A device life and fault type prediction method and system based on digital twin technology

Abstract

The application discloses a device life and fault type prediction method and system based on digital twin technology, comprising: (1) building a mutually mapped physical device entity model and a digital twin virtual model (2) obtaining twin data, extracting comprehensive indexes of running state twin data; (3) selecting specific parameters indicating the end of the life of the device from the twin data of the comprehensive indexes and the device performance data, and setting a critical value for the specific parameters; if the selected specific parameter is a comprehensive index, go to step (4); if the selected specific parameter is device performance twin data, go to step (5); (4) using Kalman filtering algorithm to predict the remaining life; (5) calculating the time when each parameter in the device performance data reaches the critical value, and combining the service life of the device to obtain the remaining life. The application combines twin technology to predict the life of different device life determining characteristic parameters.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of digital twinning, in particular to a device life and fault type prediction method and system based on digital twinning technology. BACKGROUND

[0002] As an enabling technology and means for practicing the concept of intelligent manufacturing, digital twinning technology can effectively solve the information-physical fusion problem of intelligent manufacturing, and has become a hot spot of attention in the academic and industrial circles of various countries in the world. Production enterprises have an increasingly urgent need for real-time display and online monitoring of relevant state monitoring data such as device operating state and product production quality in the manufacturing process. At present, the research based on predictive maintenance is limited to current state identification and basic alarm, so that data resources are not utilized, and device operating information cannot be intuitively displayed, nor can the maintenance and management in the whole life cycle of the device be displayed.

[0003] In the prior art, data of devices in industrial production is predicted by using twinning technology to facilitate monitoring and maintenance. A Chinese patent application with the application number 202210173117.4 discloses a "rolling bearing residual service life prediction method based on time-varying Kalman filtering", which estimates the residual service life of the device by predicting future data. The life prediction method mainly predicts based on the data of the actual device, and does not realize the virtual-real mapping and interactive fusion of the physical space and the information space, so as to construct a digital twinning system that simulates the entity and the actual device linkage control and mutual coordination, and realizes the visualized digital operation and maintenance of industrial devices. A Chinese patent application with the application number 202111021923.1 discloses a "motor rolling bearing fault model construction method based on digital twinning technology", which obtains the fault type of the device in industrial production by constructing twinning data and combining a neural network. However, the method mainly uses single device operating state data, and the prediction result accuracy is not high. SUMMARY

[0004] The application discloses a device life and fault type prediction method based on digital twinning technology, which is aimed at different device life determining characteristic parameters, combines twinning technology and Kalman filtering to predict the life from device operating state data, and uses twinning technology to predict the life of device performance data. The application also provides a device life and fault type prediction system based on digital twinning technology, which can realize the above method to obtain the final predicted life.

[0005] Technical scheme: To solve the above problems, the application provides a device life and fault type prediction method based on digital twinning technology, which specifically includes the following steps:

[0006] (1) build a physical equipment entity model and a digital twin virtual model, and associate the physical equipment entity model and the digital twin virtual model through virtual mapping technology;

[0007] (2) obtain twin data corresponding to historical running state data, fault data and equipment performance data of the entity equipment; pre-process the obtained twin data of the running state data and extract comprehensive indexes of the running state data by principal component analysis; the equipment performance data includes availability, average trouble-free working time and average repair time; the fault data includes fault type;

[0008] (3) select a specific parameter indicating the end of the life of the equipment from the comprehensive indexes and the twin data of the equipment performance data, and set a critical value for the specific parameter; if the selected specific parameter indicating the end of the life of the equipment is the comprehensive index, go to step (4); if the selected specific parameter indicating the end of the life of the equipment is the availability or the average trouble-free working time or the average repair time, go to step (5);

[0009] (4) obtain an estimated value of the comprehensive index by using Kalman filtering algorithm and obtain a time corresponding to the critical value of the estimated value of the comprehensive index, which corresponds to the running time of the equipment, and obtain the remaining life by combining the service life of the equipment;

[0010] (5) calculate the time corresponding to the critical value of the availability or the average trouble-free working time or the average repair time in the equipment performance data, which corresponds to the running time of the equipment, and obtain the remaining life by combining the service life of the equipment; the specific calculation formulas are as follows:

[0011] R=t / T

[0012] In the formula, R represents the availability; t represents the working time; and T represents the planned working time.

[0013]

[0014] In the formula, MTBF represents the average trouble-free working time; t represents the working time; and f(t) represents the probability density function of the time elapsed until the next failure.

[0015]

[0016] In the formula, MTTR represents the average repair time; t represents the working time; and N represents the number of repairs.

[0017] Further, step (4) specifically includes the following steps:

[0018] (4.1) obtain the parameter measurement value of the specific parameter, and the formula is:

[0019] Z(k) = HX(k) + V(k)

[0020] In the formula, Z(k) is the parameter measurement value of a specific parameter; X(k) is the state value at time k; V(k) is the measurement noise; H is the parameter of the measurement system;

[0021] (4.2) Obtain the state quantity estimation value according to the parameter measurement value, and the formula is:

[0022] X(k|k) = X(k|k-1) + Kg(k)(Z(k) - HX(k|k-1))

[0023] In the formula, X(k|k) is the estimated value of the current state; X(k|k-1) is the estimated value of the previous state; H is the parameter of the measurement system; Kg(k) is the gain factor;

[0024] (4.3) Compare the state quantity estimation value with the critical value, and when the state quantity estimation value reaches the critical value, it is determined that the equipment life ends, and the k value corresponding to the state quantity estimation value is the period of equipment operation, and the remaining life of the equipment is obtained by subtracting the period of operation from the service life of the equipment.

[0025] Further, it further includes the following steps (6): according to the twin data in step (2), a set of comprehensive indexes under different fault types of the equipment is extracted, and the set is divided into a training set and a test set according to a proportion; a DNN neural network model is constructed, the input layer of the DNN neural network model is the comprehensive index, and the output layer is the fault type, and the formula of the DNN neural network model is:

[0026]

[0027] In the formula, h W,b (x) is the output fault type; x i is the running comprehensive index; W i is the weight; b is the bias; and k is the feature quantity in the comprehensive index.

[0028] The data of the training set is used to train the DNN neural network model, and when the training reaches a preset number of times, a trained DNN neural network model is obtained; the test set data is input into the trained DNN neural network model for testing to obtain the prediction result of the fault model.

[0029] Further, the twin data of the obtained running state data in step (2) is preprocessed and the comprehensive index of the running state data is extracted by using the principal component analysis method, which specifically includes:

[0030] (2.1) The pre-processing of the twin data of the obtained running state data includes removing outliers, time domain feature extraction, and frequency domain feature extraction;

[0031] (2.2) The comprehensive index of the running state data is extracted by principal component analysis method, and specifically is:

[0032] (2.2.1) The preprocessed data is converted into a feature vector matrix;

[0033] (2.2.2) The average value of each column feature is calculated, and then each dimension needs to be subtracted from the average value of the column feature;

[0034] (2.2.3) The covariance matrix of the feature is calculated;

[0035] (2.2.4) The eigenvalue and eigenvector of the covariance matrix are calculated;

[0036] (2.2.5) The calculated eigenvalues are sorted from large to small;

[0037] (2.2.6) The first K eigenvectors and eigenvalues are taken out, and the dimensionality reduction feature matrix is obtained by backtracking.

[0038] Further, step (6) further comprises: if the proportion of the prediction result of the fault model obtained by the test set data inconsistent with the actual model reaches a certain value, the historical running state data in the entity device is preprocessed and the comprehensive index is extracted, the comprehensive index is mixed with the comprehensive index extracted from the original twin data; the comprehensive index of the mixed running state data is taken as the input of the DNN neural network model and retrained.

[0039] In addition, the application also provides a device life and fault type prediction system based on digital twin technology, comprising:

[0040] The building module is used for building a physical device entity model and a digital twin virtual model, and correlating the physical device entity model and the digital twin virtual model through virtual mapping technology;

[0041] The twin data acquisition module is used for acquiring the historical running state data, fault data and device performance data corresponding to the twin data in the entity device through the digital twin virtual model in the building module;

[0042] The data preprocessing module is used for preprocessing the acquired running state data twin data and extracting the comprehensive index of the running state data by principal component analysis method; the device performance data includes availability, average trouble-free working time and average repair time; and the fault data includes fault type.

[0043] The specific parameter selection and critical value setting module selects a specific parameter indicating the end of the service life of the equipment from the twin data of the comprehensive index and the equipment performance data, and sets a critical value for the specific parameter; if the selected specific parameter indicating the end of the service life of the equipment is the comprehensive index, the first life prediction module is used for model prediction; if the selected specific parameter indicating the end of the service life of the equipment is the availability or the mean time between failures or the mean time to repair, the second life prediction module is used;

[0044] The first life prediction module is used for obtaining an estimated value of the comprehensive index by using the Kalman filtering algorithm and obtaining a time corresponding to the estimated value of the comprehensive index reaching the critical value, the time corresponding to the running time of the equipment, and obtaining the remaining life by combining the service life of the equipment;

[0045] The second life prediction module is used for calculating a time corresponding to the availability or the mean time between failures or the mean time to repair in the equipment performance data reaching the critical value, the time corresponding to the running time of the equipment, and obtaining the remaining life by combining the service life of the equipment; the specific calculation formulas are as follows:

[0046] R=t / T

[0047] In the formula, R represents the availability, t represents the working time, and T represents the planned working time.

[0048]

[0049] In the formula, MTBF represents the mean time between failures, t represents the working time, and f(t) represents the probability density function of the time elapsed until the next failure.

[0050]

[0051] In the formula, MTTR represents the mean time to repair, t represents the working time, and N represents the number of repairs.

[0052] Further, the method further comprises:

[0053] The failure type prediction module is used for extracting a comprehensive index set under different failure types of the equipment from the twin data acquisition module, dividing the set into a training set and a test set according to a proportion, constructing a DNN neural network model, the input layer of the DNN neural network model being the comprehensive index, the output layer being the failure type, and the formula of the DNN neural network model being:

[0054]

[0055] In the formula, h W,b (x) represents the output failure type; x i represents the running comprehensive index; W iis a weight; b is a bias; k is a characteristic quantity in the comprehensive index;

[0056] The DNN neural network model is trained by using the data of the training set, and the trained DNN neural network model is obtained when the training reaches a preset number of times; the test set data is input into the trained DNN neural network model for testing to obtain a prediction result of the fault model.

[0057] In addition, the application also provides a computer device, including a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the steps of any one of the above methods when executing the computer program. A computer readable storage medium has a computer program stored thereon, and the computer program is executed by the processor to implement the steps of any one of the above methods.

[0058] Beneficial effects: the equipment life and fault type prediction method based on digital twin technology provided by the application has the following advantages compared with the prior art: 1, the digital twin system is set, the current physical world parameters are directly mapped to the twin world to obtain twin data through simulation in real time; the residual life prediction result is obtained by combining the equipment motion state twin data and the Kalman filtering mode; considering different reference factors of different aspects of equipment life, adding equipment performance information factors such as equipment availability, average trouble-free working time and average repair time, the residual life prediction result is obtained through the twin data of the reference factors; 2, the principal component analysis is carried out on the motion state twin data to extract comprehensive index data, and the neural network is combined to obtain more accurate fault type analysis result. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 The flowchart of the method is shown;

[0060] Figure 2 The virtual mapping association modeling block diagram of the data twin equipment and the entity equipment is shown;

[0061] Figure 3 The neural network unit of the fault identification model is shown;

[0062] Figure 4 The principle diagram of the system is shown. DETAILED DESCRIPTION

[0063] The technical solutions of the application are further described below with reference to the drawings.

[0064] As Figure 1 The application provides an equipment life and fault type prediction method based on digital twin technology, which specifically includes the following contents:

[0065] Step one, build a physical equipment entity model and a digital twin virtual model, and associate the physical equipment entity model and the digital twin virtual model through virtual mapping technology.

[0066] (1) Collect data of physical equipment, build a physical equipment entity model, and the model is as follows:

[0067]

[0068] In the formula, PE represents the physical entity of the equipment, PS represents the physical space information of the equipment, including environmental information and physical parameters of the equipment; PC represents the equipment state data, which is obtained by intelligent acquisition equipment to obtain the vibration, temperature, pressure, current voltage, etc. of the equipment in operation; PO is the equipment basic information data, including the metadata of the equipment (service life on the factory specification, equipment parameters, etc.), equipment maintenance record information data, and equipment operation data (total output, production rate, yield, etc.). represents the natural connection between PS, PC, and PO, indicating the natural interaction relationship among the three. PS, PC, and PO are dynamic sets, and the set elements and their states are updated continuously with the dynamic operation of the manufacturing process.

[0069] (2) Build a digital twin virtual model, and the model is specifically as follows:

[0070]

[0071] In the formula, VE represents the digital twin of the equipment, DS represents the physical space information of the twin, and the environmental information is a simulative information element, while the physical information is based on the geometric size, physical structure, motion posture, motion characteristics, etc. of the equipment physical entity; DC is the equipment state data of the twin, which is obtained by implementing mapping or simulation; DO is the equipment twin basic information data, including the metadata inherited from the equipment entity and the operation data generated in the free growth process of the twin. represents the natural connection between DS, DC, and DO, indicating the natural interaction relationship among the three. DS, DC, and DO are also dynamic sets, and are updated continuously with the continuous growth of the twin.

[0072] (3) On the basis of establishing the physical space entity model PE and the information space twin model VE, further establish the virtual-real mapping association between the physical equipment entity model and the digital twin virtual model, and the specific mapping relationship model is as follows:

[0073]

[0074] In the formula, represents the bidirectional real mapping between the physical space entity model and the digital twin.

[0075] As Figure 2 shown, the whole virtual mapping association modeling includes simulation and free growth. The running status of the equipment is collected in the early stage of the equipment life cycle. According to the known basic data such as physical information, functional information and performance information of the equipment, based on the clear space environment, the equipment is digitally mapped by using cloud computing, Internet of Things, three-dimensional modeling and graphic rendering technology, the mirror image of the physical entity is generated in the virtual space, the physical simulation process is reduced, some characteristics and parameters of the physical simulation are extracted, and the simulation running situation of the physical entity is presented. The actual data of the equipment generated by the external sensor and the central processor is taken as the input of the twin body for simulation. The simulation process is to directly map the current physical world parameters to the twin world. Free generation refers to the digital twin body prototype generated based on simulation. By using industrial big data and artificial intelligence (AI) machine learning technology, the construction of the twin world is realized.

[0076] Step two, according to the entity model and digital twin model established above, the corresponding twin data of the historical running state data, fault data and equipment performance data in the entity equipment are obtained respectively. The performance data of the equipment mainly includes the availability of the equipment, the average trouble-free working time and the average repair time; the fault data includes the fault type of the specific equipment.

[0077] Since the predictive diagnosis of the equipment through the digital twin is a prediction method for the whole life cycle of the equipment, in the case of only having a small amount of samples, the equipment data needs to be preprocessed to obtain effective parameters for model training from discrete data. Specifically,

[0078] (1) The twin data of the obtained running state data is preprocessed; including:

[0079] (1.1) Removing outliers

[0080] The obtained various equipment state parameters are respectively subjected to preliminary data cleaning, and the outliers are removed after data correction.

[0081] (1.2) Time domain feature extraction

[0082] The statistical characteristic quantities related to time domain feature extraction include but are not limited to mean, standard deviation, root mean square, kurtosis, waveform factor, kurtosis factor and pulse factor.

[0083] (1.3) Frequency domain feature extraction

[0084] In order to improve the sensitivity and stability of the equipment predictability maintenance, in addition to the time domain features, Fourier transform is used for frequency domain feature extraction.

[0085]

[0086]

[0087] In the formula, x(t) represents a signal, X(f) represents a spectrum after FFT of x(t), FFT(i) represents a root mean square value of the i-th sub-band after FFT, FFT(m) represents the m-th spectral line of the i-th sub-band, and N represents a number of spectral lines. rms

[0088] (1.4) A principal component analysis method is used to extract a comprehensive index of the operating state data.

[0089] (1.4.1) The preprocessed data is converted into a feature vector matrix.

[0090] (1.4.2) The average value of each column feature is calculated, and then the average value of the column feature needs to be subtracted in each dimension.

[0091] (1.4.3) The covariance matrix of the feature is calculated.

[0092] (1.4.4) The eigenvalue and eigenvector of the covariance matrix are calculated.

[0093] (1.4.5) The calculated eigenvalues are sorted from large to small.

[0094] (1.4.6) The first K eigenvectors and eigenvalues are taken out, and a reduced feature matrix is obtained by backtracking, and K is a positive integer.

[0095] Step three, constructing an RUL prediction model for life prediction

[0096] (1) A specific parameter indicating the end of the life of the device is selected from the twin data of the comprehensive index and the device performance data, and a critical value CV of the specific parameter is set. i .

[0097] If it is considered that the comprehensive index can better reflect the characteristics of the device life, that is, the selected specific parameter indicating the end of the life of the device is any index in the comprehensive index, then go to step four.

[0098] If it is considered that the parameters in the twin data of the device performance data can better reflect the characteristics of the device life, that is, the selected specific parameter indicating the end of the life of the device is one or a combination of the availability, the average trouble-free operating time, and the average repair time, then go to step five.

[0099] Step four, predicting the remaining life of the device using a Kalman filter algorithm

[0100] (1) First, the parameter measurement value of the specific parameter is obtained, and the formula is:

[0101] ​Z(k) = HX(k) + V(k)

[0102] In the formula, Z(k) is the parameter measurement value of a specific parameter; X(k) is the state value at time k; V(k) is measurement noise; H is the parameter of the measurement system;

[0103] (2) Obtain the state quantity estimation value according to the parameter measurement value, and the formula is:

[0104] X(k|k) = X(k|k-1) + Kg(k)(Z(k) - HX(k|k-1))

[0105] In the formula, X(k|k) is the estimated value of the current state; X(k|k-1) is the estimated value of the previous state; H is the parameter of the measurement system; Kg(k) is the gain factor;

[0106] (3) Compare the state quantity estimation value with the critical value CV i , when the state quantity estimation value X(k|k) reaches the critical value CV i , it is determined that the equipment life ends, and the k value corresponding to the state quantity estimation value is the period of equipment operation, and the remaining life of the equipment is obtained by subtracting the period of operation from the service life of the equipment.

[0107] Step five, according to the selected specific parameter, if the selected characteristic parameter is the availability, the equipment has been running time is calculated according to the formula, and the remaining life is obtained combined with the service life of the equipment; the formula is:

[0108] R = t / T

[0109] In the formula, R represents availability; t represents working time; T represents planned working time;

[0110] If the selected characteristic parameter is the average failure-free working time, the equipment has been running time is calculated according to the formula, and the remaining life is obtained combined with the service life of the equipment; the formula is:

[0111]

[0112] In the formula, MTBF represents the average failure-free working time; t represents working time; f(t) represents the probability density function of the time elapsed until the next failure;

[0113] If the selected characteristic parameter is the average repair time, the equipment has been running time is calculated according to the formula, and the remaining life is obtained combined with the service life of the equipment; the formula is:

[0114]

[0115] In the formula, MTTR represents the average repair time; t represents working time; N is the number of repairs.

[0116] Step 6: Construct a fault type judgment model

[0117] (1) Based on the digital twin virtual model, extract the twin data of the operating status of several devices under different fault types, and obtain the comprehensive index under different fault types through the above principal component analysis. The comprehensive indexes corresponding to different fault types are used to form a training set, and the set is divided into training set and test set in a 7:3 ratio.

[0118] (2) Construct a DNN neural network model, such as Figure 3 As shown, the input layer of the DNN neural network model is a comprehensive index, and the output layer is the fault type. The formula for the DNN neural network model is:

[0119]

[0120] In the formula, h W,b (x) Output fault type; x i To achieve comprehensive operational indicators; W i b is the weight; b is the bias; k is the feature quantity in the comprehensive index;

[0121] (3) Use the training set data to train the DNN neural network model. When the training reaches the preset number of times N, obtain the trained DNN neural network model; input the test set data into the trained DNN neural network model to test and obtain the prediction results of the fault model.

[0122] Furthermore, if the proportion of inconsistencies between the fault prediction model obtained from the test set data and the actual model reaches a certain value, the accuracy of the fault prediction model is considered low, and the model needs to be retrained. Historical operating status data from the physical equipment is preprocessed, and a comprehensive index is extracted. This comprehensive index is then mixed with the comprehensive index extracted from the original twin data. The comprehensive index of the mixed operating status data is used as input to the DNN neural network model for retraining.

[0123] Specifically, this invention takes ticket gates as an example, using raw data sources including historical operating data, fault data, and basic equipment information of the gates in use. The statistically analyzed fault types of the ticket gates include: motor damage, fan damage, swing door failure, power supply failure, and mainboard failure. Using the principal component analysis method described in this invention, temperature, equipment power, motor torque, and door opening / closing time are selected as comprehensive indicators. These comprehensive indicators are compared with performance indicators, and combined with the lifespan parameters of the gate equipment itself, the mean time between failures (MTBF) is selected as the parameter marking the end of the gate's lifespan, with a calibrated value of 80,000 hours. The prediction results for five in-use gates using the method described in this invention at a certain point in time are shown in Table 1 below.

[0124] Table 1 prediction results

[0125] Gate A Gate B Gate C Gate D Gate E Fault type - Swing door failure - - - Remaining life (h) 89762 83641 87695 90352 88634

[0126] In addition, as Figure 4 The application also provides a device life and fault type prediction system based on digital twin technology, comprising: a building module for building a physical device entity model and a digital twin virtual model, and correlating the physical device entity model and the digital twin virtual model through virtual mapping technology; a twin data acquisition module for acquiring twin data corresponding to historical running state data, fault data and device performance data in the entity device through the digital twin virtual model in the building module; a data preprocessing module for preprocessing the acquired twin data of the running state data and extracting comprehensive indexes of the running state data using principal component analysis; the device performance data includes availability, average trouble-free working time and average repair time; the fault data includes fault type; a specific parameter selection and critical value setting module for selecting a specific parameter indicating the end of the device life from the twin data of the comprehensive indexes and the device performance data, and setting a critical value for the specific parameter; if the selected specific parameter indicating the end of the device life is the comprehensive index, a first life prediction module is used for model prediction; if the selected specific parameter indicating the end of the device life is the availability or the average trouble-free working time or the average repair time, a second life prediction module is used; the first life prediction module is used for acquiring an estimated value of the comprehensive index using Kalman filtering algorithm and acquiring a time corresponding to the estimated value of the comprehensive index reaching the critical value, which corresponds to the running time of the device, and combining the service life of the device to obtain the remaining life; the second life prediction module is used for calculating a time corresponding to the availability or the average trouble-free working time or the average repair time in the device performance data reaching the critical value, which corresponds to the running time of the device, and combining the service life of the device to obtain the remaining life; the specific calculation formulas are as follows:

[0127] R=t / T

[0128] In the formula, R represents availability; t represents working time; T represents planned working time;

[0129]

[0130] In the formula, MTBF represents average trouble-free working time; t represents working time; f(t) represents probability density function of time elapsed until next failure;

[0131]

[0132] In the formula, MTTR represents average repair time; t represents working time; N is the number of repairs.

[0133] The fault type prediction module is used for forming a set of comprehensive indexes of the equipment in different fault types in the twin data acquisition module, dividing the set into a training set and a test set in proportion, constructing a DNN neural network model, the input layer of the DNN neural network model being the comprehensive indexes, the output layer being the fault types, and the formula of the DNN neural network model being:

[0134]

[0135] In the formula, h W,b (x) is the output fault type; x i is the running comprehensive index; W i is the weight; b is the bias; k is the characteristic quantity in the comprehensive index; the data of the training set is used to train the DNN neural network model, the trained DNN neural network model is obtained when the training reaches a preset number of times, the test set data is input into the trained DNN neural network model for testing to obtain the prediction result of the fault model.

[0136] The system can acquire equipment data, predict data, store, calculate and process twin data. The system can rely on big data analysis and processing and visualization related technologies to build an equipment full life cycle comprehensive management platform, based on the equipment full life cycle history, the equipment twin model is displayed, and management means such as centralized management, real-time monitoring and timely repair are provided. The functions of the comprehensive management system include but are not limited to equipment management, emergency repair, planned maintenance, statistical analysis and board management, and an integrated interactive platform is provided to support high-level decision-making, middle-level control and basic-level operation.

Claims

1. A method for predicting the life and failure type of a device based on digital twin technology, characterized by, Specifically comprising the following steps: (1) build a physical equipment entity model and a digital twin virtual model, and associate the physical equipment entity model with the digital twin virtual model through virtual mapping technology; (2) obtain the twin data corresponding to the historical running state data, fault data and equipment performance data of the entity equipment; preprocess the obtained twin data of the running state data and extract the comprehensive index of the running state data by using principal component analysis; the equipment performance data includes availability, average trouble-free working time and average repair time; the fault data includes fault type; (3) select a specific parameter indicating the end of the life of the equipment from the comprehensive index and the twin data of the equipment performance data, and set a critical value for the specific parameter; if the selected specific parameter indicating the end of the life of the equipment is the comprehensive index, go to step (4); if the selected specific parameter indicating the end of the life of the equipment is the availability or the average trouble-free working time or the average repair time, go to step (5); (4) obtain the estimated value of the comprehensive index by using Kalman filtering algorithm and obtain the time corresponding to the critical value of the estimated value of the comprehensive index, which corresponds to the running time of the equipment, and obtain the remaining life by combining the service life of the equipment; (5) calculate the time corresponding to the critical value of the availability or the average trouble-free working time or the average repair time in the equipment performance data, which corresponds to the running time of the equipment, and obtain the remaining life by combining the service life of the equipment; the specific calculation formulas are as follows: R=t / T In the formula, R represents the availability; t represents the working time; T represents the planned working time; In the formula, MTBF represents the average trouble-free working time; t represents the working time; f(t) represents the probability density function of the time elapsed until the next failure; In the formula, MTTR represents the average repair time; t represents the working time; N is the number of repairs. 2.The digital twin technology-based equipment life and failure type prediction method of claim 1, wherein, Step (4) specifically comprises the following steps: (4.1) obtain the parameter measurement value of the specific parameter, and the formula is: Z(k)=HX(k)+V(k) In the formula, Z(k) is the parameter measurement value of the specific parameter; X(k) is the state value at time k; V(k) is the measurement noise; H is the parameter of the measurement system; (4.2) obtain the state quantity estimate value according to the parameter measurement value, and the formula is: X(k|k)=X(k|k-1)+Kg(k)(Z(k)-HX(k|k-1)) In the formula, X(k|k) is the estimated value of the current state; X(k|k-1) is the estimated value of the previous state; H is the parameter of the measurement system; Kg(k) is the gain factor; (4.3) compare the state quantity estimate value with the critical value, when the state quantity estimate value reaches the critical value, it is determined that the life of the equipment ends, and the k value corresponding to the state quantity estimate value is the running period of the equipment, and the remaining life of the equipment is obtained by subtracting the running period from the service life of the equipment. 3.The device life and failure type prediction method based on digital twin technology according to claim 1, characterized in that, Also includes: (6) According to the comprehensive index set under different fault types of the twin data extraction device in step (2), the set is divided into a training set and a test set in proportion; a DNN neural network model is constructed, the input layer of the DNN neural network model is the comprehensive index, the output layer is the fault type, and the formula of the DNN neural network model is: where h W,b (x) outputted fault type; x i is a running aggregate; W i is a weight; b is a bias; k is a characteristic quantity in the comprehensive index; The DNN neural network model is trained using the data of the training set, and the trained DNN neural network model is obtained when the training reaches a preset number of times; the test set data is input into the trained DNN neural network model for testing to obtain the prediction result of the fault model. 4.The device life and failure type prediction method based on digital twin technology according to claim 1, wherein, The twin data of the obtained running state data in step (2) is preprocessed and the comprehensive index of the running state data is extracted by principal component analysis, which specifically includes: (2.1) The preprocessing of the obtained twin data of the running state data includes removing outliers, time domain feature extraction and frequency domain feature extraction; (2.2) The comprehensive index of the running state data is extracted by principal component analysis, which specifically includes: (2.2.1) The preprocessed data is converted into a feature vector matrix; (2.2.2) The average value of each column feature is calculated, and then each dimension needs to be subtracted from the average value of the column feature; (2.2.3) Calculate the covariance matrix of the feature; (2.2.4) Calculate the eigenvalue and eigenvector of the covariance matrix; (2.2.5) Sort the calculated eigenvalues from large to small; (2.2.6) Take out the first K eigenvectors and eigenvalues, and perform rollback to obtain the reduced feature matrix. 5.The device life and failure type prediction method based on digital twin technology according to claim 2, characterized in that, Step (6) further includes: if the proportion of the prediction result of the fault model obtained by the test set data and the actual model is inconsistent reaches a certain value, the historical running state data in the entity device is preprocessed and the comprehensive index is extracted, the comprehensive index is mixed with the comprehensive index extracted from the original twin data; the comprehensive index of the mixed running state data is used as the input of the DNN neural network model for retraining.

6. A system for predicting the life and failure type of a device based on digital twin technology, characterized by, Including: Building module: used for building physical device entity model and digital twin virtual model, and associating physical device entity model and digital twin virtual model through virtual mapping technology; Twin data acquisition module; used for acquiring historical running state data, fault data and device performance data corresponding to the twin data in the entity device through the digital twin virtual model in the building module; Data preprocessing module: used for preprocessing the twin data of the obtained running state data and extracting the comprehensive index of the running state data by principal component analysis; the device performance data includes availability, average trouble-free working time and average repair time; the fault data includes fault type; The specific parameter selection and threshold setting module selects a specific parameter indicating the end of the life of the equipment from the twin data of the comprehensive index and the equipment performance data, and sets a threshold value for the specific parameter; if the selected specific parameter indicating the end of the life of the equipment is a comprehensive index, a first life prediction module is used for model prediction; if the selected specific parameter indicating the end of the life of the equipment is an availability or an average failure-free operating time or an average repair time, a second life prediction module is used; The first life prediction module is used to obtain an estimated value of the comprehensive index by using a Kalman filtering algorithm and obtain a time corresponding to the estimated value of the comprehensive index reaching the threshold value, which corresponds to the operating time of the equipment, and obtain the remaining life by combining the service life of the equipment; The second life prediction module is used to calculate a time corresponding to the availability or the average failure-free operating time or the average repair time in the equipment performance data reaching the threshold value, which corresponds to the operating time of the equipment, and obtain the remaining life by combining the service life of the equipment; the specific calculation formulas are as follows: R=t / T In the formula, R represents the availability, t represents the operating time, and T represents the planned operating time. In the formula, MTBF represents the average failure-free operating time, t represents the operating time, and f(t) represents the probability density function of the time elapsed until the next failure. In the formula, MTTR represents the average repair time, t represents the operating time, and N represents the number of repairs. 7.The digital twin technology based equipment life and failure type prediction system as claimed in claim 5, wherein, Further comprising: The fault type prediction module is used to obtain a comprehensive index set under different fault types of the equipment from the twin data acquisition module, divide the set into a training set and a test set according to a proportion, construct a DNN neural network model, the input layer of the DNN neural network model is the comprehensive index, the output layer is the fault type, and the formula of the DNN neural network model is as follows: where h W,b (x) outputted fault type; x i is a running aggregate; W i is a weight; b is a bias k is a feature quantity in the comprehensive index. The data of the training set is used to train the DNN neural network model, and when the training reaches a preset number of times, a trained DNN neural network model is obtained; the test set data is input 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, characterized in that, The processor executes the computer program to realize the steps of the method of any one of claims 1 to 5.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method of any one of claims 1 to 5.

Images