase and discrete phase respectively, and based on the solid phase control equation of Newton's second law and the gas phase control equation based on mass conservation, momentum conservation and energy conservation, the CFD-DEM method is used to conduct gas-solid coupling numerical simulation experiments on the large particle pneumatic conveying system. The transportation data of large particles of different volumes and masses are simulated, and the optimal conveying wind speed for transporting large particles of different volume and mass combinations is found through simulation experiments.

[0047] Step 3: Use multivariate linear regression analysis in machine learning to fit the simulation data and establish a calculation model for the optimal wind speed in the pipeline. In the model, volume and mass are used as independent variables, and the optimal conveying wind speed is used as the dependent variable. By analyzing the residual distribution characteristics, the adjusted R-square of the fitting results, and the p-value of the fitting coefficient, the order of the higher-order terms of the independent variables is determined. To expand:

[0048] Multiple regression analysis was used to fit the simulation data obtained in step 2, and a regression model was established with volume and mass as independent variables and the optimal conveying wind speed as the dependent variable. The general form of the multiple linear regression equation is as follows:

[0049] y=a0+a1x1+a2x2+…+a n x n (1)

[0050] Where a0, a1, …, a n is called the regression coefficient, y is called the dependent variable, x1, x2, …, x n is called the independent variable.

[0051] Adjusted R 2 The p-value is a measure of the significance of the multiple regression equation. 2 is the original R 2 It is a modification of the test statistic that takes into account the number of independent variables in the model and is used to evaluate the model's fit under different combinations of independent variables and which model is better at explaining the variation in the dependent variable. The p-value is the probability value corresponding to the test statistic and is used to determine whether the effect of the independent variable on the dependent variable is statistically significant. When evaluating a multivariate linear regression model, the p-value is usually compared with the adjusted R 2 Use together, R 2 The closer the value is to 1, the more significant the multivariate linear regression equation is. The significance probability p is generally between 0.01 and 0.05, the smaller the better. When p<α=0.05, the regression model is established.

[0052] First, the binary linear regression equation was used to fit the data to obtain the residual distribution diagram, R 2 and p-value; then, observe the residual distribution diagram. The residual should be randomly distributed around the zero point. If the residual is in a curved shape, it indicates that there may be nonlinear relationships or variance drift in the model. For the independent variables with curved residuals, introduce higher-order terms and refit; then, gradually increase the order of higher-order terms and correct the model until R 2 After reaching the maximum value, it starts to decrease, or until the p value of a certain independent variable is greater than 0.05, the order at this time is used as the high-order term finally selected as the independent variable of the model; finally, the corrected model output is the optimal pipeline wind speed calculation model.

[0053] The large particles of each different volume and mass combination obtained after step 1 and step 2 of this embodiment and their corresponding optimal conveying wind speeds are shown in Tables 1 to 8.

[0054] Table 1 1.44L particle transport experimental data

[0055]

[0056] Table 2 2.14L particle transport experimental data

[0057]

[0058] Table 3 3.05L particle transport experimental data

[0059]

[0060] Table 4 4.19L particle transport experimental data

[0061]

[0062]

[0063] Table 5 5.58L particle transport experimental data

[0064]

[0065] Table 6 6.37L particle transport experimental data

[0066]

[0067] Table 7 7.24L particle transport experimental data

[0068]

[0069] Table 8 8.18L particle transport experimental data

[0070]

[0071] The binary linear fitting results in step 3 of this embodiment are shown in Table 9, and the distribution of the independent variable residuals is as follows: Figure 3 As shown in the figure, it can be clearly seen that the p-values ​​of volume and mass are significantly less than 0.05, indicating that the volume and mass have a significant impact on the transport wind speed. 2 The value is 0.84729, which is too small, indicating that the fitting effect is not good. It is necessary to consider adding other independent variables to improve the fitting accuracy. By observing the residual distribution of the independent variables, it can be seen that the residual of the volume has no obvious distribution pattern (such as Figure 3a), which can be regarded as a random distribution, and the residual of the quality has a relatively obvious curve shape (such as Figure 3 b), indicating that there may be a nonlinear relationship between quality and the model. Therefore, the high-order terms of quality were added to the multiple linear regression model and refitted.

[0072] Table 9 Binary linear fitting results

[0073]

[0074] The results of fitting and correcting the model by introducing high-order terms such as second-order, third-order, fourth-order, and fifth-order terms of mass in step 3 of this embodiment are shown in Tables 10 to 13. It can be concluded that when the second-order, third-order, and fourth-order terms of mass are added, the R 2 The value gradually increases, and the p-value of each independent variable is lower than 0.05. When the fifth-order term of mass is added, the p-values ​​of the higher-order terms of mass are all higher than 0.05.

[0075] Table 10 Ternary regression fitting correction parameters

[0076]

[0077] Table 11 Quaternary regression fitting correction parameters

[0078]

[0079] Table 12 Five-element regression fitting correction parameters

[0080]

[0081] Table 13 Six-element regression fitting correction parameters

[0082]

[0083] Therefore, it is determined that the regression model has the best fitting effect when the fourth-order mass term is introduced. After comprehensive analysis of all the above processes and data, the regression model equation of large particle volume and mass and conveying wind speed is as follows:

[0084]

[0085] In the formula, x1 is the volume of the particle, x2 is the mass of the particle, x3 is the square of the mass of the particle, x4 is the cube of the mass of the particle, and x5 is the fourth power of the mass of the particle.

[0086] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application patent shall be subject to the attached claims.

Claims

1. A method for calculating the optimal wind speed of large particle pneumatic conveying pipeline based on digital twin and machine learning, characterized in that the steps include: S1: Digital twin modeling of gas fluid regions and large volume particles, including overall pipeline fluid model, fluid region meshing, large particle model, and large particle transport volume fraction algorithm; S2: CFD-DEM is used to perform gas-solid coupling numerical simulation experiments on the pneumatic conveying system to calculate simulation data and obtain the optimal conveying wind speed for transporting spherical particles of different volumes and masses; S3: Multivariate linear regression analysis in machine learning is used to fit the simulation data and establish a calculation model for the optimal wind speed in the pipeline. The volume and mass are used as independent variables, and the optimal conveying wind speed is used as the dependent variable. The residual distribution characteristics and the adjusted R of the fitting results are analyzed. 2 The significance probability p-value of the fitting coefficient is used to determine the order of the higher-order terms of the independent variables, and the model is fitted and corrected. The corrected model output is the optimal pipeline wind speed calculation model.

2. The method for calculating the optimal wind speed of large particle pneumatic conveying pipeline based on digital twin and machine learning according to claim 1 is characterized in that: In the step 1, The overall pipeline fluid model is obtained by using SpaceClaim software to perform proportional modeling of the gas fluid area; then, the overall pipeline fluid model is meshed using Gambit software and then imported into Fluent to set the fluid parameters; The large particle model is created using a round sphere in EDEM; The large particle transport volume fraction algorithm uses the virtual grid method to calculate the volume fraction of the particles on the virtual coarse grid, and add up the volumes of all small grids in the area as the volume of the virtual coarse grid unit. At this time, the volume fraction is calculated as the particle volume divided by the volume of the virtual coarse grid unit, and then the calculated result is evenly distributed to each small grid.

3. The method for calculating the optimal wind speed of large particle pneumatic conveying pipeline based on digital twin and machine learning according to claim 1 is characterized in that: The step 2 specifically includes: The gas phase and solid phase are regarded as continuous phase and discrete phase respectively. Based on the solid phase control equation of Newton's second law and the gas phase control equation based on conservation of mass, conservation of momentum and conservation of energy, the CFD-DEM method is used to carry out gas-solid coupling numerical simulation experiments on the large particle pneumatic conveying system. The transportation data of large particles of different volumes and masses are simulated. Through simulation experiments, the optimal conveying wind speed for transporting large particles of each different volume and mass combination is found.

4. The method for calculating the optimal wind speed of large particle pneumatic conveying pipeline based on digital twin and machine learning according to claim 1 is characterized in that: In step 3: Multiple regression analysis was used to fit the simulation data obtained in step 2, and a regression model was established with volume and mass as independent variables and the optimal conveying wind speed as the dependent variable. The multiple linear regression equation is shown below: y=a0+a1x1+a2x2+…+a n x n (1) Where a0, a1, …, a n is called the regression coefficient, y is called the dependent variable, x1, x2, …, x n is called the independent variable.

5. The method for calculating the optimal wind speed of large particle pneumatic conveying pipeline based on digital twin and machine learning according to claim 4 is characterized in that: In step 3, by analyzing the residual distribution characteristics and the adjusted R 2 The specific steps to determine the order of the higher-order terms of the independent variables include: First, the binary linear regression equation was used to fit the data to obtain the residual distribution diagram corresponding to volume and mass, the original R 2 and p-value; Then, observe the residual distribution graph. The residuals should be randomly distributed around the zero point. If the residuals are in a curved shape, it indicates that there is a nonlinear relationship or variance drift in the model. For the independent variables with curved residuals, introduce higher-order terms and refit. After that, the order of the higher-order terms is gradually increased, and the model is fitted and corrected until the adjusted R 2 After reaching the maximum value, it starts to decrease, or until the p-value of a certain independent variable is greater than 0.05, the order at this time is used as the high-order term finally selected as the independent variable of the model; Finally, the corrected model output is the optimal wind speed calculation model for the pipeline.

6. The method for calculating the optimal wind speed of large particle pneumatic conveying pipeline based on digital twin and machine learning according to claim 5 is characterized in that: In step 3, after adjustment, R 2 The p-value is a measure of the significance of the multiple regression equation. 2 is the original R 2 A modification of , which takes into account the number of independent variables in the model, and is used to evaluate the model fit under different combinations of independent variables, and which model is better in explaining the variation of the dependent variable; The p-value is the probability value corresponding to the test statistic, which is used to determine whether the effect of the independent variable on the dependent variable is statistically significant. When evaluating a multivariate linear regression model, the p-value is similar to the adjusted R 2 Use together, R 2 The closer the value is to 1, the more significant the multiple linear regression equation is; When the significance probability p<α=0.05, the regression model is established.

7. The method for calculating the optimal wind speed of large particle pneumatic conveying pipeline based on digital twin and machine learning according to claim 6 is characterized in that: The significance probability p is between 0.01 and 0.05, and the smaller the p value, the better.

8. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute the optimal wind speed calculation method for large particle pneumatic conveying pipeline based on digital twin and machine learning as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions which, when executed by one or more processors, cause the electronic device to execute the method for calculating the optimal wind speed for a large particle pneumatic conveying pipeline based on digital twins and machine learning as described in any one of claims 1 to 7.

Images