A fault prediction method, device, terminal and medium based on digital twin

Abstract

The present application discloses a fault prediction method, device, terminal and medium based on digital twins. The method includes the following steps: three-dimensional modeling of the equipment; using the three-dimensional model of the equipment, constructing a digital twin model of the equipment, constructing a simulation program in the digital twin model to simulate the equipment operation status, and collecting the equipment operation data generated during the simulation process; constructing several fault simulation programs in the digital twin model to simulate the fault state of the equipment, and collecting the equipment operation data generated during the fault simulation process; collecting the equipment operation data during the actual operation of the equipment, and predicting the equipment failure by comparing it with the equipment operation data generated during the fault simulation process. The present application simulates the equipment operation status through the digital twin model and collects data during the simulation process. Based on this data, it is beneficial to improve the ability and accuracy of equipment fault prediction.

Description

Technical Field

[0001] The present application relates to the field of fault prediction technology, and in particular to a fault prediction method, device, terminal and medium based on digital twins. Background Art

[0002] In daily production and life, equipment failures such as elevators, lifts, vehicles, forklifts, etc. occur frequently, which not only seriously affects daily production and life, but may also cause serious safety accidents.

[0003] Currently, monitoring and fault prediction for these devices rely on the collection of equipment failure data. However, the vast majority of data collected during operation is normal, resulting in a relatively small amount of data strongly correlated with equipment failures. This leads to a low level of fault prediction capability. Often, the system can only predict a fault after it is about to occur or has already occurred. This lack of advance prediction makes it difficult to effectively guide equipment maintenance and prevent failures. Summary of the Invention

[0004] The present application provides a fault prediction method, device, terminal and medium based on digital twins, which has the advantage of improving the fault prediction capability of equipment.

[0005] The technical solution of this application is as follows:

[0006] In one aspect, the present application provides a fault prediction method based on digital twins, comprising the following steps:

[0007] S1: 3D modeling of the equipment;

[0008] S2: Use the 3D model of the device to build a digital twin model of the device. In the digital twin model, build a simulation program to simulate the device's operating status and collect the device operation data generated during the simulation process;

[0009] S3: In the digital twin model, build several fault simulation programs to simulate the fault status of the equipment and collect the equipment operation data generated during the fault simulation process;

[0010] S4: Collecting equipment operation data during the actual operation of the equipment, and predicting equipment failure by comparing it with the data obtained in step S3.

[0011] Furthermore, in steps S2 and S3, the Unity3D engine is used to build a simulation program and a fault simulation program, and the three-dimensional model of the equipment is given material mapping and physical properties, including lighting effects, quality, strength, collision properties and a real kinematic model.

[0012] Furthermore, in step S2, the simulation program is used to simulate the operating state of the device during normal operation;

[0013] In step S3, the fault simulation program is used to simulate the operating state of the device when one or more faults occur.

[0014] Furthermore, step S4 specifically includes:

[0015] According to step S3, data is obtained, and feature extraction is performed on the equipment operation data when the equipment fails to obtain a first data feature;

[0016] Extracting features of the device operation data during the actual operation of the device to obtain a second data feature;

[0017] The consistency of the second data feature and the first data feature is detected in real time, and when the consistency of the second data feature and the first data feature reaches a preset threshold, it is determined that the device has failed.

[0018] Furthermore, the method further includes step S5: building a fault prediction model based on a deep learning algorithm, using the data obtained in step S3 to train the fault prediction model, and predicting faults of the equipment using the trained fault prediction model;

[0019] The fault prediction model is based on a convolutional neural network algorithm and includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer. The operating data of the equipment at the time of failure obtained in step S3 is used as input, and the fault type is used as output to train the model.

[0020] Furthermore, the method further includes step S6:

[0021] S61: According to the data obtained in steps S2 and S3, the fault type and the device data are correlated to obtain the fault type. Related device data ;

[0022] The process of correlating the fault type with the device data includes the following steps:

[0023] Based on the data obtained in steps S2 and S3, a correlation analysis is performed on the individual fault types and device data, and the fault types and device data whose correlation is greater than a preset threshold are associated;

[0024] S62: Collect historical operation data of the equipment and extract equipment data associated with the fault type ;

[0025] S63: Traverse the device data of each natural day in the historical data , computing device data The extreme value in each natural day and , d represents natural day;

[0026] S64: Calculate the device data for several natural days Middle extreme value and The mean and variance of 、 、 and ; obtained by curve fitting 、 、 and The changing rules of 、 、 and The time it takes to change to a dangerous interval value is used to predict the time of failure;

[0027] The 、 、 and The dangerous interval value of is determined by the following method: using the data obtained in S3, calculate the extreme value of the equipment data when the equipment fails, and calculate the mean and variance of the extreme value as 、 、 and The risk interval value.

[0028] Furthermore, step S7 is also included: collecting the operating data of the equipment in real time, and importing it into the digital twin model for real-time updating of the digital twin model, and visually displaying the operating status of the equipment through the digital twin model.

[0029] In another aspect, the present application provides a fault prediction device based on digital twins, comprising:

[0030] A model building unit, used for three-dimensional modeling of the equipment;

[0031] The digital twin unit is used to construct a digital twin model of the device using the three-dimensional model of the device. Within the digital twin model, a simulation program is constructed to simulate the device's operating status and collect the device operating data generated during the simulation process. Within the digital twin model, several fault simulation programs are constructed to simulate the device's fault status and collect the device operating data generated during the fault simulation process.

[0032] A data acquisition unit collects the operating status data of the device using the device host and sensors arranged on the device;

[0033] The first fault prediction unit collects equipment operation data during actual operation of the equipment and predicts equipment failure by comparing it with the equipment operation data obtained by the digital twin unit at the time of equipment failure;

[0034] The second fault prediction unit is used to build a fault prediction model based on a deep learning algorithm, use the data obtained by the digital twin unit to train the fault prediction model, and use the trained fault prediction model to predict equipment faults;

[0035] The third fault prediction unit is configured to perform fault prediction on the device according to the following steps:

[0036] According to the data obtained by the digital twin unit, the fault type and equipment data are correlated to obtain the fault type. Related device data ;

[0037] Collect historical operating data of the equipment and extract equipment data associated with the fault type ;

[0038] Traverse the device data of each natural day in the historical data , computing device data The extreme value in each natural day and , d represents natural day;

[0039] Calculate the device data for several natural days in history Middle extreme value and The mean and variance of 、 、 and ; obtained by curve fitting 、 、 and The changing rules of 、 、 and The time it takes to change to a dangerous interval value is used to predict the time of occurrence of a fault; 、 、 and The dangerous interval value of is determined by the following method: using the data obtained in S3, calculate the extreme value of the equipment data when the equipment fails, and calculate the mean and variance of the extreme value as 、 、 and The risk interval value of

[0040] And a visual management and control platform to visually display the operating status of the equipment.

[0041] On the other hand, the present application provides a fault prediction terminal, including a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is called and executed by the processor, the method described above is implemented.

[0042] On the other hand, the present application provides a computer-readable medium, wherein the computer-readable medium stores a computer program, and when the computer program is called and executed by a computer, the method described above is implemented.

[0043] In summary, the beneficial effects of this application are:

[0044] 1. This application uses a digital twin model to simulate the operating status of equipment and collect data from the simulation process, including equipment operating data at the time of failure. This data can be used to determine whether the equipment has failed through data comparison and to train a fault prediction model. Increasing the amount of training data improves model training effectiveness, which is conducive to improving the ability and accuracy of equipment fault prediction.

[0045] 2. This application screens historical equipment operating data for data associated with equipment failures, calculates the extreme values ​​of the data for each calendar day, and the mean and variance of these extreme values. Using curve fitting, it identifies patterns in the historical equipment operating data. This enables long-term failure prediction, effectively improving the foresight of predictions and providing a data basis for equipment maintenance. Converting massive amounts of historical equipment data into daily extreme values ​​not only reduces the amount of data, but also allows the daily extreme values ​​to better reflect extreme operating conditions of the equipment, characterizing equipment aging. This can then be used to predict the moment of equipment failure, achieving the goal of forward-looking failure prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is the system architecture diagram of this application;

[0047] Figure 2 This is the digital twin model diagram of this application;

[0048] Figure 3 It is the data flow diagram in this application;

[0049] Figure 4 This is a typical data diagram in this application. DETAILED DESCRIPTION

[0050] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0051] A specific embodiment of the present application provides a fault prediction method based on digital twins for predicting equipment faults. The equipment described in this embodiment includes special equipment such as elevators, escalators, lifts, forklifts, etc. Figure 2 , the method comprises the following steps:

[0052] S1: 3D modeling of the equipment. Taking the elevator as an example, this step creates a 1:1 3D model of the elevator and the building environment, including details such as the building, elevator shaft, and various elevator components, to ensure high-precision restoration of the model.

[0053] S2: Use the three-dimensional model of the equipment to build a digital twin model of the equipment. In the digital twin model, build a simulation program to simulate the equipment operating status and collect the equipment operating data generated during the simulation process; the simulation program is used to simulate the operating status of the equipment during normal operation.

[0054] S3: Within the digital twin model, build several fault simulation programs to simulate equipment failure states and collect equipment operating data generated during the simulation. These programs simulate the equipment's operating state when experiencing one or more faults. These programs simulate typical equipment failure types. For elevators, typical failure types include emergency stops, door opening and closing failures, hitting the top, and hitting the bottom.

[0055] The 3D model was imported using the Unity3D simulation engine and assigned material mapping and physical properties, including lighting effects, mass, strength, collision properties, and realistic kinematics, ensuring that the model behaves in the virtual environment consistent with its real-world behavior. Simulation and fault simulation programs were built using the Unity3D engine, using C# scripting to programmatically simulate different faults. The Unity3D simulation engine also supports multi-person online collaboration, facilitating multi-departmental collaboration in elevator supervision and maintenance.

[0056] S4: Collect equipment operating data during actual operation and compare it with the data obtained in step S3 to predict equipment failures. For elevators, the collected data includes various types of data (e.g., static and dynamic) about the elevator and its operating environment, including but not limited to basic elevator attributes, historical operating data, historical failure data, building environment data (e.g., architectural drawings, 3D point cloud data), meteorological data (e.g., temperature, humidity), historical elevator operation data (e.g., cumulative number of elevator runs, cumulative duration), and real-time elevator operation data (e.g., direction of travel, floor, door opening and closing, presence of occupants, alarm buttons, etc.). By comparing real-time monitoring data with simulation results, potential failure risks can be promptly identified and early warnings issued to operation and maintenance personnel.

[0057] Specifically, according to step S3, data is obtained, and feature extraction is performed on the device operation data when the device fails to obtain a first data feature;

[0058] Extracting features of the device operation data during the actual operation of the device to obtain a second data feature;

[0059] The consistency of the second data feature and the first data feature is detected in real time, and when the consistency of the second data feature and the first data feature reaches a preset threshold, it is determined that the device has failed.

[0060] Specifically, for numerical data such as voltage and temperature, a difference method is used to determine consistency. When the difference reaches a preset threshold, the two are considered inconsistent. For vibration signals, feature extraction extracts the signal's frequency domain features, and consistency is determined by comparing these features.

[0061] S5: Build a fault prediction model based on a deep learning algorithm, use the data obtained in step S3 to train the fault prediction model, and use the trained fault prediction model to predict equipment faults;

[0062] The fault prediction model is based on a convolutional neural network algorithm and includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer. The operating data of the equipment at the time of failure obtained in step S3 is used as input, and the fault type is used as output to train the model.

[0063] Step S6:

[0064] S61: According to the data obtained in steps S2 and S3, the fault type and the device data are correlated to obtain the fault type. Related device data ;

[0065] The process of correlating the fault type with the device data includes the following steps:

[0066] Based on the data obtained in steps S2 and S3, a correlation analysis is performed on the individual fault types and device data, and the fault types and device data whose correlation is greater than a preset threshold are associated;

[0067] S62: Collect historical operation data of the equipment and extract equipment data associated with the fault type ;

[0068] S63: Traverse the device data of each natural day in the historical data , computing device data The extreme value in each natural day and , d represents natural day;

[0069] S64: Calculate the device data for several natural days Middle extreme value and The mean and variance of 、 、 and ; obtained by curve fitting 、 、 and The changing rules of 、 、 and The time it takes to change to a dangerous interval value is used to predict the time of failure;

[0070] The 、 、 and The dangerous interval value of is determined by the following method: using the data obtained in S3, calculate the extreme value of the equipment data when the equipment fails, and calculate the mean and variance of the extreme value as 、 、 and The risk interval value.

[0071] Step S7: Real-time data collection of equipment operation is imported into the digital twin model for real-time updating. The digital twin model provides a visual display of the equipment's operating status. Elevator operation status and fault warning information are intuitively displayed through charts, animations, and other forms, improving supervision efficiency and decision-making support capabilities.

[0072] Another embodiment of the present application provides a fault prediction device based on digital twins, such as Figure 1 and 3 ,include:

[0073] A model building unit, used for three-dimensional modeling of the equipment;

[0074] The digital twin unit is used to construct a digital twin model of the device using the three-dimensional model of the device. Within the digital twin model, a simulation program is constructed to simulate the device's operating status and collect the device operating data generated during the simulation process. Within the digital twin model, several fault simulation programs are constructed to simulate the device's fault status and collect the device operating data generated during the fault simulation process.

[0075] The data acquisition unit uses the device host and sensors arranged on the equipment to collect the operating status data of the equipment; the data acquisition unit is responsible for obtaining various types of data on the elevator and its operating environment (such as static data and dynamic data), including but not limited to the basic attribute data of the elevator, historical working data, historical fault data, building environment data (such as architectural drawing data, three-dimensional point cloud data, etc.), meteorological environment data (such as temperature, humidity, etc.), historical elevator operation data (such as the cumulative number of elevator operations, cumulative duration, etc.) and real-time elevator operation data (such as the direction of elevator operation, floor, door opening and closing action, whether there is someone, alarm button, etc.).

[0076] The data acquisition unit is connected to the data processing unit and is used to clean, integrate and pre-process the collected data. The data acquisition unit and the data processing unit constitute the Internet of Things platform.

[0077] The first fault prediction unit collects equipment operation data during actual operation of the equipment and predicts equipment failure by comparing it with the equipment operation data obtained by the digital twin unit at the time of equipment failure;

[0078] The second fault prediction unit is used to build a fault prediction model based on a deep learning algorithm, use the data obtained by the digital twin unit to train the fault prediction model, and use the trained fault prediction model to predict equipment faults;

[0079] The third fault prediction unit is configured to perform fault prediction on the device according to the following steps:

[0080] According to the data obtained by the digital twin unit, the fault type and equipment data are correlated to obtain the fault type. Related device data ;

[0081] Collect historical operating data of the equipment and extract equipment data associated with the fault type ;

[0082] Traverse the device data of each natural day in the historical data , computing device data The extreme value in each natural day and , d represents natural day;

[0083] Calculate the device data for several natural days in history Middle extreme value and The mean and variance of 、 、 and ; obtained by curve fitting 、 、 and The changing rules of 、 、 and The time it takes to change to a dangerous interval value is used to predict the time of occurrence of a fault; 、 、 and The dangerous interval value of is determined by the following method: using the data obtained in S3, calculate the extreme value of the equipment data when the equipment fails, and calculate the mean and variance of the extreme value as 、 、 and The risk interval value of

[0084] And a visual management and control platform to visually display the operating status of the equipment.

[0085] Another embodiment of the present application provides a fault prediction terminal, including a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is called and executed by the processor, the method described above is implemented.

[0086] Another embodiment of the present application provides a computer-readable medium, wherein the computer-readable medium stores a computer program, and when the computer program is called and executed by a computer, the method described above is implemented.

[0087] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this field, several variations and improvements can be made without departing from the creative concept of the present application, and these all fall within the scope of protection of the present application.

Claims

1. A fault prediction method based on digital twins, characterized in that: The following steps are involved: S1: 3D modeling of the equipment; S2: Use the 3D model of the device to build a digital twin model of the device. In the digital twin model, build a simulation program to simulate the device's operating status and collect the device operation data generated during the simulation process; S3: In the digital twin model, build several fault simulation programs to simulate the fault status of the equipment and collect the equipment operation data generated during the fault simulation process; S4: Collecting equipment operation data during actual operation of the equipment, and predicting equipment failure by comparing it with the data obtained in step S3; S5: Build a fault prediction model based on a deep learning algorithm, use the data obtained in step S3 to train the fault prediction model, and use the trained fault prediction model to predict equipment faults; The step S6 is also included: S61: According to the data obtained in steps S2 and S3, the fault type and the device data are correlated to obtain the fault type. Related device data ; S62: Collect historical operation data of the equipment and extract equipment data associated with the fault type ; S63: Traverse the device data of each natural day in the historical data , computing device data The extreme value in each natural day and , d represents natural day; S64: Calculate the device data for several natural days Middle extreme value and The mean and variance of 、 、 and ; obtained by curve fitting 、 、 and The changing rules of 、 、 and The time it takes to change to a dangerous interval value is used to predict the time of failure; The 、 、 and The dangerous interval value of is determined by the following method: using the data obtained in S3, calculate the extreme value of the equipment data when the equipment fails, and calculate the mean and variance of the extreme value as 、 、 and The risk interval value.

2. The fault prediction method based on digital twin according to claim 1, characterized in that: In steps S2 and S3, the Unity3D engine is used to build a simulation program and a fault simulation program, and the three-dimensional model of the device is given material mapping and physical properties, including light and shadow effects, mass, strength, collision properties and a real kinematic model.

3. The fault prediction method based on digital twin according to claim 2, characterized in that: In step S2, the simulation program is used to simulate the operating state of the device during normal operation; In step S3, the fault simulation program is used to simulate the operating state of the device when one or more faults occur.

4. The fault prediction method based on digital twin according to claim 1, characterized in that: Step S4 specifically includes: According to step S3, data is obtained, and feature extraction is performed on the equipment operation data when the equipment fails to obtain a first data feature; Extracting features of the device operation data during the actual operation of the device to obtain a second data feature; The consistency of the second data feature and the first data feature is detected in real time, and when the consistency of the second data feature and the first data feature reaches a preset threshold, it is determined that the device has failed.

5. The fault prediction method based on digital twin according to claim 1, characterized in that: The fault prediction model is based on a convolutional neural network algorithm and includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer. The operating data of the equipment at the time of failure obtained in step S3 is used as input, and the fault type is used as output to train the model.

6. The fault prediction method based on digital twin according to claim 1, characterized in that: In step S61, the process of correlating the fault type with the device data includes the following steps: Based on the data obtained in steps S2 and S3, correlation analysis is performed on individual fault types and device data, and fault types and device data having correlations greater than a preset threshold are associated.

7. The fault prediction method based on digital twin according to claim 1, characterized in that: It also includes step S7: collecting the operating data of the equipment in real time and importing it into the digital twin model for real-time updating of the digital twin model, and visually displaying the operating status of the equipment through the digital twin model.

8. A fault prediction device based on digital twins, characterized in that: include: A model building unit, used for three-dimensional modeling of the equipment; The digital twin unit is used to construct a digital twin model of the device using the three-dimensional model of the device. Within the digital twin model, a simulation program is constructed to simulate the device's operating status and collect the device operating data generated during the simulation process. Within the digital twin model, several fault simulation programs are constructed to simulate the device's fault status and collect the device operating data generated during the fault simulation process. A data acquisition unit collects the operating status data of the device using the device host and sensors arranged on the device; The first fault prediction unit collects equipment operation data during actual operation of the equipment and predicts equipment failure by comparing it with the equipment operation data obtained by the digital twin unit at the time of equipment failure; The second fault prediction unit is used to build a fault prediction model based on a deep learning algorithm, use the data obtained by the digital twin unit to train the fault prediction model, and use the trained fault prediction model to predict equipment faults; The third fault prediction unit is configured to perform fault prediction on the device according to the following steps: According to the data obtained by the digital twin unit, the fault type and equipment data are correlated to obtain the fault type. Related device data ; Collect historical operating data of the equipment and extract equipment data associated with the fault type ; Traverse the device data of each natural day in the historical data , computing device data The extreme value in each natural day and , d represents natural day; Calculate the device data for several natural days in history Middle extreme value and The mean and variance of 、 、 and ; obtained by curve fitting 、 、 and The changing rules of 、 、 and The time it takes to change to a dangerous interval value is used to predict the time of occurrence of a fault; 、 、 and The dangerous interval value of is determined by the following method: using the data obtained by the digital twin unit, the extreme value of the equipment data at the time of failure is calculated, and the mean and variance of the extreme value are calculated as 、 、 and The risk interval value of And a visual management and control platform to visually display the operating status of the equipment.

9. A fault prediction terminal, characterized in that: The method comprises a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is called and executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable medium, characterized in that The computer-readable medium stores a computer program, and when the computer program is called and executed by a computer, the method according to any one of claims 1 to 7 is implemented.

Images