A Method for Dynamic Evolution and Credibility Evaluation of Equipment Digital Twin Driven by Data

Abstract

The present invention discloses a method for dynamic evolution and credibility evaluation driven by digital twin data of equipment, which relates to the field of digital twin technology. The method includes: obtaining historical data of a physical object and sensor data collected in real time; constructing an equipment digital twin body according to the historical data, and performing credibility evaluation on the sensor data according to a set sampling period to determine whether the sensor data triggers evolution. If so, evolving the equipment digital twin body according to an evolution trigger control method. If not, directly updating the equipment digital twin body according to the sensor data; wherein, the evolution trigger control method is composed of a deep learning network and a Koopman operator. The present invention can improve the accuracy, real-time performance and reliability of model establishment.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twins, and particularly to a method for dynamic evolution and credibility evaluation of equipment digital twins driven by data. Background Art

[0002] The data-driven dynamic evolution method of equipment digital twins refers to using a data-driven method to perform dynamic evolution modeling and simulation of equipment through digital twin technology to achieve real-time state monitoring, prediction, and optimal management of equipment. The background of this method stems from the development of digital twin technology and the needs of the equipment maintenance field. The traditional equipment maintenance mode is usually based on experience and fixed-cycle maintenance plans, which cannot make full use of real-time data and advanced prediction technologies and are difficult to cope with the changes in the real-time state of equipment and the diversification of maintenance requirements. Therefore, data-driven digital twin technology has emerged, aiming to achieve precise monitoring and intelligent maintenance management of equipment status through real-time data collection, modeling, and simulation.

[0003] Currently, many researchers combine artificial intelligence algorithms such as machine learning and deep learning for dynamic modeling of digital twin models. The dynamic modeling method based on artificial intelligence algorithms refers to using technologies such as machine learning and deep learning to perform dynamic modeling of the digital twin model of equipment through real-time data. However, there are still the following technical drawbacks:

[0004] High model complexity: Deep learning models usually have high complexity and require a large amount of computing resources and time to train and optimize the model, with high requirements for computing resources.

[0005] Poor model interpretability: Deep learning models are usually regarded as "black box" models and it is difficult to explain their internal decision-making processes and prediction results, which to a certain extent limits the reliability of the model in practical applications.

[0006] Insufficient real-time performance: The training and inference processes of deep learning models usually take a long time and it is difficult to achieve rapid monitoring and prediction of the real-time state of equipment, especially in scenarios with high real-time requirements, there are certain challenges.

[0007] Therefore, there are still great challenges in aspects such as the accuracy, real-time performance, and reliability of model establishment. Summary of the Invention

[0008] The purpose of the present invention is to provide a method for dynamic evolution and credibility evaluation of equipment digital twins driven by data, which can improve the accuracy, real-time performance, and reliability of model establishment.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] A method for dynamic evolution and credibility evaluation driven by digital twin data of equipment, comprising:

[0011] Obtaining historical data of a physical object and sensor data collected in real time;

[0012] Constructing a digital twin of the equipment according to the historical data, and performing credibility evaluation on the sensor data according to a set sampling period to determine whether the sensor data triggers evolution. If so, evolving the digital twin of the equipment according to an evolution trigger control method. If not, directly updating the digital twin of the equipment according to the sensor data; wherein, the evolution trigger control method is composed of a deep learning network and a Koopman operator.

[0013] Optionally, the calculation formula for the credibility evaluation is:

[0014]

[0015] where Y is the sensor data of the physical object, is the output data of the digital twin of the equipment, m is the number of samples, y i is the i-th sensor data, is the output data of the i-th digital twin of the equipment.

[0016] Optionally, determining whether the sensor data triggers evolution specifically includes:

[0017] Setting a buffer, where the size of the detection time window is adjusted according to the specific scenario. The size of the set time window is q, the sampling period is p, then the amount of data within the time window is q / p, and a scene threshold is given as γ. Two trigger mechanisms are constructed, and when either trigger mechanism is satisfied, evolution is started.

[0018] Optionally, the two trigger mechanisms are respectively: a credibility comparison mechanism for a fixed detection time period and a credibility comparison mechanism for any detection time length; the credibility comparison mechanism for a fixed detection time period is to detect once every fixed time. If the trigger condition is satisfied at the n-th detection, evolution is started; the credibility comparison mechanism for any detection time length is that the error at time tx satisfies the trigger condition, then evolution is triggered.

[0019] Optionally, the credibility comparison mechanism for a fixed detection time period is expressed as:

[0020]

[0021] where E is the credibility, Y is the sensor data of the physical object, is the output data of the digital twin of the equipment, t0 is the initial time, and n is the number of time windows.

[0022] Optionally, the credibility comparison mechanism for any of the detection time lengths is expressed as:

[0023]

[0024] where E is the credibility, Y is the sensor data of the physical object, is the output data of the equipment digital twin.

[0025] Optionally, the evolution trigger control method is specifically to perform a correction operation on the deep learning network based on the Koopman operator through multiple loss functions; the multiple loss functions include L pos loss, L liner loss and L neg loss.

[0026] Optionally, the multiple loss functions are specifically expressed as:

[0027]

[0028] where α i , i = 1,…,6 represents the weighted scalar, θ e , θ d are respectively the regularization terms used to avoid overfitting in the encode and decode processes of the Koopman operator.

[0029] According to the specific embodiments provided by the present invention, the following technical effects are disclosed:

[0030] The present invention discloses a method for dynamic evolution and credibility evaluation driven by equipment digital twin data. The method includes obtaining historical data of a physical object and sensor data collected in real time; constructing an equipment digital twin according to the historical data, and performing credibility evaluation on the sensor data according to a set sampling period to determine whether the sensor data triggers evolution. If so, evolving the equipment digital twin according to the evolution trigger control method, and if not, directly updating the equipment digital twin according to the sensor data; wherein, the evolution trigger control method is composed of a deep learning network and a Koopman operator. The present invention can improve the accuracy, real-time performance and reliability of model establishment. Description of the Drawings

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0032] Figure 1 Schematic flow chart of the method for dynamic evolution and credibility evaluation driven by digital twin data of the equipment of the present invention;

[0033] Figure 2 Logical flow chart of the method for dynamic evolution and credibility evaluation driven by digital twin data of the equipment in this embodiment;

[0034] Figure 3 Schematic time detection diagram of the first triggering mechanism in this embodiment;

[0035] Figure 4 Schematic time detection diagram of the second triggering mechanism in this embodiment;

[0036] Figure 5 Flow chart of the credibility evolution algorithm based on the Koopman operator in this embodiment;

[0037] Figure 6 Structural diagram of the evolution network based on the Koopman operator in this embodiment. Detailed implementation manners

[0038] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0039] The purpose of the present invention is to provide a method for dynamic evolution and credibility evaluation driven by digital twin data of equipment, which can improve the accuracy, real-time performance and reliability of model establishment.

[0040] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0041] As Figure 1 shown, the present invention provides a method for dynamic evolution and credibility evaluation driven by digital twin data of equipment, including:

[0042] Step 100: Obtain the historical data of the physical object and the sensor data collected in real time;

[0043] Step 200: Construct an equipment digital twin based on the historical data, and evaluate the credibility of the sensor data according to a set sampling period to determine whether the sensor data triggers evolution. If so, evolve the equipment digital twin according to the evolution trigger control method. If not, directly update the equipment digital twin according to the sensor data; among them, the evolution trigger control method consists of a deep learning network and a Koopman operator.

[0044] Based on the above technical solution, the overall structure diagram shown in Figure 2 is provided. Among them, the digital twin is a digital model of a physical object, and this model can evolve in real time by receiving data from the physical object, so as to be consistent with the physical object throughout the life cycle.

[0045] In the first part, the physical object collects real-time data through sensors and transmits the real-time data to the digital twin. The transmitted information can support the dynamic real-time evolution of the digital twin, enabling the digital twin to be consistent with the physical object, and thus credibly characterizing or predicting the actual state of the physical object.

[0046] In the second part, after receiving the real-time data, the digital twin performs preliminary processing and then transmits it to the evolution trigger control module. The evolution trigger control module determines whether to trigger the evolution of the digital twin, so that the twin and the physical object are consistent throughout the life cycle.

[0047] Further detailed description of the above steps:

[0048] As shown in Figure 3 , first perform a credibility evaluation on the sensor data X k to judge the credibility of the output of the digital twin model at a certain moment. The calculation method is as follows:

[0049]

[0050] Among them, Y is the sensor data of the physical object, is the output data of the equipment digital twin, m is the number of samples, y i is the i-th sensor data, is the output data of the i-th equipment digital twin.

[0051] When making an evolution trigger judgment, set a buffer, and the size of the detection time window is adjusted according to the specific scenario. Assuming the size of the time window is q and the sampling period is p, then the amount of data within the time window is q / p. Given the scenario threshold is γ, there are two trigger mechanisms, and evolution is started if either condition is met.

[0052] a) Detect once every fixed time. If the following situation occurs during the nth detection (as shown in Figure 4 ), then start the evolution.

[0053]

[0054] Among them, E is the credibility, Y is the sensor data of the physical object, is the output data of the equipment digital twin, t0 is the initial time, and n is the number of time windows.

[0055] b) If the error is large enough at time tx (as shown in Figure 5 ), then trigger the evolution.

[0056]

[0057] When performing the evolution of the digital twin, if the error between the output of the digital twin and the physical equipment of the entity is greater than the threshold, that is, the credibility of the system is low, then trigger the evolution. And the network model based on Koopman is as shown in Figure 6 , and its evolution principle is as follows:

[0058] Use three parts of Loss to correct the network, which are:

[0059]

[0060] Then we can get:

[0061]

[0062] Among them, α i , i = 1,..., 6 represents the weighted scalar, θ e , θ d are the regularization terms used to avoid overfitting in the Koopman operator encode and decode processes respectively.

[0063] Therefore, the evolution algorithm can be realized by solving the following optimization problem:

[0064] Therefore, through the process of this embodiment, it can be seen that compared with the prior art, the present invention combines a deep learning network and uses the Koopman operator to realize the evolution of the digital twin. The advantage is that the Koopman algorithm can achieve dynamic modeling with high real-time requirements through fast calculation of the system. At the same time, the Koopman algorithm is based on the linear representation of the system, which can provide a more intuitive and interpretable description of the dynamic behavior of the system and improve the reliability of the model.

[0065] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other.

[0066] In this article, specific examples are used to elaborate on the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the core idea of the present invention. At the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation on the present invention.

Claims

1. A method for dynamic evolution and credibility evaluation of equipment digital twin data-driven, characterized in that, Including: Obtaining historical data of a physical object and sensor data collected in real time; Constructing an equipment digital twin according to the historical data, and evaluating the credibility of the sensor data according to a set sampling period, determining whether the sensor data triggers evolution, if so, evolving the equipment digital twin according to an evolution trigger control method, if not, directly updating the equipment digital twin according to the sensor data; wherein, the evolution trigger control method is composed of a deep learning network and a Koopman operator; Determining whether the sensor data triggers evolution, specifically including: Setting a buffer, wherein the size of the detection time window is adjusted according to a specific scenario, setting the size of the time window as q, the sampling period as p, then the data volume within the time window is q / p, and giving a scenario threshold as γ, constructing two trigger mechanisms, and when any one of the trigger mechanisms is satisfied, evolution is started; The two trigger mechanisms are respectively: a credibility comparison mechanism with a fixed detection time period and a credibility comparison mechanism with any detection time length; the credibility comparison mechanism with a fixed detection time period is to detect once every fixed time, if the trigger condition is satisfied at the nth detection, evolution is started; the credibility comparison mechanism with any detection time length is that the error at the moment of tx satisfies the trigger condition, then evolution is triggered; The evolution trigger control method is specifically to perform a correction operation on a deep learning network based on the Koopman operator through a multi-loss function; the multi-loss function is specifically expressed as: where α i , i = 1, …, 6 represents a weighted scalar, and θ e , θ d are regularization terms used to avoid overfitting in the encode and decode processes of the Koopman operator, respectively.

2. The method for dynamic evolution and credibility evaluation driven by digital twin data of the equipment according to claim 1, wherein The calculation formula for the credibility evaluation is: Among them, Y is the sensor data of the physical object, is the output data of the equipment digital twin, m is the number of samples, and y i is the i-th sensor data, is the output data of the i-th equipment digital twin.

3. The method for dynamic evolution and credibility evaluation of equipment digital twin data-driven according to claim 1, wherein The credibility comparison mechanism with a fixed detection time period is expressed as: Among them, E is the credibility, Y is the sensor data of the physical object, is the output data of the equipment digital twin, t0 is the initial moment, and n is the number of time windows.

4. The method for dynamic evolution and credibility evaluation driven by digital twin data of equipment according to claim 1, wherein The credibility comparison mechanism with any detection time length is expressed as: Among them, E is the credibility, and Y is the sensor data of the physical object. It is the output data of the equipment digital twin.

Images