CN115421400A - Digital twin virtual-real consistency discrimination method for service health state intelligent recognition decision - Google Patents
Digital twin virtual-real consistency discrimination method for service health state intelligent recognition decision Download PDFInfo
- Publication number
- CN115421400A CN115421400A CN202211136899.0A CN202211136899A CN115421400A CN 115421400 A CN115421400 A CN 115421400A CN 202211136899 A CN202211136899 A CN 202211136899A CN 115421400 A CN115421400 A CN 115421400A
- Authority
- CN
- China
- Prior art keywords
- data
- consistency
- digital twin
- accuracy
- simulation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000036541 health Effects 0.000 title claims abstract description 73
- 238000012850 discrimination method Methods 0.000 title claims description 6
- 238000000034 method Methods 0.000 claims abstract description 30
- 238000004364 calculation method Methods 0.000 claims abstract description 10
- 238000004088 simulation Methods 0.000 claims description 62
- 238000012549 training Methods 0.000 claims description 57
- 238000003745 diagnosis Methods 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 30
- 238000013527 convolutional neural network Methods 0.000 claims description 16
- 238000005259 measurement Methods 0.000 claims description 15
- 238000005516 engineering process Methods 0.000 claims description 8
- 230000002452 interceptive effect Effects 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000005094 computer simulation Methods 0.000 claims description 5
- 238000012800 visualization Methods 0.000 claims description 5
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 230000001360 synchronised effect Effects 0.000 claims description 4
- 238000005260 corrosion Methods 0.000 claims description 3
- 230000007797 corrosion Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000010223 real-time analysis Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 abstract description 2
- 238000002474 experimental method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000007246 mechanism Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000005299 abrasion Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 239000006247 magnetic powder Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B17/00—Systems involving the use of models or simulators of said systems
- G05B17/02—Systems involving the use of models or simulators of said systems electric
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Automation & Control Theory (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The application discloses a digital twin virtual-real consistency judging method for service health state intelligent recognition decision, which is a digital twin consistency evaluation method for constructing an initial consistency, value consistency, trend consistency and engineering consistency four-dimensional structure based on health state recognition accuracy calculation, provides a judging method for digital twin consistency and provides a basis for evolution iteration of a digital twin model to a certain extent.
Description
Technical Field
The application relates to a digital twin virtual-real consistency judging method, in particular to a digital twin virtual-real consistency judging method for intelligent identification decision of service health states.
Background
The digital twin technology guides the real-time evolution of the twin model according to the change of the running state of the physical object through the interaction of data between the physical object and the virtual model, feeds back the prejudgment result to a diagnosis control center of the digital twin system through simulation analysis and helps the physical entity to optimize and make decisions.
By reflecting the physical entity to the digital model, various high-simulation operations can be performed on the model in the digital field, the defects of high implementation difficulty and high cost in the operation of the physical entity are avoided, and the digital twin technology is combined with various modern disciplines simultaneously and has distinct technical advantages.
At present, the application of the digital twinning technology in various fields is researched, but rarely is the research on the consistency judgment of the digital twinning physical entity and the digital model.
Disclosure of Invention
The method is based on the calculation of the health state identification accuracy, and a digital twin consistency evaluation method of a four-dimensional structure with initial consistency, value consistency, trend consistency and engineering consistency is constructed, so that a discrimination method is provided for the digital twin consistency, and a basis is provided for the evolution iteration of a digital twin model to a certain extent.
In some embodiments of the present application, the present application provides a digital twin consistency discrimination method for service health state intelligent recognition decision, the method is formed based on an adjustment training set, so as to drive health state recognition accuracy calculation, and the method mainly includes the following steps:
(1) Judging the initial consistency of the digital twin: constructing a digital twin of a physical object in modeling software and performing dynamic simulation to obtain simulation data in a normal state, a fault state and different health states; comparing simulation data and actual measurement data under different health states, and theoretically calculating frequency and fault characteristic frequency to verify the initial consistency of the virtual space and the real space;
(2) Judging the consistency of the digital twin values: driving a deep convolutional neural network model by simulation data, and calculating the health state identification accuracy; adopting simulation data of different health states as a training set and actual measurement data as a test set, calculating the accuracy of fault diagnosis, and verifying the value consistency of virtual and real spaces in the aspect of intelligent identification decision of service health states;
(3) Judging the consistency of the digital twin trend: constructing a training set of simulation data and measured data, training a deep convolutional neural network model, keeping a test set unchanged, calculating the recognition accuracy of the health state, and verifying the trend consistency of virtual and real space data when the simulation data in the mixed data occupy different proportions;
(4) Judging the consistency of digital twin engineering: respectively constructing training sets by using the measured data of the mixed data, training a deep convolutional neural network model, keeping the testing sets unchanged, and calculating the recognition accuracy of the health state; and when the number of the actually measured data in the training set accounts for different proportions, acquiring the accuracy of fault diagnosis and verifying the engineering consistency of the digital twin.
In some embodiments of the present application, the method further comprises (5) a step of determining the digital twin precision consistency of the service decision: and (3) iteratively optimizing digital twins through virtual-real space interactive feedback, then repeating the steps (1), (2), (3) and (4), and respectively observing the accuracy of the health state identification until the digital twins are advanced into an accurate consistency model which accords with the intelligent identification decision of the health state.
In some embodiments of the present application, the physical object is a planetary gearbox.
In some embodiments of the present application, the physical object is a planetary gearbox sun gear.
In some embodiments of the present application, four health states common to planetary gearbox sun gears are determined, respectively worn gears, pitted gears, broken gears, and normal gears.
In some embodiments of the present application, when the physical object is a planetary gearbox, the following steps are performed:
(1) Judging the initial consistency of the digital twin: the method comprises the steps that a planetary gearbox digital twin is built in CAD software, and dynamic simulation is carried out through ADMAS to obtain simulation data under four different health states, namely a normal state, a wear fault, a tooth breakage fault and a pitting corrosion fault; verifying simulation data and actual measurement data under four health states, and meshing frequency and fault characteristic frequency of theoretical calculation;
(2) And (3) judging the consistency of the digital twin values: driving a deep convolution neural network model by simulation data, and calculating the health state identification accuracy; adopting simulation data of different health states as a training set and actual measurement data as a test set, calculating the accuracy of fault diagnosis, and verifying the value consistency of virtual and real spaces in the aspect of intelligent identification decision of service health states;
(3) Judging the consistency of the digital twin trend: constructing a training set of the simulation data and the measured data, training a deep convolution neural network model, keeping the test set unchanged, and calculating the recognition accuracy of the health state; when simulation data in the mixed data account for different proportions, the accuracy of fault diagnosis is calculated, and the consistency of the trend of virtual and real space data is verified;
(4) Judging the consistency of the digital twin engineering: and respectively constructing training sets by using the measured data of the mixed data, training a deep convolution neural network model, keeping the testing sets unchanged, and calculating the recognition accuracy of the health state. When the number of the measured data in the training set is 300 groups and 600 groups respectively, the accuracy of fault diagnosis is calculated, and the project consistency of digital twins is verified in the actual scene that the data is insufficient and the field is difficult to obtain in the fault diagnosis of the rotary machine.
In some embodiments of the present application, in the step (2), simulation data of different health states are used as a training set, and measured data is used as a test set, wherein 300 groups of four health states are used as the training set, 1200 groups of simulation data are used as the training set, and 100 groups of measured data are used as the test set.
In some embodiments of the present application, in the (3), the simulation data in the mixed data account for 100%, 75%, 50%, and 25%, respectively.
In some embodiments of the present application, the method further comprises (5) a digital twin exact-consistency determination of the service decision: and (3) iteratively optimizing digital twins through virtual-real space interactive feedback, then repeating the steps (1), (2), (3) and (4), and respectively observing the accuracy of the health state identification until the digital twins are advanced into an accurate consistency model which accords with the intelligent identification decision of the health state.
In some embodiments of the present application, the virtual-real space interactive feedback is based on a technology that edge computing and cloud-edge cooperative scheduling are used as a core and edge-side-oriented data real-time processing is performed, so that data for real-time analysis is transmitted to an edge device closest to a data source through a network and is locally stored, and time-delay insensitive data is directly transmitted to a cloud center; the method comprises the following steps of realizing self-adaptive optimization scheduling of cloud-edge data based on an intelligent algorithm, and realizing digital twin high-fidelity behavior simulation and real-time state visualization driven by cloud-end data; and then the physical space is fed back and controlled, and real-time accurate mapping and synchronous evolution of the virtual space and the real space are realized.
Drawings
To the accomplishment of the foregoing and related ends, the application, then, describes certain illustrative aspects in connection with the following description and the annexed drawings, indicating various ways in which the principles disclosed herein may be practiced, and all aspects and equivalents thereof are intended to be within the scope of the claimed subject matter. The above and other objects, features and advantages of the present application will become more apparent from the following detailed description when read in conjunction with the accompanying drawings. Throughout this application, like reference numerals generally refer to like parts or elements.
FIG. 1 is a flowchart illustrating the determination of one embodiment of the present application;
FIG. 2 is a schematic diagram illustrating four common health states of a planetary gearbox sun gear in accordance with one embodiment of the present application;
FIG. 3 is a schematic view of a planetary gear configuration according to one embodiment of the present application;
FIG. 4 is simulation data of the planet gear wear health status according to one embodiment of the present application;
FIG. 5 is a planetary gearbox fault diagnosis test stand of one embodiment of the present application;
FIG. 6 is measured data of the planetary gearbox in a healthy wear state according to one embodiment of the present application;
FIG. 7 is a comparison of simulated values of the planetary gearbox of one embodiment of the present application with theoretical and measured values;
FIG. 8 is a network topology diagram of one embodiment of the present application;
FIG. 9 is a data partitioning for data enhancement of acquired simulated and measured data according to one embodiment of the present application;
FIG. 10 is a graph of diagnostic accuracy for one embodiment of the present application;
FIG. 11 is a T-SNE visualization presentation diagram of one embodiment of the present application;
FIG. 12 is a graph of accuracy versus scale for one embodiment of the present application;
FIG. 13 is a comparison graph of accuracy of unadditized simulation data for one embodiment of the present application;
FIG. 14 is a schematic diagram of a six-dimensional structure digital twin evolutionary mechanism of one embodiment of the present application;
FIG. 15 is a schematic diagram of virtual-real interaction and feedback control of cloud-edge coordination according to an embodiment of the present application;
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without making any creative effort belong to the protection scope of the present application.
Taking a planetary gearbox as an example, a digital twin consistency discrimination method for service health state intelligent recognition decision mainly comprises the following steps:
(1) And (5) judging the initial consistency of the digital twin. The method comprises the steps of constructing a planetary gearbox digital twin in CAD software and carrying out dynamic simulation through ADMAS to obtain simulation data under four different health states of a normal state, a wear fault, a tooth breakage fault and a pitting corrosion fault. The simulation data and the measured data under the four health states, the meshing frequency calculated theoretically and the fault characteristic frequency are almost completely equal, and the fact that the virtual space and the real space have initial consistency is verified.
(2) And (5) judging the consistency of the digital twin values. And driving a deep convolutional neural network model by the simulation data, and calculating the health state identification accuracy. The accuracy of fault diagnosis calculation is 80.2% by adopting 300 groups of four health states, 1200 groups of simulation data as a training set and 100 groups of actual measurement data as a test set, and the fact that the virtual space and the real space have higher value consistency in the aspect of intelligent identification decision of the health states is verified.
(3) And (5) judging the consistency of the digital twin trend. And constructing a training set of the simulation data and the measured data, training a deep convolution neural network model, keeping the test set unchanged, and calculating the recognition accuracy of the health state. When the simulation data in the mixed data account for 100%, 75%, 50% and 25%, the accuracy rates of fault diagnosis are respectively 80.2%, 84.3%, 90.5% and 92.6%, and it is verified that the virtual and real space data have trend consistency but do not reach accurate consistency.
(4) And (5) judging the consistency of the digital twin engineering. And respectively constructing a training set by using the measured data of the mixed data, training a deep convolutional neural network model, keeping the testing set unchanged, and calculating the recognition accuracy of the health state. When the number of the measured data in the training set is 300 groups and 600 groups respectively, the accuracy of fault diagnosis is 69.7 percent, 78.1 percent and lower than 80.2 percent respectively. The method proves that the digital twin has engineering consistency in the actual scene that data is insufficient in fault diagnosis of the rotating machinery and the site is difficult to obtain.
Preferably, the method further comprises the step (5) of judging the digital twin accurate consistency of the service decision: and (3) iteratively optimizing digital twins through virtual-real space interactive feedback, then repeating the steps (1), (2), (3) and (4), and respectively observing the accuracy of the health state identification until the digital twins are advanced into an accurate consistency model which accords with the intelligent identification decision of the health state.
Specifically, the method comprises the following steps:
(1) Digital twin initial consistency discrimination
1.1 digital twinning model construction
And constructing a planetary gearbox digital twin by using three-dimensional modeling software UG. Four health states common to the sun gear of a planetary gearbox are identified, namely a worn gear, a pitted gear, a broken gear and a normal gear, as shown in fig. 2 and 3.
1.2 model simulation
Converting a digital twin model of the planetary gearbox into Parasolid (xt) format, introducing ADMAS software, calculating contact force by adopting an Impact function method (Impact), setting the simulation rotating speed to be 790.5r/min, setting the simulation time to be 1s, and setting the simulation step number to be 1500. As shown in fig. 4, a time domain diagram and a frequency domain diagram of the acceleration signal in the healthy state of sun wheel abrasion are obtained through simulation. It is evident from the spectrograms that the higher frequencies are 287.1Hz, 344.6Hz, and 401.7Hz.
1.3 theoretical calculation of meshing frequency and failure characteristic frequency
The calculation formula of the meshing frequency and the fault characteristic frequency of the planetary gearbox is as follows:
sun gear local fault characteristic frequency:
wherein N is the number of the planetary gears; n is s Indicating the rotational speed of the sun gear, Z r And Z s The number of teeth of the ring gear and the sun gear, respectively.
The theoretical output angular speed of the planet carrier is 1709.76 degrees/s and the theoretical meshing frequency f of the sun wheel can be obtained through calculation m =342Hz, the mesh frequency of the whole planetary gear system is 342Hz due to the gears in the planetary gear box meshing with each other, the theoretical rotational frequency f of the sun gear s =18.99Hz, so the theoretical failure characteristic frequency of the sun gear is 56.9Hz.
1.4 actual measurement data acquisition
1.4.1 Experimental Collection System set-up
The mechanical part of the planetary gearbox fault diagnosis test bed mainly comprises a bearing seat, a driving motor, a load magnetic powder brake and a planetary gearbox. The driving motor is a two-phase asynchronous alternating current motor, and the maximum rotating speed of the driving motor is 1500r/min. The acquisition system includes a rotary encoder, a torque sensor, and two three-dimensional accelerometers. The laboratory bench and test system are arranged as shown in fig. 5, the rotary encoder is positioned at the input end of the planetary gear; the torsion sensor is arranged at the output end of the motor; the three-way accelerometer is arranged at the input of the bearing block and the planetary gearbox. In the test system, the parameters of the acquisition card are shown in FIG. 7.
1.4.2 analysis of measured data
The time domain information acquired by the sensor is subjected to Fourier transform, and the impact signal can be obviously seen on a spectrogram. FIG. 6 is a graph of the acceleration signal of the planetary gearbox in the healthy state of sun gear abrasion, which is obtained through experiments, and a frequency domain graph, wherein the higher frequency is 340Hz and 398Hz.
As shown in fig. 7, the errors of the engagement frequency of the simulation data in the worn state, the theoretical engagement frequency, and the engagement frequency of the measured data are 0.8% and 0.9%, respectively. The error of the fault characteristic frequency is 0.7 percent and 0.4 percent respectively with the error of the theoretical fault characteristic frequency and the error of the measured data fault characteristic frequency.
(2) Digital twin value consistency discrimination
2.1 convolutional neural network model construction of attention mechanism
Firstly, a deep convolutional neural network of the attention system learns gear fault characteristics from an original vibration signal, then the attention system can adaptively acquire the importance degree of the characteristics, inhibit invalid characteristic information, and enhance the characteristics of the fault information according to the importance degree, and the network topology structure is shown in fig. 8.
The fault diagnosis model mainly comprises an input layer, a depth feature extraction layer, an attention mechanism layer and a full connection layer. After depth features are extracted through a series of convolution pooling, feature self-adaptive weighting of different signal segments is carried out for information screening by utilizing an attention mechanism technology, screened features are marked again, and then the marked feature sequences are sent to a Soft-Max classifier to realize fault diagnosis of the planetary gearbox.
2.2 data partitioning
The deep learning model needs to learn data features from a large number of samples, so that data enhancement technology processing needs to be performed on the acquired simulation and actual measurement data. Since the number of the samples is not less than the number of the data points acquired by one rotation of the rotating shaft, the length of each sample is 500, and the number of the samples in the four health states shown in fig. 9 is obtained. 1200 samples of the training set are constructed from simulation data of four health states, and 400 samples of the test set are constructed from measured data of four health states.
2.3 model parameter selection
Since the processed data is a one-dimensional vibration signal, using a larger number of convolutional layers may cause a model overfitting problem. Therefore, it is preferable to design a wide convolution kernel in the 1 st convolutional layer on the premise of fixing the 3 convolutional layers to better suppress noise and capture useful information.
2.4 diagnostic output
The divided data are input into the constructed one-dimensional convolutional neural network model for adaptive learning, and the accuracy curves of the training set and the test set obtained according to the diagnostic process are shown in fig. 10. It can be seen from the figure that the training accuracy of the model reaches 96.3%, no overfitting phenomenon exists, and the test accuracy can reach 80.2%.
FIG. 11 is a graph showing the feature visualization by mapping the high-dimensional feature to a two-dimensional space and mapping the output vector of the one-dimensional convolutional neural network to a 2-dimensional space by using a t-SNE dimension reduction algorithm.
Under the engineering background that fault data is insufficient or fault sample data cannot be obtained on site, only simulation data is used for training a one-dimensional convolutional neural network model, actual measurement data is used as a test set, the fault diagnosis accuracy rate is 80.2%, and 4 health types have obvious identification degrees as can be seen from fig. 11.
(3) Digital twin trend consistency discrimination
In order to further verify that the virtual and actual data have consistency, the measured data of the same health state is added into the simulation data of each type of health state to construct mixed data to be used as a new training set, the number of samples in the training set is still 1200, and the measured data is used as a test set and is unchanged. The training set composition for the three experiments is shown in figure 12,
experiment 1: the 1200 training set samples are formed by mixing 900 simulation data samples and 300 measured data samples, namely the ratio of the simulation data to the measured data in the training set is 3: 1.
Experiment 2:1200 training set samples are formed by mixing 600 simulation data and 600 measured data samples, namely the ratio of the simulation data to the measured data in the training set is 1: 1.
Experiment 3: the 1200 training set samples are formed by mixing 300 simulation data and 900 measured data samples, namely the ratio of the simulation data to the measured data in the training set is 1: 3.
The experimental results are as follows: the failure diagnosis accuracy of experiment 1 is 84.3%, the failure diagnosis accuracy of experiment 2 is 90.5%, and the failure diagnosis accuracy of experiment 3 is 92.6%, and the results are shown in fig. 12. It can be seen from the figure that the diagnosis accuracy can be slowly improved along with the increase of the actually measured data in the training set, and it can be judged that the constructed digital twin has not reached the accurate consistency.
(4) Digital twin engineering consistency discrimination
In order to verify the engineering value of the method for testing the fault, which is insufficient or can not be obtained, the one-dimensional convolutional neural network model is independently trained by only adopting the measured data in the mixed data, and the test set is unchanged. It can be seen from fig. 13 that the training accuracy rates of the three schemes are 95.4%, 96.4% and 97.3%, respectively, which indicates that the training accuracy rate is slowly increased with the increase of the measured data in the training set. In addition, when the number of the measured data samples in the training set is 300 and 600, the fault diagnosis accuracy is 69.7% and 78.1%, respectively, and the diagnosis result of the one-dimensional neural network model trained by the simulation data cannot reach 80.2%, and the diagnosis result of the one-dimensional neural network trained by the mixed data consisting of the measured data with the same number cannot reach 84.3% and 90.5%. Comparing fig. 13 and fig. 12 at the same time, it can be seen that the advantages of the present application are more prominent when there are fewer measured data samples in the training set.
(5) Accurate consistency discrimination
5.1 "six-dimensional Structure" digital twin evolution mechanism
On the basis of the existing five-dimensional structure, a six-dimensional structure of physical space data, digital space simulation data, human factors overall, twin data, an information channel and diagnosis and prediction services is built, real-time accurate mapping and synchronous evolution of a digital space and a physical entity are achieved, and finally the essential characteristic of digital twin visualization-simulation-feedback for accurate identification and decision of service health states is achieved.
5.2 virtual-real space real-time interaction and feedback control with cloud edge coordination
Based on the edge computing and cloud edge cooperative scheduling as a core and facing to an edge side data real-time processing technology, data for real-time analysis are transmitted to edge equipment nearest to a data source through a network and are stored locally, and delay insensitive data are directly transmitted to a cloud center; the self-adaptive optimization scheduling of cloud-side data is realized based on an intelligent algorithm, and digital twin high-fidelity behavior simulation and real-time state visualization driven by cloud-side data are realized; and then the physical space is fed back and controlled, and real-time accurate mapping and synchronous evolution of the virtual space and the real space are realized.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not necessarily depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. A digital twin consistency discrimination method for service health state intelligent recognition decision is characterized in that the method is formed based on an adjustment training set so as to drive the calculation of health state recognition accuracy, and the method mainly comprises the following steps:
(1) Judging the initial consistency of the digital twins: constructing a digital twin of a physical object in modeling software and carrying out dynamic simulation to obtain simulation data in a normal state, a fault state and different health states; comparing simulation data and actual measurement data under different health states, and theoretically calculating frequency and fault characteristic frequency to verify the initial consistency of the virtual space and the real space;
(2) And (3) judging the consistency of the digital twin values: driving a deep convolutional neural network model by simulation data, and calculating the health state identification accuracy; adopting simulation data of different health states as a training set and actual measurement data as a test set, calculating the accuracy of fault diagnosis, and verifying the value consistency of virtual and real spaces in the aspect of intelligent identification decision of service health states;
(3) Judging the consistency of the digital twin trend: constructing a training set of simulation data and measured data, training a deep convolutional neural network model, keeping a test set unchanged, calculating the recognition accuracy of the health state, and verifying the trend consistency of virtual and real space data when the simulation data in the mixed data occupy different proportions;
(4) Judging the consistency of digital twin engineering: respectively constructing training sets by using the measured data of the mixed data, training a deep convolutional neural network model, keeping the testing sets unchanged, and calculating the recognition accuracy of the health state; and when the number of the actually measured data in the training set accounts for different proportions, acquiring the accuracy of fault diagnosis and verifying the engineering consistency of the digital twin.
2. The method for judging the digital twin consistency of service health state intelligent recognition decision as claimed in claim 1, further comprising the step (5) of judging the digital twin accurate consistency of service decision: and (3) iteratively optimizing digital twins through virtual-real space interactive feedback, repeating the steps (1), (2), (3) and (4), and respectively observing the accuracy of health state identification until the digital twins are evolved into an accurate consistency model which accords with the intelligent health state identification decision.
3. The method as claimed in claim 1, wherein the physical object is a planetary gearbox.
4. The method of claim 1, wherein the physical object is a planetary gearbox sun gear.
5. The method as claimed in claim 4, wherein four health states common to sun gears of planetary gear boxes are determined, namely worn gears, pitting gears, broken gears and normal gears.
6. The method as claimed in claim 5, wherein when the physical object is a planetary gearbox, the following steps are executed:
(1) Judging the initial consistency of the digital twins: the method comprises the steps that a planetary gearbox digital twin is built in CAD software, and dynamic simulation is carried out through ADMAS to obtain simulation data under four different health states, namely a normal state, a wear fault, a tooth breakage fault and a pitting corrosion fault; verifying simulation data and actual measurement data under four health states, and meshing frequency and fault characteristic frequency of theoretical calculation;
(2) Judging the consistency of the digital twin values: driving a deep convolutional neural network model by simulation data, and calculating the health state identification accuracy; adopting simulation data of different health states as a training set and actual measurement data as a test set, calculating the accuracy of fault diagnosis, and verifying the value consistency of virtual and real spaces in the aspect of intelligent identification decision of service health states;
(3) Judging the consistency of the digital twin trend: constructing a training set of simulation data and actual measurement data, training a deep convolution neural network model, keeping the test set unchanged, and calculating the recognition accuracy of the health state; when simulation data in the mixed data account for different proportions, the accuracy rate of fault diagnosis is calculated, and the consistency of the trend of the virtual space data and the real space data is verified;
(4) Judging the consistency of the digital twin engineering: and respectively constructing a training set by using the measured data of the mixed data, training a deep convolutional neural network model, keeping the testing set unchanged, and calculating the recognition accuracy of the health state. When the number of the measured data in the training set is 300 groups and 600 groups respectively, the accuracy of fault diagnosis is calculated, and the project consistency of digital twins is verified in the actual scene that the data is insufficient in fault diagnosis of the rotary machine and the site is difficult to obtain.
7. The method according to claim 6, wherein in the step (2), simulation data of different health states are used as a training set, actual measurement data are used as a test set, 300 groups of four health states are used as the training set, 1200 groups of simulation data are used as the training set, 100 groups of simulation data are used, and 400 groups of actual measurement data are used as the test set.
8. The method as claimed in claim 6, wherein in the step (3), the simulation data in the mixed data respectively account for 100%, 75%, 50% and 25%.
9. The method as claimed in claim 6, further comprising (5) a digital twin positive consistency determination of the service decision: and (3) iteratively optimizing digital twins through virtual-real space interactive feedback, repeating the steps (1), (2), (3) and (4), and respectively observing the accuracy of health state identification until the digital twins are evolved into an accurate consistency model which accords with the intelligent health state identification decision.
10. The method for judging the digital twin consistency of the service health state intelligent recognition decision as claimed in claim 9, wherein the virtual-real space interactive feedback is based on a real-time processing technology facing to the edge side data with edge computing and cloud edge cooperative scheduling as a core, so that data for real-time analysis is transmitted to the nearest edge device close to a data source through a network and is locally stored, and time delay insensitive data is directly transmitted to a cloud center; the method comprises the following steps of realizing self-adaptive optimization scheduling of cloud-edge data based on an intelligent algorithm, and realizing digital twin high-fidelity behavior simulation and real-time state visualization driven by cloud-end data; and then the physical space is fed back and controlled, and real-time accurate mapping and synchronous evolution of the virtual space and the real space are realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211136899.0A CN115421400B (en) | 2022-09-16 | Digital twin virtual-real consistency judging method for intelligent service health state identification decision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211136899.0A CN115421400B (en) | 2022-09-16 | Digital twin virtual-real consistency judging method for intelligent service health state identification decision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115421400A true CN115421400A (en) | 2022-12-02 |
| CN115421400B CN115421400B (en) | 2024-09-24 |
Family
ID=
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116050190A (en) * | 2023-03-31 | 2023-05-02 | 中国海洋大学 | Product performance and degradation state analysis method based on digital twinning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110851966A (en) * | 2019-10-30 | 2020-02-28 | 同济大学 | Digital twin model correction method based on deep neural network |
| CN112417619A (en) * | 2020-11-23 | 2021-02-26 | 江苏大学 | Pump unit optimal operation adjusting system and method based on digital twinning |
| CN113283095A (en) * | 2021-05-31 | 2021-08-20 | 中国水利水电科学研究院 | Evolutionary digital twin watershed construction method |
| CN114611313A (en) * | 2022-03-21 | 2022-06-10 | 西南交通大学 | Complex product digital twinning construction and application method based on model fusion |
| CN114894507A (en) * | 2022-06-08 | 2022-08-12 | 天津职业技术师范大学(中国职业培训指导教师进修中心) | Fault diagnosis and health state prediction method for digital twin-driven rail vehicle transmission system |
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110851966A (en) * | 2019-10-30 | 2020-02-28 | 同济大学 | Digital twin model correction method based on deep neural network |
| CN112417619A (en) * | 2020-11-23 | 2021-02-26 | 江苏大学 | Pump unit optimal operation adjusting system and method based on digital twinning |
| CN113283095A (en) * | 2021-05-31 | 2021-08-20 | 中国水利水电科学研究院 | Evolutionary digital twin watershed construction method |
| CN114611313A (en) * | 2022-03-21 | 2022-06-10 | 西南交通大学 | Complex product digital twinning construction and application method based on model fusion |
| CN114894507A (en) * | 2022-06-08 | 2022-08-12 | 天津职业技术师范大学(中国职业培训指导教师进修中心) | Fault diagnosis and health state prediction method for digital twin-driven rail vehicle transmission system |
Non-Patent Citations (1)
| Title |
|---|
| 陶飞;刘蔚然;刘检华;刘晓军;刘强;屈挺;胡天亮;张执南;向峰;徐文君;王军强;张映锋;刘振宇;李浩;程江峰;戚庆林;张萌;张贺;隋芳媛;何立荣;易旺民;程辉;: "数字孪生及其应用探索", 计算机集成制造系统, no. 01, 15 January 2018 (2018-01-15) * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116050190A (en) * | 2023-03-31 | 2023-05-02 | 中国海洋大学 | Product performance and degradation state analysis method based on digital twinning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jiao et al. | Deep coupled dense convolutional network with complementary data for intelligent fault diagnosis | |
| CN109918752B (en) | Mechanical fault diagnosis method, equipment and medium based on migration convolution neural network | |
| CN107702922B (en) | Rolling bearing fault diagnosis method based on LCD and stacking automatic encoder | |
| CN113255078A (en) | Bearing fault detection method and device under unbalanced sample condition | |
| CN109211546B (en) | Rotary machine fault diagnosis method based on noise reduction automatic encoder and increment learning | |
| CN109186964B (en) | rotary machine fault diagnosis method based on angle resampling and ROC-SVM | |
| Gecgel et al. | Gearbox fault diagnostics using deep learning with simulated data | |
| CN108106841A (en) | Epicyclic gearbox intelligent failure diagnosis method based on built-in encoder signal | |
| CN112665852B (en) | Variable working condition planetary gearbox fault diagnosis method and device based on deep learning | |
| CN116956048B (en) | Industrial equipment fault diagnosis method and device based on cross-domain generalized label | |
| CN109060347B (en) | Planetary gear fault identification method based on stacking noise-eliminating automatic encoder and gated cyclic unit neural network | |
| CN111122161B (en) | Variable working condition bearing fault diagnosis method based on fast kurtogram and deep residual learning | |
| Jie et al. | Spectral kurtosis based on evolutionary digital filter in the application of rolling element bearing fault diagnosis | |
| CN115563853A (en) | Rolling bearing fault diagnosis method based on digital twinning | |
| CN114970628B (en) | Rotating part fault diagnosis method based on generation countermeasure network under condition of unbalanced samples | |
| CN112733612A (en) | Cross-domain rotating machinery fault diagnosis model establishing method and application thereof | |
| CN116223020A (en) | Intelligent fault diagnosis method for gearbox based on Li-MDCAN | |
| CN115683604A (en) | Engineering machinery fault diagnosis method based on unbalanced data | |
| CN115481666A (en) | Fault diagnosis method, system and equipment for small sample of gear box | |
| CN117828531A (en) | Bearing fault diagnosis method based on multi-sensor multi-scale feature fusion | |
| CN115421400B (en) | Digital twin virtual-real consistency judging method for intelligent service health state identification decision | |
| CN115421400A (en) | Digital twin virtual-real consistency discrimination method for service health state intelligent recognition decision | |
| CN112446326B (en) | Canned motor pump fault mode identification method and system based on deep rewinding and accumulating network | |
| CN117606798A (en) | Tobacco machinery bearing fault diagnosis method and diagnosis system | |
| CN116720442A (en) | Wear prediction method based on improved Fourier neural operator |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |