Construction method of plunger pump digital twin system based on CFD result

Abstract

The invention discloses a method for constructing a plunger pump digital twin system based on a CFD result. The method comprises the following steps: constructing a flow field calculation model of a plunger pump; the working condition parameters serve as input, physical field data in the plunger pump are obtained through a CFD solver, and a set of input and output data pairs are obtained; determining the change gradient according to the weight of the working condition parameter, updating the working condition parameter, repeatedly carrying out CFD solution until the value of the working condition parameter exceeds a set range, and gathering all input and output data pairs to obtain a sample data set; constructing an order reduction model of the plunger pump by utilizing a model order reduction technology, and training the order reduction model by adopting a neural network until a standard mean absolute error of the order reduction model meets a set requirement; and applying the trained order reduction model to a digital twin system, and inputting local real-time data of the plunger pump detected by a physical sensor. According to the invention, real-time monitoring and display of operation of the plunger pump are realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent monitoring and operation and maintenance of plunger pumps, and in particular to a method for constructing a digital twin system of a plunger pump based on CFD results. Background Art

[0002] As a key component in the hydraulic system, the operating status of the plunger pump directly affects the performance and safety of the entire system. The plunger pump has a complex structure and variable operating conditions. Traditional plunger pump operation monitoring methods mostly rely on empirical formulas and simple sensor data, which are difficult to accurately reflect the complex flow conditions and operating status inside the pump.

[0003] With the development of computational fluid dynamics (CFD) technology, the flow conditions inside the plunger pump can be simulated and analyzed more accurately. CFD simulation of the plunger pump generally requires geometric pre-processing, meshing, calculation settings, calculation results, post-processing and other steps. CFD solution usually also requires the support of a high-performance computing platform. Therefore, direct use of CFD simulation for real-time monitoring will face problems such as large amount of calculation and poor real-time performance. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention proposes a method for constructing a digital twin system of a plunger pump based on CFD results. This method can accurately reflect the flow conditions and operating status inside the plunger pump in real time, provide technical support for real-time monitoring of the status of the plunger pump unit, and improve the intelligence level of the overall unit.

[0005] The specific technical solutions are as follows:

[0006] A method for constructing a plunger pump digital twin system based on CFD results includes the following steps:

[0007] S1: Construct a flow field calculation model consistent with the actual size of the plunger pump;

[0008] S2: Inputting working condition parameters into the flow field calculation model, then meshing the flow field calculation model, setting boundary conditions, and dynamic mesh, and performing numerical calculations through the solver to obtain the physical field data in the plunger pump as output, thereby obtaining a set of input-output data pairs;

[0009] S3: According to the weights of different working condition parameters, determine their change gradients and obtain updated working condition parameters; determine whether the values ​​of the updated working condition parameters exceed the set range. If not, repeat S2-S3 using the updated working condition parameters; if yes, collect all existing input and output data pairs to obtain a sample data set; the setting ranges of different working condition parameters are obtained according to the actual working conditions of the plunger pump;

[0010] S4: Use model reduction technology to build a reduced-order model of the plunger pump, and use a neural network to train the reduced-order model; the standard mean absolute error of the trained reduced-order model meets the set requirements;

[0011] S5: Apply the trained reduced-order model to the real-time online digital twin system of the plunger pump, wherein the input of the digital twin system is the local real-time data detected by the physical sensors arranged on the plunger pump, and the output is the physical field data of the plunger pump obtained by online simulation of the trained reduced-order model.

[0012] Furthermore, in S2, the operating parameters include: oil properties, plunger pump speed, swash plate inclination angle, oil inlet pressure, oil return pressure, and oil film thickness.

[0013] Furthermore, in S2, the physical field data includes pressure, velocity, and cavitation data in the plunger pump.

[0014] Furthermore, in S3, the change gradient of the operating condition parameter is obtained by analyzing the measured data of the whole machine factory within a certain period of time.

[0015] Furthermore, in S4, the calculation expression of the standard mean absolute error is as follows:

[0016]

[0017] In the formula, MAE represents the mean absolute error, i represents the i-th operating parameter, N represents the number of samples in the training set, S(i) represents the sample value of the i-th operating parameter, that is, the output value corresponding to the operating parameter in the training set; S′(i) represents the model simulation value of the i-th operating parameter, and S represents the corresponding rule for mapping the independent variable “i-th operating parameter” to the dependent variable “sample value”.

[0018] A digital twin system is obtained according to the method for constructing a digital twin system of a plunger pump based on CFD results.

[0019] Furthermore, the simulation error result is displayed in real time, and the simulation error result is: the standard mean absolute error between the data collected by other sensors except the physical sensors that input local real-time data into the digital twin system and the predicted output data of the digital twin system.

[0020] Furthermore, when an abnormality or failure is detected, the digital twin system automatically sends out a warning signal, and the warning signal is given in the form of sound or light.

[0021] Furthermore, the output result display forms include: table, line graph, bar graph, scatter plot, contour map, profile map, 3D stereogram, dynamic picture and video.

[0022] The beneficial effects of the present invention are:

[0023] The present invention uses high-precision simulation calculations of CFD technology and a neural network to train CFD results to generate a corresponding reduced-order model, and then applies the trained reduced-order model to the real-time online digital twin system of the plunger pump to achieve automatic monitoring and early warning. The digital twin system prepared by this method has a smaller capacity than the CFD calculation model and a higher degree of model lightweight. It can quickly obtain output physical field data according to the input working conditions of the plunger pump, and the simulation accuracy remains unchanged, which is more efficient than the CFD simulation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flow chart of a method for constructing a digital twin system of a plunger pump based on CFD results in an embodiment of the present invention.

[0025] Figure 2 It is a flowchart of constructing a training set and a test set in an embodiment of the present invention.

[0026] Figure 3 It is a flowchart of the construction and use of the digital twin system in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The present invention will be described in detail below according to the accompanying drawings and preferred embodiments, and the purpose and effect of the present invention will become more clear. The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0028] like Figure 1 As shown, a method for constructing a plunger pump digital twin system based on CFD results specifically includes the following steps:

[0029] S1: Construct a flow field calculation model consistent with the actual size of the plunger pump.

[0030] S2: Input the working parameters into the flow field calculation model, the working parameters include: oil properties, plunger pump speed, swash plate inclination, oil inlet pressure, oil return pressure, oil film thickness. Then mesh the flow field calculation model, set boundary conditions, set dynamic mesh, and perform numerical calculations through the solver to obtain the physical field data in the plunger pump as output, and obtain a set of input and output data pairs. Preferably, the physical field data includes output data such as pressure, velocity, and cavitation in the plunger pump.

[0031] Specifically, the principle of meshing is to use a variety of sizes or densities to divide the mesh and verify the output data of the numerical calculation, that is, to perform mesh independence verification. When the mesh density increases to a certain level, the result basically does not change. The calculation result and efficiency under this critical mesh density are the highest.

[0032] The boundary condition settings include: the piston pump's cylinder speed, oil inlet pressure, oil return pressure, and saturated vapor pressure.

[0033] The numerical calculation by the solver is to solve the NS equations by using the stiffness matrix formed by the grid. Preferably, the solver is selected from various commercial solvers and open source solvers, such as Fluent, CFX, STAR-CCM+, PumpLinx, OpenFoam, etc.

[0034] S3: Figure 2 As shown, according to the weights of different working condition parameters, their change gradients are determined to obtain updated working condition parameters, and it is determined whether the values ​​of the updated working condition parameters exceed the set range. The setting ranges of different working condition parameters are obtained according to the actual working conditions of the plunger pump. If not, S2-S3 are repeatedly executed using the updated working condition parameters; if so, all existing input and output data pairs are collected to obtain a sample data set, each input and output data pair is a sample, and the sample data set is divided into a training set and a test set.

[0035] The change gradient of the operating parameters is obtained based on the actual measured data analysis within a certain period of time (for example, one year) of the whole machine factory. For example, the output power of the plunger pump is used as a weighted measurement indicator to obtain the change gradient of the plunger pump speed and return oil pressure in the input operating parameters. For another example, the working range of the return oil pressure of the whole machine is in the range of 5-30MPa. After analyzing the actual plunger pump operation data, it is found that the plunger pump works in the range of 10-20MPa most of the time. This range is recorded as the commonly used range, and the remaining ranges are recorded as uncommon ranges. In order to ensure the simulation accuracy of the reduced-order model constructed subsequently, the change gradient of the return oil pressure is 0.5% in this commonly used range; at the same time, in order to improve the training efficiency of the reduced-order model constructed subsequently, the change gradient of the return oil pressure in the uncommon range can be appropriately expanded, such as to 2%.

[0036] S4: Use model reduction technology to build a reduced-order model of the plunger pump. Model reduction technology includes the Proper Orthogonal Decomposition (POD) method and the Galerkin method. A neural network is used to train the reduced-order model, where the training set is used to train the reduced-order model, and the test set is used to verify the training effect of the reduced-order model. Finally, a trained reduced-order model is obtained. The standard mean absolute error NMAE of the trained reduced-order model is required to meet the set requirements (i.e., less than the set threshold). The smaller the value of the standard mean absolute error, the higher the accuracy of the reduced-order model. Its calculation expression is as follows:

[0037]

[0038] Wherein, MAE represents the mean absolute error, i represents the i-th operating parameter, N represents the number of samples in the training set, S(i) represents the sample value of the i-th operating parameter (i.e., the output value corresponding to the operating parameter in the training set), S′(i) represents the model simulation value of the i-th operating parameter, and S represents the corresponding rule for mapping the independent variable “i-th operating parameter” to the dependent variable “sample value”.

[0039] S5: Figure 3 As shown, the trained reduced-order model is applied to the real-time online digital twin system of the plunger pump. The input of the digital twin system is the local real-time data detected by the physical sensors arranged on the plunger pump, and the output is the physical field data of the plunger pump obtained by online simulation of the trained reduced-order model in the digital twin system.

[0040] This embodiment also proposes a digital twin system, which is obtained by adopting the above-mentioned method for constructing a digital twin system of a plunger pump based on CFD results. Optionally, data collected by other sensors arranged on the plunger pump (i.e., other sensors except physical sensors that input local real-time data into the digital twin system) are compared with the predicted output data of the digital twin system and the accuracy is confirmed, that is, the real-time standard mean absolute error of the digital twin system is calculated, and the error result is displayed in real time in the digital twin system, thereby realizing accurate operation and maintenance and real-time monitoring of the plunger pump.

[0041] Preferably, when an abnormality or fault is detected, the digital twin system can automatically send out a warning signal to prompt the operator to take corresponding measures; the warning signal is given by sound or light to ensure that the operator can respond in time.

[0042] Preferably, the output result display forms of the digital twin system include: tables, line graphs, bar graphs, scatter plots, contour plots, profile graphs, 3D stereograms, dynamic pictures and videos.

[0043] The present invention establishes a flow field calculation model of the plunger pump, and takes the CFD calculation input and output results of different working conditions as sample data sets, and trains the reduced-order model through neural network training, thereby obtaining a real-time online digital twin system applied to the plunger pump, which realizes real-time monitoring and display of the plunger pump operation, thereby improving the reliability and stability of the plunger pump operation.

[0044] Those skilled in the art can understand that the above are only preferred examples of the invention and are not intended to limit the invention. Although the invention is described in detail with reference to the above examples, those skilled in the art can still modify the technical solutions recorded in the above examples or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, etc. made within the spirit and principle of the invention shall be included in the protection scope of the invention.

Claims

1. A method for constructing a plunger pump digital twin system based on CFD results, characterized in that: The following steps are involved: S1: Construct a flow field calculation model consistent with the actual size of the plunger pump; S2: Inputting working condition parameters into the flow field calculation model, then meshing the flow field calculation model, setting boundary conditions, and dynamic mesh, and performing numerical calculations through the solver to obtain the physical field data in the plunger pump as output, thereby obtaining a set of input-output data pairs; S3: According to the weights of different working condition parameters, determine their change gradients and obtain updated working condition parameters; determine whether the values ​​of the updated working condition parameters exceed the set range. If not, repeat S2-S3 using the updated working condition parameters; if yes, collect all existing input and output data pairs to obtain a sample data set; the setting ranges of different working condition parameters are obtained according to the actual working conditions of the plunger pump; S4: Use model reduction technology to build a reduced-order model of the plunger pump, and use a neural network to train the reduced-order model; the standard mean absolute error of the trained reduced-order model meets the set requirements; S5: Apply the trained reduced-order model to the real-time online digital twin system of the plunger pump, wherein the input of the digital twin system is the local real-time data detected by the physical sensors arranged on the plunger pump, and the output is the physical field data of the plunger pump obtained by online simulation of the trained reduced-order model.

2. The method for constructing a plunger pump digital twin system based on CFD results according to claim 1, characterized in that: In S2, the operating parameters include: oil properties, plunger pump speed, swash plate inclination, oil inlet pressure, oil return pressure, and oil film thickness.

3. The method for constructing a plunger pump digital twin system based on CFD results according to claim 1, characterized in that: In S2, the physical field data includes pressure, velocity, and cavitation data in the plunger pump.

4. The method for constructing a plunger pump digital twin system based on CFD results according to claim 1, characterized in that: In S3, the change gradient of the operating condition parameter is obtained by analyzing the measured data of the whole machine factory within a certain period of time.

5. The method for constructing a plunger pump digital twin system based on CFD results according to claim 1, characterized in that: In S4, the calculation expression of the standard mean absolute error is as follows: In the formula, MAE represents the mean absolute error, i represents the i-th operating parameter, N represents the number of samples in the training set, and S(i) represents the sample value of the i-th operating parameter, that is, the output value corresponding to the operating parameter in the training set; S′(i) represents the model simulation value of the i-th operating condition parameter, and S represents the corresponding rule mapping the independent variable "i-th operating condition parameter" to the dependent variable "sample value".

6. A digital twin system, characterized in that: The method for constructing a digital twin system of a plunger pump based on CFD results is obtained according to any one of claims 1 to 5.

7. The digital twin system according to claim 6, characterized in that: The simulation error results are displayed in real time. The simulation error results are: the standard mean absolute error between the data collected by other sensors except the physical sensors that input local real-time data into the digital twin system and the predicted output data of the digital twin system.

8. The digital twin system according to claim 6, characterized in that: When an abnormality or failure is detected, the digital twin system automatically sends out a warning signal, which is given in the form of sound or light.

9. The digital twin system according to claim 6, characterized in that: The output results can be displayed in the following forms: table, line graph, bar graph, scatter plot, contour graph, profile graph, 3D stereogram, dynamic picture and video.

Images