CN115438726A - Device life and fault type prediction method and system based on digital twin technology - Google Patents
Device life and fault type prediction method and system based on digital twin technology Download PDFInfo
- Publication number
- CN115438726A CN115438726A CN202211018147.4A CN202211018147A CN115438726A CN 115438726 A CN115438726 A CN 115438726A CN 202211018147 A CN202211018147 A CN 202211018147A CN 115438726 A CN115438726 A CN 115438726A
- Authority
- CN
- China
- Prior art keywords
- equipment
- data
- time
- service life
- twin
- 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.)
- Pending
Links
Images
Classifications
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a method and a system for predicting the service life and the fault type of equipment based on a digital twin technology, wherein the method comprises the following steps: (1) Building a physical equipment entity model and a digital twin virtual model (2) which are mapped with each other to obtain twin data, and extracting a comprehensive index of the twin data in an operating state; (3) Selecting a specific parameter for marking the service life of the equipment from twin data of the comprehensive index and the equipment performance data, and setting a critical value for the specific parameter; if the selected specific parameters are comprehensive indexes, turning to the step (4); if the selected specific parameters are equipment performance twin data, turning to the step (5); (4) Predicting the residual life of the service life by adopting a Kalman filtering algorithm; (5) And calculating the time corresponding to each parameter in the performance data of the equipment when reaching the critical value, and obtaining the residual life by combining the service life of the equipment. The invention combines the twin technology to predict the service life of different equipment according to the life decisive characteristic parameters.
Description
Technical Field
The invention relates to the technical field of digital twinning, in particular to a method and a system for predicting service life and fault types of equipment based on a digital twinning technology.
Background
The digital twin technology is taken as an enabling technology and means for practicing an intelligent manufacturing idea, can effectively solve the problem of information physical fusion of intelligent manufacturing, and has become a hot spot concerned by academic circles and industrial circles of various countries in the world at present. The real-time display and online monitoring of relevant state monitoring data such as equipment running state, product production quality and the like in the manufacturing process are increasingly urgent for production enterprises. At present, research based on predictive maintenance is limited to current state identification and basic alarm, so that data resources are not utilized, equipment operation information cannot be displayed visually, and maintenance and management in the whole life cycle process of equipment cannot be displayed.
In the prior art, twin technology is utilized to predict data of equipment in industrial production so as to facilitate monitoring and maintenance. Chinese patent application No. 202210173117.4 discloses a method for predicting the residual service life of a rolling bearing based on time-varying Kalman filtering, which estimates the residual service life of equipment by predicting future data. The method for predicting the service life is mainly based on data of actual equipment, virtual-real mapping and interactive fusion of a physical space and an information space are not realized, so that a digital twin system for realizing linkage control of a simulation entity and the actual equipment and mutual cooperation cannot be constructed, and the digital operation maintenance of the visual industrial equipment is realized. The Chinese patent application with the patent application number of 202111021923.1 discloses a method for constructing a fault model of a motor rolling bearing based on a digital Lisheng technology, wherein the fault type of equipment in industrial production is obtained by constructing twin data and combining a neural network, but the method mainly adopts the mode that the running state data of the equipment is single, and the accuracy of a prediction result is not high.
Disclosure of Invention
The invention aims to: aiming at the defects, the invention discloses a device service life and fault type prediction method based on a digital twin technology, aiming at different device service life decisive characteristic parameters, a twin technology and Kalman filtering are combined to predict the service life from device operation state data, and the twin technology is adopted to predict the service life of device performance data. The invention also provides a system for predicting the service life and the fault type of the equipment based on the digital twin technology, and the method can be used for obtaining the final predicted service life.
The technical scheme is as follows: in order to solve the above problems, the present invention provides a method for predicting a lifetime and a failure type of a device based on a digital twin technology, which specifically comprises the following steps:
(1) Building a physical equipment entity model and a digital twin virtual model, and associating the physical equipment entity model and the digital twin virtual model through a virtual mapping technology;
(2) Twin data corresponding to historical operating state data, fault data and equipment performance data in entity equipment are obtained; preprocessing twin data of the acquired running state data and extracting a comprehensive index of the running state data by adopting a principal component analysis method; the equipment performance data comprises the availability ratio, the average failure-free working time and the average repair time; the fault data includes a fault type;
(3) Selecting a specific parameter marking the service life of the equipment from twin data of the comprehensive index and the equipment performance data, and setting a critical value for the specific parameter; if the selected specific parameter marking the service life of the equipment is a comprehensive index, turning to the step (4); if the selected specific parameter marking the service life of the equipment is the availability or the average failure-free working time or the average repair time, then the step (5) is carried out;
(4) Acquiring a predicted value of the comprehensive index by adopting a Kalman filtering algorithm, acquiring time corresponding to the time when the predicted value of the comprehensive index reaches a critical value, wherein the time corresponds to the running time of equipment, and obtaining the residual life by combining the service life of the equipment;
(5) Calculating the time corresponding to the time when the availability or the average failure-free working time or the average repair time in the equipment performance data reaches a critical value, wherein the time corresponds to the running time of the equipment and is combined with the service life of the equipment to obtain the residual life; the specific calculation formulas are respectively as follows:
R＝t/T
wherein R represents the availability; t represents the working time; t represents the planned work time;
wherein MTBF represents mean time to failure; t represents the working time; f (t) represents the probability density function over time until the next failure;
wherein MTTR represents the mean repair time; t represents the working time; and N is the number of times of repair.
Further, the step (4) specifically comprises the following steps:
(4.1) acquiring a parameter measured value of a specific parameter, wherein the formula is as follows:
Z(k)＝HX(k)+V(k)
wherein Z (k) is a parameter measurement for a particular parameter; x (k) is a state value at the time k; v (k) is measurement noise; h is a parameter of the measuring system;
(4.2) acquiring a state quantity estimated value according to the parameter measured value, wherein the formula is as follows:
X(k|k)＝X(k|k-1)+Kg(k)(Z(k)-HX(k|k-1))
in the formula, X (k | k) is an estimated value of the current state; x (k | k-1) is an estimated value of the previous state; h is a parameter of the measuring system; kg (k) is a gain factor;
and (4.3) comparing the state quantity estimated value with a critical value, judging that the service life of the equipment is ended when the state quantity estimated value reaches the critical value, wherein the k value corresponding to the state quantity estimated value is the period of the equipment which has been operated, and subtracting the operated period from the service life of the equipment to obtain the residual service life of the equipment.
Further, the method also comprises the step (6): forming a set according to comprehensive indexes of the twin data extraction equipment in different fault types in the step (2), and dividing the set into a training set and a test set according to a proportion; constructing a DNN neural network model, wherein an input layer of the DNN neural network model is a comprehensive index, an output layer of the DNN neural network model is a fault type, and a formula of the DNN neural network model is as follows:
in the formula, h W,b (x) The type of fault output; x is a radical of a fluorine atom i Is an operation comprehensive index; w is a group of i Is a weight; b is an offset; k is a characteristic quantity in the comprehensive index;
training a DNN neural network model by using data of a training set, and acquiring the trained DNN neural network model when the training reaches a preset number; and inputting the test set data into the trained DNN neural network model for testing to obtain a prediction result of the fault model.
Further, the step (2) of preprocessing twin data of the acquired operation state data and extracting comprehensive indexes of the operation state data by adopting a principal component analysis method specifically comprises the following steps:
(2.1) preprocessing twin data of the acquired running state data, including removing abnormal values, extracting time domain features and extracting frequency domain features;
(2.2) the comprehensive indexes of the operation state data extracted by adopting a principal component analysis method are specifically as follows:
(2.2.1) converting the preprocessed data into a characteristic vector matrix;
(2.2.2) calculating the average value of each column of characteristics, and then subtracting the characteristic average value of the column from each dimension;
(2.2.3) calculating a covariance matrix of the features;
(2.2.4) performing calculation of eigenvalue and eigenvector for the covariance matrix;
(2.2.5) sorting the calculated characteristic values from large to small;
and (2.2.6) taking out the first K eigenvectors and eigenvalues, and performing backspacing to obtain a dimensionality-reduced eigenvector matrix.
Further, the step (6) further comprises: if the proportion of the inconsistent prediction result of the fault model obtained by the test set data and the actual model reaches a specific value, selecting historical operation state data in the entity equipment for preprocessing, extracting a comprehensive index, and mixing the comprehensive index with the comprehensive index extracted by the twin data; and (4) taking the comprehensive index of the mixed running state data as the input of the DNN neural network model for retraining.
In addition, the invention also provides a system for predicting the service life and the fault type of equipment based on the digital twin technology, which comprises the following steps:
building a module: the system comprises a virtual mapping technology, a physical device entity model and a digital twin virtual model, wherein the virtual mapping technology is used for establishing the physical device entity model and the digital twin virtual model and associating the physical device entity model and the digital twin virtual model through the virtual mapping technology;
a twin data acquisition module; the device comprises a building module, a data acquisition module and a data acquisition module, wherein the building module is used for acquiring the twin data corresponding to the historical operating state data, the fault data and the equipment performance data in the entity equipment through a digital twin virtual model in the building module;
a data preprocessing module: the system comprises a data acquisition unit, a data processing unit and a data processing unit, wherein the data acquisition unit is used for acquiring twin data of running state data; the equipment performance data comprises the availability ratio, the average failure-free working time and the average repair time; the fault data includes a fault type;
specific parameter selection and threshold setting module: selecting a specific parameter for marking the service life of the equipment from twin data of the comprehensive index and the equipment performance data, and setting a critical value for the specific parameter; if the selected specific parameter marking the service life termination of the equipment is a comprehensive index, a first service life prediction module is adopted for model prediction; if the selected specific parameter marking the service life termination of the equipment is the availability or the average failure-free working time or the average repair time, adopting a second service life prediction module;
a first life prediction module: the system is used for acquiring the pre-estimated value of the comprehensive index by adopting a Kalman filtering algorithm and acquiring the time corresponding to the pre-estimated value of the comprehensive index reaching a critical value, wherein the time corresponds to the running time of equipment, and the residual life is obtained by combining the service life of the equipment;
a second life prediction module: the method is used for calculating the time corresponding to the time when the availability or the average failure-free working time or the average repair time in the equipment performance data reaches a critical value, wherein the time corresponds to the running time of the equipment, and the residual life is obtained by combining the service life of the equipment; the specific calculation formulas are respectively as follows:
R＝t/T
wherein R represents the availability; t represents the working time; t represents the planned work time;
wherein MTBF represents mean time to failure; t represents the working time; f (t) represents a probability density function until the next lapse of time to failure;
wherein MTTR represents the mean repair time; t represents the working time; and N is the number of times of repair.
Further, the method also comprises the following steps:
a fault type prediction module: the device is used for extracting comprehensive indexes of equipment under different fault types from the twin data acquisition module to form a set, and dividing the set into a training set and a test set according to a proportion; constructing a DNN neural network model, wherein an input layer of the DNN neural network model is a comprehensive index, an output layer of the DNN neural network model is a fault type, and a formula of the DNN neural network model is as follows:
in the formula, h W,b (x) The type of fault output; x is a radical of a fluorine atom i Is an operation comprehensive index; w i Is a weight; b is an offset; k is a characteristic quantity in the comprehensive index;
training a DNN neural network model by using data of a training set, and acquiring the trained DNN neural network model when the training reaches a preset number; and inputting the test set data into the trained DNN neural network model for testing to obtain the prediction result of the fault model.
Furthermore, the present invention also provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of any one of the above methods when executing the computer program. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of any of the methods described above.
Has the advantages that: compared with the prior art, the device life and fault type prediction method based on the digital twin technology has the advantages that: 1. setting a digital twin system, and directly mapping the current physical world parameters to a twin world in real time through simulation to obtain twin data; obtaining a residual life prediction result by combining twin data of the motion state of the equipment and a Kalman filtering mode; considering reference factors of different equipment service lives in different aspects, adding equipment performance information factors such as equipment availability, average fault-free working time and average repair time, and obtaining a residual service life prediction result through twin data of the reference factors; 2. and performing principal component analysis on the twin data of the motion state to extract comprehensive index data, and combining the comprehensive index data with a neural network to obtain a more accurate fault type analysis result.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a block diagram illustrating modeling of virtual mapping association between data twin devices and physical devices according to the present invention;
FIG. 3 illustrates a neural network element of the fault identification model of the present invention;
fig. 4 shows a schematic diagram of the system according to the invention.
Detailed Description
The technical scheme of the invention is further explained by combining the attached drawings.
As shown in fig. 1, the present invention provides a method for predicting the lifetime and the failure type of a device based on a digital twin technology, which specifically includes the following steps:
step one, a physical equipment entity model and a digital twin virtual model are set up, and the physical equipment entity model and the digital twin virtual model are associated through a virtual mapping technology.
(1) Collecting data of physical equipment, and building physical equipment entity modeling, wherein the model is as follows:
in the formula, PE represents a physical entity of the device, and PS represents physical space information of the device, including environment information and physical parameters of the device; the PC represents equipment state data, and vibration, temperature, pressure, current and voltage and the like of the equipment during operation are acquired through intelligent acquisition equipment; the PO is the basic information data of the equipment, including metadata of the equipment (service life, equipment parameters, etc. on the factory specifications), past maintenance record information data of the equipment, and operation data of the equipment (total output, production rate, yield, etc.).
Indicates natural connection among PS, PC and PO, and indicates thatThe natural interaction relationship among the three is realized. PS, PC, PO are all dynamic sets, the set elements and their states are constantly updated with the dynamic operation of the manufacturing process.
(2) Constructing a digital twin virtual model, wherein the model specifically comprises the following steps:
wherein VE represents a digital twin of the equipment, DS represents physical space information of the twin, wherein the environment information is an information element which can be simulated, and the physical information is information such as the geometric dimension, the physical structure, the motion attitude, the motion characteristic and the like based on the physical entity of the equipment; the DC twin equipment state data is obtained by implementing mapping or simulation and the like; the DO is device twin basic information data including metadata inherited from a device entity, operation data generated during a twin free growth process, and the like.
Natural connection among DS, DC and DO is shown, and natural interaction relation among the DS, the DC and the DO is shown. DS, DC, DO are also dynamic sets that are constantly updated as the twins grow.
(3) On the basis of establishing a physical space entity model PE and an information space twin model VE, virtual-real mapping association between a physical equipment entity model and a digital twin virtual model is further established, and a specific mapping relation model is as follows:
in the formula (I), the compound is shown in the specification,
representing a two-way true mapping between the physical space mockup and the digital twins.
As shown in fig. 2, the entire virtual mapping association modeling includes simulation and free growth. The method comprises the steps of collecting the running state of equipment at the early stage of the life cycle of the equipment, carrying out digital mapping on the equipment according to known basic data such as physical information, function information, performance information and the like of the equipment and based on a definite space environment by utilizing technologies such as cloud computing, internet of things, three-dimensional modeling, graphic rendering and the like, generating a mirror image of a physical entity in a virtual space, reducing the physical simulation process, and extracting certain characteristics and parameters of physical simulation so as to present the simulation running condition of the physical entity, taking the actual data of the equipment generated by an external sensor and a central processing unit as the input of a twin body for simulation, wherein the simulation process is to directly map the current physical world parameters into a world. The free generation is based on a digital twin prototype generated by analog simulation, and the construction of a twin world is realized by utilizing industrial big data and an Artificial Intelligence (AI) machine learning technology.
And secondly, acquiring twin data corresponding to historical operating state data, fault data and equipment performance data in the entity equipment respectively according to the established entity model and the digital twin model. The performance data of the equipment mainly comprises the availability ratio, the average fault-free working time and the average repair time of the equipment; the fault data includes a fault type for a particular device.
Since the device predictive diagnosis realized by the digital twin is a prediction method aiming at the whole life cycle of the device, under the condition of only possessing a small amount of samples, the device data needs to be preprocessed first, and effective parameters of model training are obtained from discrete data. In particular, the method comprises the following steps of,
(1) Preprocessing twin data of the acquired running state data; the method comprises the following steps:
(1.1) removing abnormal values
And respectively carrying out primary data cleaning on the acquired various equipment state parameters, and removing abnormal values after data correction is realized.
(1.2) time-domain feature extraction
Statistical feature quantities related to time domain feature extraction include, but are not limited to, mean, standard deviation, root mean square, kurtosis, form factor, kurtosis factor, and impulse factor.
(1.3) frequency domain feature extraction
In order to improve the sensitivity and stability of equipment predictability maintenance, frequency domain feature extraction is performed by utilizing Fourier transform besides time domain features.
Wherein X (t) represents a signal, X (f) represents a spectrum of X (t) after FFT, and FFT (i) rms Represents the root mean square value of the ith subband after FFT, FFT (m) represents the mth spectral line of the ith subband, and N represents the number of spectral lines.
(1.4) extracting comprehensive indexes of the running state data by adopting a principal component analysis method;
(1.4.1) converting the preprocessed data into a characteristic vector matrix;
(1.4.2) calculating the average value of each column of characteristics, and then subtracting the characteristic average value of the column from each dimension;
(1.4.3) calculating a covariance matrix of the features;
(1.4.4) performing calculation of eigenvalue and eigenvector for the covariance matrix;
(1.4.5) sorting the calculated characteristic values from large to small;
(1.4.6) taking out the first K eigenvectors and eigenvalues, and performing backspacing to obtain a dimensionality-reduced eigenvector matrix, wherein K is a positive integer.
Step three, constructing a RUL prediction model for life prediction
(1) Selecting specific parameters marking the service life of the equipment from twin data of acquiring comprehensive indexes and equipment performance data, and setting a critical value CV for the specific parameters i 。
If the comprehensive index can better reflect the characteristics of the service life of the equipment, namely the selected specific parameter marking the service life termination of the equipment is any index in the comprehensive index, the step is switched to the step four.
If the parameters in the twin data of the device performance data are considered to be more capable of characterizing the device life, that is, the selected specific parameters for marking the end of the device life are one or a combination of the available rate, the average failure-free working time and the average repair time, the step five is executed.
Step four, adopting a Kalman filter algorithm to predict the residual service life of the equipment
(1) Firstly, acquiring a parameter measurement value of a specific parameter, wherein the formula is as follows:
Z(k)＝HX(k)+V(k)
wherein Z (k) is a parameter measurement for a particular parameter; x (k) is a state value at the time k; v (k) is measurement noise; h is a parameter of the measuring system;
(2) And acquiring a state quantity estimated value according to the parameter measured value, wherein the formula is as follows:
X(k|k)＝X(k|k-1)+Kg(k)(Z(k)-HX(k|k-1))
in the formula, X (k | k) is an estimated value of the current state; x (k | k-1) is an estimated value of the previous state; h is a parameter of the measuring system; kg (k) is a gain factor;
(3) The state quantity estimated value and a critical value CV are calculated i Comparing, when the state quantity estimated value X (k | k) reaches the critical value CV i And judging that the service life of the equipment is ended, wherein the k value corresponding to the state quantity estimated value is the running period of the equipment, and the residual service life of the equipment is obtained by subtracting the running period from the service life of the equipment.
Step five, according to the selected specific parameters, if the selected characteristic parameters are the availability, calculating the running time of the equipment according to a formula, and obtaining the remaining life by combining the service life of the equipment; the formula is as follows:
R＝t/T
wherein R represents the availability; t represents the working time; t represents the scheduled work time;
if the selected characteristic parameter is the average fault-free working time, calculating the running time of the equipment according to a formula, and obtaining the residual service life by combining the service life of the equipment; the formula is as follows:
wherein MTBF represents mean time to failure; t represents the working time; f (t) represents the probability density function over time until the next failure;
if the selected characteristic parameter is the average repair time, calculating the running time of the equipment according to a formula, and obtaining the residual life by combining the service life of the equipment; the formula is as follows:
wherein MTTR represents the mean repair time; t represents the working time; and N is the number of times of repair.
Step six, constructing a fault type judgment model
(1) And extracting twin data of the equipment in running states under different fault types from the digital twin virtual model, and obtaining comprehensive indexes under different fault types through the principal component analysis. And (3) forming a training set by the obtained comprehensive indexes corresponding to different fault types, and dividing the set into the training set and the test set according to the proportion of 7.
(2) Constructing a DNN neural network model, as shown in FIG. 3, wherein an input layer of the DNN neural network model is a comprehensive index, an output layer of the DNN neural network model is a fault type, and a formula of the DNN neural network model is as follows:
in the formula, h W,b (x) The type of fault output; x is the number of i Is an operation comprehensive index; w i Is a weight; b is an offset; k is a characteristic quantity in the comprehensive index;
(3) Training a DNN neural network model by using data of a training set, and acquiring the trained DNN neural network model when the training reaches a preset number N; and inputting the test set data into the trained DNN neural network model for testing to obtain the prediction result of the fault model.
In addition, if the inconsistent ratio of the prediction result of the fault model obtained by the test set data and the actual model reaches a specific value, the accuracy of the fault prediction model is determined to be low, and the model needs to be trained again. Selecting historical running state data in entity equipment for preprocessing, extracting comprehensive indexes, and mixing the comprehensive indexes with comprehensive indexes extracted from the original twin data; and (5) training again by taking the comprehensive index of the mixed running state data as the input of the DNN neural network model.
Specifically, the ticket selling and checking gate is taken as an example, and original data sources are all from historical operating data, fault data and basic information of the gate which is put into use. Wherein, statistics after sales ticket checking floodgate equipment fault type includes: motor damage, fan damage, swing door failure, power damage, mainboard damage. The principal component analysis method is adopted to extract and analyze the temperature, the equipment power, the motor torque force and the door opening and closing time which are taken as comprehensive indexes, compare the characteristics of the comprehensive indexes, performance indexes and the like, combine the parameter characteristics of the service life of the gate equipment and select the average fault-free working time as a parameter for marking the service life termination of the gate, and the calibration value is 80,000h. The method of the invention is adopted for 5 pieces of gate equipment in use, and the prediction result at a certain time point is shown in the following table 1:
TABLE 1 prediction results
| Gate A | Gate B | Gate C | Gate D | Gate E | |
| Type of failure | - | Failure of the swing door | - | - | - |
| Residual life (h) | 89762 | 83641 | 87695 | 90352 | 88634 |
In addition, as shown in fig. 4, the present invention further provides a system for predicting the life and failure type of a device based on a digital twin technology, comprising: building a module: the system comprises a virtual mapping module, a physical equipment entity model and a digital twin virtual model, wherein the virtual mapping module is used for establishing the physical equipment entity model and the digital twin virtual model and associating the physical equipment entity model and the digital twin virtual model through a virtual mapping technology; a twin data acquisition module; the device comprises a building module, a data acquisition module and a data acquisition module, wherein the building module is used for acquiring the twin data corresponding to the historical operating state data, the fault data and the equipment performance data in the entity equipment through a digital twin virtual model in the building module; a data preprocessing module: the system comprises a data acquisition unit, a data processing unit and a data processing unit, wherein the data acquisition unit is used for acquiring twin data of running state data; the equipment performance data comprises the availability ratio, the average failure-free working time and the average repair time; the fault data includes a fault type; specific parameter selection and threshold setting module: selecting a specific parameter for marking the service life of the equipment from twin data of the comprehensive index and the equipment performance data, and setting a critical value for the specific parameter; if the selected specific parameter marking the service life termination of the equipment is a comprehensive index, a first service life prediction module is adopted for model prediction; if the selected specific parameter marking the service life termination of the equipment is the availability or the average failure-free working time or the average repair time, adopting a second service life prediction module; a first life prediction module: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring a pre-evaluation value of a comprehensive index by adopting a Kalman filtering algorithm and acquiring time corresponding to a critical value when the pre-evaluation value of the comprehensive index reaches the critical value, the time corresponds to the running time of equipment, and the residual life is acquired by combining the service life of the equipment; a second life prediction module: the method is used for calculating the time corresponding to the time when the availability or the average failure-free working time or the average repair time in the equipment performance data reaches a critical value, wherein the time corresponds to the running time of the equipment, and the residual life is obtained by combining the service life of the equipment; the specific calculation formulas are respectively as follows:
R＝t/T
wherein R represents the availability; t represents the working time; t represents the planned work time;
wherein MTBF represents mean time to failure; t represents the working time; f (t) represents a probability density function until the next lapse of time to failure;
wherein MTTR represents the mean repair time; t represents the working time; and N is the number of times of repair.
A fault type prediction module: the comprehensive indexes used for the equipment in the self-twinning data acquisition module under different fault types form a set, and the set is divided into a training set and a test set according to a proportion; constructing a DNN neural network model, wherein an input layer of the DNN neural network model is a comprehensive index, an output layer of the DNN neural network model is a fault type, and a formula of the DNN neural network model is as follows:
in the formula, h W,b (x) The type of fault output; x is the number of i Is an operation comprehensive index; w i Is a weight; b is an offset; k is a characteristic quantity in the comprehensive index; training a DNN neural network model by using data of a training set, and acquiring the trained DNN neural network model when the training reaches a preset number; and inputting the test set data into the trained DNN neural network model for testing to obtain the prediction result of the fault model.
The system can be used for acquiring equipment data, storing, calculating and processing prediction data and twin data. The system can rely on big data analysis processing and visualization related technologies, a comprehensive management platform of the whole life cycle of the equipment is set up, and management means of technologies such as centralized management, real-time monitoring and timely repair reporting are provided while the twin model of the equipment is displayed based on the history of the whole life cycle of the equipment. The functions of the integrated management system include but are not limited to equipment management, emergency maintenance, planned maintenance, statistical analysis, billboard management and the like, and an integrated interactive platform supporting high-level decision, middle-level control and basic-level operation is provided.
Claims (9)
1. A device life and fault type prediction method based on a digital twin technology is characterized by comprising the following steps:
(1) Building a physical equipment entity model and a digital twin virtual model, and associating the physical equipment entity model and the digital twin virtual model through a virtual mapping technology;
(2) Twin data corresponding to historical operating state data, fault data and equipment performance data in entity equipment are obtained; preprocessing twin data of the acquired running state data and extracting comprehensive indexes of the running state data by adopting a principal component analysis method; the equipment performance data comprises the availability ratio, the average failure-free working time and the average repair time; the fault data includes a fault type;
(3) Selecting a specific parameter for marking the service life of the equipment from twin data of the comprehensive index and the equipment performance data, and setting a critical value for the specific parameter; if the selected specific parameter marking the service life of the equipment is a comprehensive index, turning to the step (4); if the selected specific parameter for marking the service life of the equipment is the available rate or the average failure-free working time or the average repair time, turning to the step (5);
(4) Acquiring a pre-evaluation value of the comprehensive index by adopting a Kalman filtering algorithm and acquiring time corresponding to the pre-evaluation value of the comprehensive index reaching a critical value, wherein the time corresponds to the running time of equipment, and the residual service life is obtained by combining the service life of the equipment;
(5) Calculating the time corresponding to the time when the available rate or the average failure-free working time or the average repair time in the performance data of the equipment reaches a critical value, wherein the time corresponds to the running time of the equipment, and the residual life is obtained by combining the service life of the equipment; the specific calculation formulas are respectively as follows:
R＝t/T
wherein R represents the availability; t represents the working time; t represents the scheduled work time;
wherein MTBF represents mean time to failure; t represents the working time; f (t) represents the probability density function over time until the next failure;
wherein MTTR represents the mean repair time; t represents the working time; and N is the number of times of repair.
2. The method for predicting the life span and the fault type of the equipment based on the digital twinning technique as claimed in claim 1, wherein the step (4) comprises the following steps:
(4.1) acquiring a parameter measurement value of a specific parameter, wherein the formula is as follows:
Z(k)＝HX(k)+V(k)
wherein Z (k) is a parameter measurement for a particular parameter; x (k) is a state value at the time k; v (k) is measurement noise; h is a parameter of the measuring system;
(4.2) acquiring a state quantity estimated value according to the parameter measured value, wherein the formula is as follows:
X(k|k)＝X(k|k-1)+Kg(k)(Z(k)-HX(k|k-1))
in the formula, X (k | k) is an estimated value of the current state; x (k | k-1) is an estimated value of the previous state; h is a parameter of the measuring system; kg (k) is a gain factor;
and (4.3) comparing the state quantity estimated value with a critical value, judging that the service life of the equipment is ended when the state quantity estimated value reaches the critical value, wherein the k value corresponding to the state quantity estimated value is the period of the equipment which has run, and subtracting the running period from the service life of the equipment to obtain the residual service life of the equipment.
3. The method for predicting the service life and the fault type of the equipment based on the digital twin technology as claimed in claim 1, further comprising:
(6) Forming a set according to comprehensive indexes of the twin data extraction equipment in different fault types in the step (2), and dividing the set into a training set and a test set according to a proportion; constructing a DNN neural network model, wherein an input layer of the DNN neural network model is a comprehensive index, an output layer of the DNN neural network model is a fault type, and a formula of the DNN neural network model is as follows:
in the formula, h W,b (x) The type of fault output; x is a radical of a fluorine atom i Is an operation comprehensive index; w i Is a weight; b is an offset; k is a characteristic quantity in the comprehensive index;
training a DNN neural network model by using data of the training set, and acquiring the trained DNN neural network model when the training reaches a preset number of times; and inputting the test set data into the trained DNN neural network model for testing to obtain a prediction result of the fault model.
4. The method for predicting the service life and the fault type of the equipment based on the digital twin technology as claimed in claim 1, wherein the step (2) of preprocessing the twin data of the acquired operation state data and extracting the comprehensive index of the operation state data by adopting a principal component analysis method specifically comprises the following steps:
(2.1) preprocessing twin data of the acquired running state data, including removing abnormal values, extracting time domain features and extracting frequency domain features;
(2.2) the comprehensive indexes of the operation state data extracted by adopting a principal component analysis method are specifically as follows:
(2.2.1) converting the preprocessed data into a characteristic vector matrix;
(2.2.2) calculating the average value of each column of features, and then subtracting the average value of the column of features from each dimension;
(2.2.3) calculating a covariance matrix of the features;
(2.2.4) performing calculation of eigenvalue and eigenvector for the covariance matrix;
(2.2.5) sorting the calculated characteristic values from large to small;
and (2.2.6) taking out the first K eigenvectors and eigenvalues, and performing backspacing to obtain a dimensionality-reduced eigenvector matrix.
5. The method for predicting the service life and the fault type of the equipment based on the digital twin technology as claimed in claim 2, wherein the step (6) further comprises the following steps: if the inconsistent ratio of the prediction result of the fault model obtained by the test set data and the actual model reaches a specific value, selecting historical operation state data in the entity equipment for preprocessing, extracting a comprehensive index, and mixing the comprehensive index with the comprehensive index extracted by the original twin data; and (4) taking the comprehensive index of the mixed running state data as the input of the DNN neural network model for retraining.
6. A system for predicting life and failure type of equipment based on digital twinning technique, comprising:
building a module: the system comprises a virtual mapping technology, a physical device entity model and a digital twin virtual model, wherein the virtual mapping technology is used for establishing the physical device entity model and the digital twin virtual model and associating the physical device entity model and the digital twin virtual model through the virtual mapping technology;
a twin data acquisition module; the system comprises a building module, a data acquisition module and a data transmission module, wherein the building module is used for building a digital twin virtual model;
a data preprocessing module: the system comprises a data acquisition unit, a data processing unit and a data processing unit, wherein the data acquisition unit is used for acquiring twin data of running state data; the equipment performance data comprises the availability ratio, the average failure-free working time and the average repair time; the fault data includes a fault type;
the specific parameter selection and threshold setting module: selecting a specific parameter marking the service life of the equipment from twin data of the comprehensive index and the equipment performance data, and setting a critical value for the specific parameter; if the selected specific parameter marking the service life termination of the equipment is a comprehensive index, a first service life prediction module is adopted for model prediction; if the selected specific parameter marking the service life termination of the equipment is the availability or the average failure-free working time or the average repair time, adopting a second service life prediction module;
a first life prediction module: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring a pre-evaluation value of a comprehensive index by adopting a Kalman filtering algorithm and acquiring time corresponding to a critical value when the pre-evaluation value of the comprehensive index reaches the critical value, the time corresponds to the running time of equipment, and the residual life is acquired by combining the service life of the equipment;
a second life prediction module: the method is used for calculating the time corresponding to the time when the availability or the average failure-free working time or the average repair time in the performance data of the equipment reaches a critical value, wherein the time corresponds to the running time of the equipment and is combined with the service life of the equipment to obtain the residual life; the specific calculation formulas are respectively as follows:
R＝t/T
wherein R represents the availability; t represents the working time; t represents the planned work time;
wherein MTBF represents mean time to failure; t represents the working time; f (t) represents the probability density function over time until the next failure;
wherein MTTR represents the mean repair time; t represents the working time; and N is the number of times of repair.
7. The system of claim 5, further comprising:
a fault type prediction module: the comprehensive indexes used for the equipment in the self-twinning data acquisition module under different fault types form a set, and the set is divided into a training set and a test set according to a proportion; constructing a DNN neural network model, wherein an input layer of the DNN neural network model is a comprehensive index, an output layer of the DNN neural network model is a fault type, and a formula of the DNN neural network model is as follows:
in the formula, h W,b (x) The type of fault output; x is a radical of a fluorine atom i Is an operation comprehensive index; w is a group of i Is a weight; b is an offset; k is a characteristic quantity in the comprehensive index;
training a DNN neural network model by using data of the training set, and acquiring the trained DNN neural network model when the training reaches a preset number of times; and inputting the test set data into the trained DNN neural network model for testing to obtain the prediction result of the fault model.
8. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the steps of the method of any one of claims 1 to 5 are performed when the computer program is executed by the processor.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211018147.4A CN115438726A (en) | 2022-08-24 | 2022-08-24 | Device life and fault type prediction method and system based on digital twin technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211018147.4A CN115438726A (en) | 2022-08-24 | 2022-08-24 | Device life and fault type prediction method and system based on digital twin technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115438726A true CN115438726A (en) | 2022-12-06 |
Family
ID=84245459
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211018147.4A Pending CN115438726A (en) | 2022-08-24 | 2022-08-24 | Device life and fault type prediction method and system based on digital twin technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115438726A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116070840A (en) * | 2022-12-26 | 2023-05-05 | 北京国网富达科技发展有限责任公司 | Transformer collaborative management method and system based on power grid digital twin model |
| CN117254596A (en) * | 2023-10-10 | 2023-12-19 | 雷玺智能科技(上海)有限公司 | Digital twinning-based energy storage power station full life cycle monitoring system and method |
| CN117592977A (en) * | 2024-01-19 | 2024-02-23 | 陕西万禾数字科技有限公司 | Intelligent digital twin guarantee system for equipment full life cycle |
-
2022
- 2022-08-24 CN CN202211018147.4A patent/CN115438726A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116070840A (en) * | 2022-12-26 | 2023-05-05 | 北京国网富达科技发展有限责任公司 | Transformer collaborative management method and system based on power grid digital twin model |
| CN116070840B (en) * | 2022-12-26 | 2023-10-27 | 北京国网富达科技发展有限责任公司 | Transformer collaborative management method and system based on power grid digital twin model |
| CN117254596A (en) * | 2023-10-10 | 2023-12-19 | 雷玺智能科技(上海)有限公司 | Digital twinning-based energy storage power station full life cycle monitoring system and method |
| CN117254596B (en) * | 2023-10-10 | 2024-04-09 | 雷玺智能科技(上海)有限公司 | Digital twinning-based energy storage power station full life cycle monitoring system and method |
| CN117592977A (en) * | 2024-01-19 | 2024-02-23 | 陕西万禾数字科技有限公司 | Intelligent digital twin guarantee system for equipment full life cycle |
| CN117592977B (en) * | 2024-01-19 | 2024-04-09 | 陕西万禾数字科技有限公司 | Intelligent digital twin guarantee system for equipment full life cycle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106951695B (en) | Method and system for calculating residual service life of mechanical equipment under multiple working conditions | |
| CN111830408B (en) | Motor fault diagnosis system and method based on edge calculation and deep learning | |
| CN115438726A (en) | Device life and fault type prediction method and system based on digital twin technology | |
| Lindemann et al. | Anomaly detection and prediction in discrete manufacturing based on cooperative LSTM networks | |
| CN114723285B (en) | Power grid equipment safety evaluation prediction method | |
| CN111639783A (en) | Line loss prediction method and system based on LSTM neural network | |
| CN112859822A (en) | Equipment health analysis and fault diagnosis method and system based on artificial intelligence | |
| CN116150897A (en) | Machine tool spindle performance evaluation method and system based on digital twin | |
| CN114493049A (en) | Production line optimization method and device based on digital twin, electronic equipment and medium | |
| CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
| CN109299208B (en) | Intelligent visual risk assessment method for transmission tower in typhoon disaster | |
| CN111638988A (en) | Cloud host fault intelligent prediction method based on deep learning | |
| CN114266289A (en) | Complex equipment health state assessment method | |
| CN114519923A (en) | Intelligent diagnosis and early warning method and system for power plant | |
| CN115204491A (en) | Production line working condition prediction method and system based on digital twinning and LSTM | |
| CN112507479A (en) | Oil drilling machine health state assessment method based on manifold learning and softmax | |
| CN110750455B (en) | Intelligent online self-updating fault diagnosis method and system based on system log analysis | |
| CN106096170B (en) | Wind turbines multivariable failure prediction method based on data-driven | |
| CN114548494A (en) | Visual cost data prediction intelligent analysis system | |
| CN116821610B (en) | Method for optimizing wind power generation efficiency by utilizing big data | |
| CN117371207A (en) | Extra-high voltage converter valve state evaluation method, medium and system | |
| CN112286088A (en) | Method and application system for online application of power equipment fault prediction model | |
| CN108345996B (en) | System and method for reducing wind power assessment electric quantity | |
| CN116205044A (en) | Real-time health analysis method and system for complex aerospace equipment | |
| CN115600695A (en) | Fault diagnosis method of metering equipment |
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 |