CN116050190B - Product performance and degradation state analysis method based on digital twinning - Google Patents
Product performance and degradation state analysis method based on digital twinning Download PDFInfo
- Publication number
- CN116050190B CN116050190B CN202310331163.7A CN202310331163A CN116050190B CN 116050190 B CN116050190 B CN 116050190B CN 202310331163 A CN202310331163 A CN 202310331163A CN 116050190 B CN116050190 B CN 116050190B
- Authority
- CN
- China
- Prior art keywords
- layer
- data
- product
- digital twin
- digital
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/04—Ageing analysis or optimisation against ageing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention belongs to the technical field of intelligent and digital information, and particularly discloses a product performance and degradation state analysis method based on digital twinning. The digital twin system has the advantages that the digital twin system can fully mine the implicit design requirements related to the use environment or working condition, the performance and efficiency of the electromechanical product and the like, simultaneously avoids the problem of inaccurate subjective data and model caused by manual analysis in the traditional optimization design, and has important significance for updating and competitive force improvement of the product.
Description
Technical Field
The invention belongs to the technical field of intelligent and digital information, and particularly discloses a product performance and degradation state analysis method based on digital twinning.
Background
With the intervention of informatization and the rapid development of products, data collected over time in monitoring the operational status of the product can provide valuable information about the current and historical status of the asset. This evolution of the machine can be used to predict how the asset will perform over time and how it may degrade, allowing maintenance to be scheduled based on these predictions.
In the prior art, complex operation environments and working conditions of electromechanical products are difficult to truly and completely reflect, so that a certain deviation exists between input of corresponding design information and actual conditions, and the digital twin simulation operation result and product actual monitoring data have errors of different degrees and cannot be completely consistent with an optimal design model.
Disclosure of Invention
Based on the problems, the application provides a digital twinning-based product performance and degradation state analysis method capable of reducing errors of operation results and product real monitoring data. The technical proposal is that,
a product performance and degradation state analysis method based on digital twin comprises a physical entity layer, a complete machine disassembly layer, a digital twin layer, a performance monitoring layer, a prediction integration module and an application service layer; the physical entity layer comprises a complex electromechanical product and a data acquisition device; the whole machine disassembly and layering is to disassemble the whole machine of the product into a plurality of key parts and map the subsequent digital twin modules; the digital twin layer comprises a data processing module and a twin module, wherein the data processing module comprises reading of acquired data, preprocessing of the data, extraction of characteristic values and data analysis; the data acquisition device acquires the multi-element data type of the product, processes the multi-element data type through the data processing layer and then applies the multi-element data type to the twin model layer; the performance monitoring module adjusts design parameters by adopting a digital twin correction method, and evaluates the influence of the monitoring parameters on the product and the influence of the digital twin prediction method on the performance of the whole machine product by predicting the design parameters and working conditions; the prediction integration module performs weighted integration on the parts of the whole machine and outputs a prediction result to the service control module; the service control module comprises docking with clients, demand analysis, and electromechanical product design process of different stages of external packaging and scheme design and detailed design of products, and is a process of interaction and feedback of digital twin module design adjustment information.
Preferably, the complete machine disassembly layer maps the functions and the service performance of the complete machine to the action layer through disassembly according to a meta action theory, and researches and controls meta actions and meta action units from the action layer, disassembles complex products of the complete machine into single parts which are easy to establish a digital model according to a power transmission route, and the method comprises the following specific steps,
s101, analyzing a power transmission route of a product;
s102, grading a transmission route;
selecting an FMA decomposition method, and dividing a product into three layers, namely a functional layer, a motion layer and an action layer from the function of the whole machine; the functional layer is used for analyzing the specific function of the whole machine; the motion layer is used for analyzing the motion form of the whole machine part for realizing the sub-function; the action layer decomposes the element action form for realizing the movement according to the transmission path of the movement and the power; the constituent elements of the action layer at the bottommost layer are meta actions;
s103, classifying according to the action of the meta action on the action layer.
Preferably, the data reading in the data processing module is to summarize and classify the signals collected by the sensor; the data preprocessing is to reduce or eliminate interference components in the acquired data for the read sensor signals by adopting a five-point three-time smoothing method, and to smooth the data by adopting a weighted moving average method.
Preferably, the digital twin layer processing steps are as follows:
s201, inquiring parameters and corresponding working conditions of each part according to the parts split by the product splitting module, and acquiring multiple data types through real-time monitoring of sensors;
s202, transmitting the multi-element data type to a data processing layer, preprocessing acquired data by the data processing layer to remove interference, and extracting time-domain and frequency-domain characteristic values of the processed signals;
s203, constructing a digital twin model according to the finite element model,
wherein the method comprises the steps of
Representative ofDigital twin architecture->
Representing physical layer, & gt>
Representing a digital twin model->
Representing digital twin data including physical entity operation data and digital twin model generation data,/->
Related services representing digital twinning +.>
Represents->
、/>
、/>
、/>
Connection interactions between them.
Preferably, the step S203 of establishing a digital twin model is trained, and the actual vibration signal is used for correcting the theoretical vibration signal, so that consistency between the physical entity and the digital twin model is maintained;
training and predicting theoretical data through real data by adopting a convolutional neural network, wherein the method comprises the following steps of:
convolution layer input:
；
convolution layer output:
；
wherein conv2 () is a function of a convolution operation, W is a convolution kernel matrix, which is essentially a filter, performing element-by-element multiplication and addition; x is an input matrix, input is operation data of the split parts, b represents offset,
is an activation function;
and calculating the total error of the connecting layer according to the deviation of the input value and the output value, wherein the calculation formula is as follows:
where L represents the number of layers of the output layer, d represents the vector desired to be output, y represents the vector output by the convolutional neural network,
representative vector (+)>
) The specific calculation formula is as follows:
Preferably, the corrected digital twin model is divided into full life cycles, and the specific operation is as follows:
s301, performing simulation operation on the modified part twin model, generating theoretical operation data, preprocessing the theoretical operation data,
s302, adding kurtosis operation, setting thresholds of different stages, and dividing the performance of the part product into a normal stage and a slight damage stage and a serious damage stage through comparing the kurtosis with the set thresholds;
the juncture of the normal working stage and the slight damage stage is defined as a first monitoring point FPT; the juncture point between the slight damage stage and the next stage is needed to be noted, namely a start monitoring point FOT, the selection of the point is required to require that the vibration signal exceeds an alarm threshold value for m times continuously, and the meeting point at the moment is set as a performance degradation start monitoring point; important monitoring is required for severe injury phases and concerns about failure points.
Preferably, the life distribution of the product is set as
On the premise of no failure of the part, < +.>
The remaining lifetime at the moment satisfies the following formula:
and c is the used time during detection, T is the total life time of the part, P is a part life calculation function, a bi-directional long-short-time memory network BILSTM is adopted, and corrected part digital twin model operation data are input into a built BILSTM network to obtain a life prediction result.
Preferably, the score of each part is calculated according to the Scoring function,
wherein the method comprises the steps of
Error for the percentage of the ith sample, +.>
For the actual life time of the component>
For the predicted lifetime of the component, +.>
RUL estimation accuracy transfer function representing part i, < ->
Representing the final score for the part.
Preferably, in the prediction integration module, the analytic hierarchy process, the expert evaluation method and the weighted analysis method or the fusion of the three methods are adopted to carry out weighted integration on the parts of the whole machine, so as to realize the performance evaluation of the product; the parts of the electromechanical product of the whole machine are divided into core parts, basic parts and easily damaged auxiliary parts; the weight of the core part is 0.4-0.7; the specific gravity of the vulnerable part is 0-0.1; the weight of the basic part is 0.1-0.4, and the sum of the highest upper limits of the weights of the three is not more than 1 during calculation.
Compared with the prior art, the application has the following advantages:
1. the method comprises the steps of establishing a comprehensive product digital twin model, mainly applying modeling knowledge such as mathematical formula modeling, finite element theory and computational fluid mechanics according to the characteristics of parts in different element action fields, selecting a geometric model, a finite element model, a dynamic model or a mixed modeling form of three models, and establishing a complete product coupling equation, so that the coupling of multiple physical fields of the digital twin model is realized.
2. Firstly, modeling data of a digital twin model are all derived from real physical entity data; secondly, correcting parameters of the digital twin model which is built preliminarily by using actual monitoring data as a reference object; finally, online parameter adjustment is carried out on the split part digital twin model after the previous correction according to the product real-time monitoring operation data, so that the real-time updating property of the digital twin model is ensured.
3. Aiming at the problems of information interaction and feedback of a service control layer and a product performance monitoring and service life predicting module and a physical system and signal acquisition system layer, the information feedback problem of a product monitoring and service life predicting stage is solved, firstly, the predicting precision of the parts is subjected to separate precision evaluation, and the mutual noninterference between part predictions is ensured. And then, the calculated part prediction precision is subjected to weighted evaluation, so that a more objective and accurate monitoring result is obtained. And the monitoring result is input into a service control layer system of the product, customer information is acquired and interacted with the product, and the interacted information is fed back to a physical entity, so that the perfection of product performance monitoring is realized.
4. The digital twin model of the electromechanical product split established based on the digital twin product performance and degradation state analysis method can be reused, and repeated domain knowledge analysis is avoided. The designer reduces the difference in concept and term by constructing a unified framework or a normative model, and ensures the uniformity and rapidity of data and information interaction in the design process of performance monitoring and life prediction.
Drawings
FIG. 1 is an overall framework of the digital twin product performance and degradation state analysis method of the present invention.
FIG. 2 is an exploded view of the truck drive train elements.
FIG. 3 is a block diagram of a CNN-BILSTM cell.
Fig. 4 is a basic frame of the fusion of the evaluation accuracy of the parts.
Detailed Description
The following detailed description, which is illustrative in nature and is intended to provide further explanation of the present application, will be clearly and fully described with reference to the accompanying drawings in which embodiments of the invention are shown. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
FIG. 1 shows a digital twinning-based product performance and degradation state analysis method, which comprises a physical entity layer, a complete machine disassembly layer, a digital twinning layer, a performance monitoring layer, a prediction integration module and an application service layer;
the physical entity layer comprises a complex electromechanical product and a data acquisition device;
the data acquisition device mainly comprises various sensors, wherein the types of the sensors mainly comprise vibration, acceleration, speed and other types of sensors;
the whole machine disassembly and layering is to disassemble the whole machine of the product into a plurality of key parts and map the subsequent digital twin modules;
the digital twin layer comprises a data processing module and a twin module, wherein the data processing module comprises reading of acquired data, preprocessing of the data, extraction of characteristic values and data analysis;
the data acquisition device acquires the multi-element data type of the product, processes the multi-element data type through the data processing layer and then applies the multi-element data type to the twin model layer;
the performance monitoring module adjusts design parameters by adopting a digital twin correction method, and evaluates the influence of the monitoring parameters on the product and the influence of the digital twin prediction method on the performance of the whole machine product by predicting the design parameters and working conditions;
the prediction integration module performs weighted integration on the parts of the whole machine and outputs a prediction result to the service control module;
the service control module comprises docking with clients, demand analysis, and electromechanical product design process of different stages of external packaging and scheme design and detailed design of products, and is a process of interaction and feedback of digital twin module design adjustment information.
Taking a bogie transmission system as a typical complex electromechanical product as an example, a digital twin module is established. The multiple data types of the product first need to be processed. The method specifically comprises the steps of firstly arranging a plurality of data acquisition devices to perform multi-element acquisition on detailed information of products, preprocessing acquired data, inputting the processed data into a convolutional neural network as test data, and training the training data.
Step 1, data preprocessing: after the data type and the attribute of the acquired data of the transmission system of the bogie mechanism are determined, the data processing module is used for preprocessing the vibration signals, and a digital twin module is established. Firstly, data preprocessing is needed, a five-point three-time smoothing method is adopted to reduce or eliminate interference components in the acquired data for the read sensor signals, and a weighted sliding average method is adopted to carry out smoothing on the data level, so that the acquired data is attached to the actual working condition of the product.
The analysis steps of the whole machine splitting of the product in the step 2 are as follows:
step 101, analyzing a bogie transmission route: the bogie transmits gravity, traction and braking forces and also transmits transverse forces when passing through curves. The analysis is made here by taking gravity as an example, the transmission route of the traction force is as follows:
the weight of the upper part of the vehicle body, a secondary spring suspension device, a bogie frame, an axle box spring suspension device, a wheel set and steel rails.
Step 102, grading the transmission route: according to the element action theory, selecting the FMA decomposition method, the bogie can be divided into three layers from top to bottom, namely a functional layer (F), a motion layer (M) and an action layer (A), wherein the constituent elements of the action layer at the bottom layer are element actions. The action of the element action on the action layer can be divided into five major categories, namely an input piece, a middle piece, an output piece, a fastener and a supporting piece.
The motor and other elements providing power are input parts, the bearings, rotating shafts, keys and other elements transmitting motion and power and guaranteeing the positions of the input parts and the output parts are middle parts, the motion executing parts of the equivalent unit actions of the wheels are output parts, the bogie frame and other elements providing assembly references for other parts are supporting parts, the bolts, the bearing end covers and the like are connected with two or more parts, and the elements which do not move relatively or play a role in sealing are fastening parts, as shown in fig. 2.
The construction steps of the digital twin model of the part are as follows:
201, inquiring parameters and corresponding working conditions of each part according to the parts split in the step 2, and acquiring multiple data types through real-time monitoring of sensors;
s202, transmitting the multi-element data type to a data processing layer, preprocessing acquired data by the data processing layer to remove interference, and extracting time-domain and frequency-domain characteristic values of the processed signals;
s203, a mathematical formula is selected according to the characteristics of each part to establish a theoretical model, a finite element model and a solidworks to establish a physical entity model, wherein the specific digital twin framework is as follows:
wherein the method comprises the steps of
Represents a digital twin architecture consisting of +.>
These five parts constitute->
Representing the physical entity of the bogie transmission of a motor vehicle,/-or->
Representing a digital twin model, mainly comprising a theoretical model, a finite element model and a three-dimensional model,/->
Representing digital twin data, mainly comprising physical entity operation data and digital twin model generation data,/->
Related services representing digital twinning (prior art, digital Twin Service towards Smart Manufacturing), a->
Represents->
、/>
、/>
、/>
Connection interactions between them.
The digital twin model training comprises the steps of correcting theoretical vibration signals by using actual vibration signals for theoretical digital twin established in the step S203, so that consistency between a physical entity and the digital twin model is maintained;
and (3) data characteristic extraction: firstly, extracting frequency domain characteristic values when the data preprocessed in the step one is subjected to the preprocessing, wherein the characteristic values mainly comprise: peak-to-peak value, mean value, variance, standard deviation, mean square value, root mean square value, kurtosis, skewness, peak factor, pulse factor, margin factor; and extracting the same characteristic value from the digital twin model.
Model parameter correction: training the digital twin model according to the error magnitude, and training and predicting theoretical data through real data by adopting a convolutional neural network as shown in fig. 3. The convolutional neural network training process is as follows:
convolution layer input:
；
convolution layer output:
；
wherein conv2 () is a function of convolution operation in Matlab, W is a convolution kernel matrix, X is an input matrix, the input is operation data of each part, b represents an offset,
to activate the function.
And calculating the total error of the connecting layer according to the deviation of the input value and the output value, wherein the calculation formula is as follows:
where L represents the number of layers of the output layer, d represents the vector of the desired output, y represents the vector of the convolutional neural network output,
representative vector (+)>
) The specific calculation formula is as follows:
And 4, monitoring the performance of the corrected part digital twin model, wherein the flow is as follows:
s301, performing simulation operation on the modified part twin model, generating theoretical operation data, preprocessing the theoretical operation data, extracting actual characteristic values, and modifying the kurtosis by using the actual data because the kurtosis easily reflects the impact characteristic of a vibration signal and takes the kurtosis as a prediction accuracy parameter, wherein the kurtosis expression is as follows:
wherein KU represents a kurtosis,
represents the amplitude of vibration waveform, Q represents the number of vibration waveform, mu represents average value, and the calculation formula is +.>
The method comprises the steps of carrying out a first treatment on the surface of the rms is root mean square value, and the calculation formula is +.>
。
301. Health stage division: and (3) performing simulation operation on the modified part twin model, generating theoretical operation data, and dividing the theoretical operation into a normal stage and a slight damage stage according to the kurtosis, wherein the normal operation is mainly divided into a severe damage stage. Taking a bearing as an example, the kurtosis value of a normal bearing is about 3, if the kurtosis value exceeds 3, the bearing is a fault bearing, and then nodes are changed according to the numerical value to divide the degradation stage until the final failure.
302. In order to reduce the waste of resources, performance monitoring is not needed in the normal working stage of the parts; the juncture of the normal working stage and the slight damage stage is defined as a first monitoring point FPT; the juncture point between the slight damage stage and the next stage is needed to be noted, namely a start monitoring point FOT, the selection of the point is required to require that the vibration signal exceeds an alarm threshold value for m times continuously, and the meeting point at the moment is set as a performance degradation start monitoring point; important monitoring is required for severe injury phases and concerns about failure points. Performance monitoring is primarily performed between FOT and failure points, as performance degradation processes of the product are of major concern.
Step 5, life prediction: predicting the residual service life of the part according to the divided health stages, and setting the service life distribution of the part as D #
) On the premise of no failure of the part, < +.>
The remaining lifetime of the moment of time mainly satisfies the following formula.
The method comprises the steps of c is the used time during detection, T is the total service life time of a part, P is a part service life calculation function, the design adopts a bi-directional long-short-time memory network BILSTM, and corrected part digital twin model operation data are input into a built BILSTM network to obtain a service life prediction result of the corresponding part.
And 6, calculating prediction precision: the evaluation index of the life prediction accuracy mainly includes MAE, RMSE, MAPE and a scaling function. The Scoring function is calculated as follows:
wherein the method comprises the steps of
Error for the percentage of the ith sample, +.>
For the actual life time of the component>
For the predicted lifetime of the component N represents +.>
Number of (I) and (II)>
RUL estimation accuracy transfer function representing part i, < ->
Representing the final score for the part.
As shown in fig. 4, in the prediction integration module in step 7, the analytic hierarchy process, the expert evaluation method and the weighted analysis method or the fusion of the three methods are adopted to perform weighted integration on the parts of the whole machine, so as to realize the product performance evaluation; the parts of the electromechanical product of the whole machine are divided into core parts, basic parts and easily damaged auxiliary parts; the weight of the core part is 0.4-0.7; the specific gravity of the vulnerable part is 0-0.1; the weight of the basic part is 0.1-0.4, and the sum of the highest upper limits of the weights of the three is not more than 1 during calculation.
Claims (5)
1. The product performance and degradation state analysis method based on digital twin is characterized by comprising a physical entity layer, a complete machine disassembly layer, a digital twin layer, a performance monitoring layer, a prediction integration module and an application service layer;
the physical entity layer comprises a complex electromechanical product and a data acquisition device;
the whole machine disassembly and layering is to disassemble the whole machine of the product into a plurality of key parts and map the subsequent digital twin modules; the digital twin layer comprises a data processing module and a twin module, wherein the data processing module comprises reading of acquired data, preprocessing of the data, extraction of characteristic values and data analysis; the parts of the electromechanical product of the whole machine are divided into core parts, basic parts and easily damaged auxiliary parts; the weight of the core part is 0.4-0.7; the specific gravity of the vulnerable part is 0-0.1; the weight of the basic part is 0.1-0.4, and the sum of the highest upper limits of the weights of the three is not more than 1 during calculation;
the data acquisition device acquires the multi-element data type of the product, processes the multi-element data type through the data processing layer and then applies the multi-element data type to the twin model layer;
the performance monitoring module adjusts design parameters by adopting a digital twin correction method, and evaluates the influence of the monitoring parameters on the product and the influence of the digital twin prediction method on the performance of the whole machine product by predicting the design parameters and working conditions;
dividing the full life cycle of the corrected digital twin model, and specifically performing the following operations:
s301, performing simulation operation on the modified part twin model, generating theoretical operation data, preprocessing the theoretical operation data,
s302, adding kurtosis operation, setting thresholds of different stages, and dividing the performance of the part product into a normal stage and a slight damage stage and a serious damage stage through comparing the kurtosis with the set thresholds;
the juncture of the normal working stage and the slight damage stage is defined as a first monitoring point FPT; the juncture point between the slight damage stage and the next stage is needed to be noted, namely a start monitoring point FOT, the selection of the point is required to require that the vibration signal exceeds an alarm threshold value for m times continuously, and the meeting point at the moment is set as a performance degradation start monitoring point; important monitoring is needed for the severe injury stage, and failure points are concerned;
predicting the remaining service life of the part according to the divided health stages, and setting the service life distribution of the part as D (t h ) T is the sum of the values of the parts under the precondition of no failure h The remaining lifetime of a moment mainly satisfies the following formula:
wherein c is the used time during detection, T is the total life time of the part, P is the life calculation function of the part, a bi-directional long-short-time memory network BILSTM is adopted, and corrected running data of the digital twin model of the part is input into the built BILSTM network to obtain the life prediction result of the corresponding part;
the prediction integration module performs weighted integration on the parts of the whole machine and outputs a prediction result to the service control module;
the service control module comprises docking with a customer, demand analysis, and electromechanical product design processes at different stages of outward packaging, scheme design and detailed design of the product, and is a process of interaction and feedback of design adjustment information of the digital twin module;
the whole machine disassembly and layering is to map the functions and the service performance of the whole machine to an action layer through disassembly according to a meta action theory, study and control meta actions and meta action units by starting from the action layer, disassemble a complex product of the whole machine into single parts which are easy to establish a digital model according to a power transmission route, and concretely comprises the following steps of,
s101, analyzing a power transmission route of a product;
s102, grading a transmission route;
selecting an FMA decomposition method, and dividing a product into three layers, namely a functional layer, a motion layer and an action layer from the function of the whole machine;
the functional layer is used for analyzing the specific function of the whole machine; the motion layer is used for analyzing the motion form of the whole machine part for realizing the sub-function; the action layer decomposes the element action form for realizing the movement according to the transmission path of the movement and the power; the constituent elements of the action layer at the bottommost layer are meta actions;
s103, classifying according to the action of the meta action on the action layer;
the digital twin layer processing steps are as follows,
s201, inquiring parameters and corresponding working conditions of each part according to the parts split by the product splitting module, and acquiring multiple data types through real-time monitoring of sensors;
s202, transmitting the multi-element data type to a data processing layer, preprocessing acquired data by the data processing layer to remove interference, and extracting time-domain and frequency-domain characteristic values of the processed signals;
s203, constructing a digital twin model according to the finite element model,
M DT ＝(E P ,M T ,D T ,S T ,C I )
wherein M is DT Representing a digital twin architecture, E P Representing physical entity layer, M T Representing a digital twin model, D T Representing digital twin data including physical entity operation data and digital twin model generation data, S T Representing a digital twinned related service, C I Represents E P 、M T 、D T 、S T Connection interactions between them.
2. The digital twinning-based product performance and degradation state analysis method according to claim 1, wherein the data reading in the data processing module is to perform summary classification on signals collected by the sensors; the data preprocessing is to reduce or eliminate interference components in the acquired data for the read sensor signals by adopting a five-point three-time smoothing method, and to smooth the data by adopting a weighted moving average method.
3. The method for analyzing product performance and degradation state based on digital twinning according to claim 1, wherein the step S203 is to build a digital twinning model for training, and the theoretical vibration signal is corrected by the actual vibration signal, so that consistency between the physical entity and the digital twinning model is maintained;
training and predicting theoretical data by adopting a convolutional neural network through real data, wherein the steps are as follows,
convolution layer input: v=conv2 (W, X, "valid") +b
Convolution layer output: y=ψ (V)
Wherein conv2 () is a function of a convolution operation, W is a convolution kernel matrix, which is essentially a filter, performing element-by-element multiplication and addition; x is an input matrix, input is operation data of the split parts, b represents offset, and psi (V) is an activation function;
and calculating the total error of the connecting layer according to the deviation of the input value and the output value, wherein the calculation formula is as follows:
where d represents the vector of the desired output, y represents the vector of the convolutional neural network output, L represents the number of layers of the output layer, |d-y L ‖ 2 Representative vector (d-y) L ) The specific calculation formula is as follows:
x i representing any numerical value in X, extracting the characteristics of the predicted data, carrying out error calculation between the predicted data and the corresponding characteristic values of the real vibration signals, and if the error threshold value is exceeded, adjusting key parameters of the split parts according to the error so as to meet the errorAs a result.
4. A digital twin based product performance and degradation state analysis method according to claim 1, wherein the score of each part is calculated according to a Scoring function,
wherein% i Actrll is the percentage error for the ith sample i For the actual life time of the part, RUL i For the predicted life time of the part, N represents A i Number A i The RUL estimate accuracy transfer function representing part i, score represents the final Score for that part.
5. The method for analyzing the product performance and the degradation state based on the digital twin according to claim 1, wherein in the prediction integration module, a hierarchical analysis method, an expert evaluation method and a weighted analysis method or three methods are adopted to integrate the components of the whole machine in a fusion manner, so that the product performance evaluation is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310331163.7A CN116050190B (en) | 2023-03-31 | 2023-03-31 | Product performance and degradation state analysis method based on digital twinning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310331163.7A CN116050190B (en) | 2023-03-31 | 2023-03-31 | Product performance and degradation state analysis method based on digital twinning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116050190A CN116050190A (en) | 2023-05-02 |
| CN116050190B true CN116050190B (en) | 2023-06-23 |
Family
ID=86133660
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310331163.7A Active CN116050190B (en) | 2023-03-31 | 2023-03-31 | Product performance and degradation state analysis method based on digital twinning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116050190B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116405897B (en) * | 2023-06-09 | 2023-08-11 | 成都航空职业技术学院 | Novel intelligent building integration method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111695734A (en) * | 2020-06-12 | 2020-09-22 | 中国科学院重庆绿色智能技术研究院 | Multi-process planning comprehensive evaluation system and method based on digital twin and deep learning |
| CN113221277A (en) * | 2021-05-14 | 2021-08-06 | 西安交通大学 | Bearing performance degradation evaluation method and system based on digital twin model |
| WO2022105072A1 (en) * | 2020-11-19 | 2022-05-27 | 中国矿业大学 | Intelligent reliability evaluation and service life prediction method for kilometer deep well hoist brake |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11403541B2 (en) * | 2019-02-14 | 2022-08-02 | Rockwell Automation Technologies, Inc. | AI extensions and intelligent model validation for an industrial digital twin |
| CN114919559B (en) * | 2022-07-05 | 2023-05-12 | 中南大学 | System and method for predicting residual service life of brake system based on digital twin |
| CN115421400A (en) * | 2022-09-16 | 2022-12-02 | 青岛理工大学 | Digital twin virtual-real consistency discrimination method for service health state intelligent recognition decision |
| CN115422814B (en) * | 2022-11-04 | 2023-04-28 | 中国海洋大学 | Digital twin-driven closed-loop optimization design method for complex electromechanical product |
-
2023
- 2023-03-31 CN CN202310331163.7A patent/CN116050190B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111695734A (en) * | 2020-06-12 | 2020-09-22 | 中国科学院重庆绿色智能技术研究院 | Multi-process planning comprehensive evaluation system and method based on digital twin and deep learning |
| WO2022105072A1 (en) * | 2020-11-19 | 2022-05-27 | 中国矿业大学 | Intelligent reliability evaluation and service life prediction method for kilometer deep well hoist brake |
| CN113221277A (en) * | 2021-05-14 | 2021-08-06 | 西安交通大学 | Bearing performance degradation evaluation method and system based on digital twin model |
Non-Patent Citations (2)
| Title |
|---|
| Research and Analysis of Power Transformer Remaining Life Prediction Based on Digital Twin Technology;Yongteng Jing等;《IEEE Xplore》;全文 * |
| 数字孪生智采工作面技术架构研究;葛世荣;张帆;王世博;王忠宾;;煤炭学报(第06期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116050190A (en) | 2023-05-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| van Dinter et al. | Predictive maintenance using digital twins: A systematic literature review | |
| CN110377984B (en) | Industrial equipment residual effective life prediction method and system and electronic equipment | |
| CN110554657B (en) | Health diagnosis system and diagnosis method for operation state of numerical control machine tool | |
| Jiang et al. | Two-stage structural damage detection using fuzzy neural networks and data fusion techniques | |
| CN107941537A (en) | A kind of mechanical equipment health state evaluation method | |
| CN116050190B (en) | Product performance and degradation state analysis method based on digital twinning | |
| CN110096827B (en) | Shield tunneling machine parameter optimization method based on deep neural network | |
| CN114323644A (en) | Gear box fault diagnosis and signal acquisition method and device and electronic equipment | |
| DE10257793A1 (en) | Model based service life monitoring system, especially for forecasting the remaining service life of motor vehicle components, whereby existing instrumentation is used to provide data for a model for calculating wear | |
| CN115422814B (en) | Digital twin-driven closed-loop optimization design method for complex electromechanical product | |
| Chen et al. | Application of nonlinear output frequency response functions and deep learning to RV reducer fault diagnosis | |
| Wang et al. | A sequence-to-sequence model with attention and monotonicity loss for tool wear monitoring and prediction | |
| CN116881819B (en) | Stay cable working state monitoring method based on isolated forest | |
| CN116011125B (en) | Response prediction method for uncertain axle coupling system | |
| CN116010900A (en) | Multi-scale feature fusion gearbox fault diagnosis method based on self-attention mechanism | |
| CN113325721A (en) | Model-free adaptive control method and system for industrial system | |
| CN116049989A (en) | Bogie digital twin model construction method, system, electronic equipment and medium | |
| Xie et al. | A life prediction method of mechanical structures based on the phase field method and neural network | |
| Ji et al. | Learning the dynamics of time delay systems with trainable delays | |
| CN115660293A (en) | Comprehensive evaluation method for complete life cycle of complex electromechanical product based on digital twinning | |
| Juuso | Expertise and uncertainty processing with nonlinear scaling and fuzzy systems for automation | |
| CN114720129B (en) | Rolling bearing residual life prediction method and system based on bidirectional GRU | |
| CN115455754A (en) | Mine hydraulic support design method based on digital twinning | |
| Wang et al. | State-wise LSTM-GRU method for ball screw prediction | |
| Zheng et al. | A novel deep learning architecture and its application in dynamic load monitoring of the vehicle system |
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 | ||
| GR01 | Patent grant |