CN114841021A - Modification method, device, electronic device and storage medium for digital twin model - Google Patents

Abstract

The embodiments of the present application provide a method, device, electronic device and storage medium for revising a digital twin model, which relate to the technical field of digital twins. By intercepting the physical space data and virtual space data obtained at the current moment and before the current moment according to the preset time sequence length, the target physical space data and target virtual space data are obtained; In the construction model, the reconstruction model is used to reconstruct the target virtual space data, and the reconstruction data in the physical space corresponding to the target virtual space data is obtained; the error index is determined according to the reconstructed data and the target physical space data, and the error index satisfies In the case of correction conditions, the digital twin model is corrected to improve the error accuracy, and then the digital twin model can be corrected in time to improve the accuracy of the digital twin model.

Description

Modification method, device, electronic device and storage medium for digital twin model

technical field

The present application relates to the technical field of digital twins, and in particular, to a method, apparatus, electronic device and storage medium for revising a digital twin model.

Background technique

At present, digital twin technology can be used for fault diagnosis and predictive maintenance of industrial equipment. The process includes digital twin model modeling, digital twin model correction, and digital twin model application. The accuracy of the digital twin model is important for the normal operation of industrial equipment. Operation is of great significance, therefore, the revision of the digital twin model plays a crucial role.

In the prior art, it is often possible to obtain the error between the digital twin model and the industrial equipment by performing original data-level error calculation on the virtual data corresponding to the digital twin model and the physical data corresponding to the industrial equipment, and determine the error based on the error. Whether to correct the digital twin model, but the error calculation at the original data level has the problem of poor error accuracy, so the digital twin model cannot be corrected in time, making the digital twin model inaccurate.

SUMMARY OF THE INVENTION

In view of this, the purpose of the present application is to provide a correction method, device, electronic device and storage medium for a digital twin model, so as to improve the error accuracy, and then the digital twin model can be corrected in time to improve the accuracy of the digital twin model.

In order to achieve the above purpose, the technical solutions adopted in the embodiments of the present application are as follows:

In a first aspect, the present application provides a method for revising a digital twin model, the method comprising:

According to the preset time sequence length, intercept the physical space data and virtual space data obtained at the current moment and before the current moment to obtain target physical space data and target virtual space data; the physical space data is each component on the industrial equipment operation data, the virtual space data is the operation data of each component in the digital twin model corresponding to the industrial equipment;

Input the target virtual space data into the pre-trained reconstruction model, use the reconstruction model to reconstruct the target virtual space data, and obtain the reconstruction in the physical space corresponding to the target virtual space data data;

An error index is determined according to the reconstructed data and the target physical space data, and when the error index satisfies a correction condition, the digital twin model is corrected.

In an optional embodiment, the reconstructed model is obtained by training in the following steps:

Acquiring a plurality of training samples of preset time series lengths; each of the training samples includes physical space data samples and virtual space data samples corresponding to each moment in the preset time series length;

Input all virtual space data samples into a pre-built reconstruction model, use the reconstruction model to reconstruct each virtual space data sample, and obtain the weight in the physical space corresponding to each virtual space data sample. structure data;

Calculate the error index corresponding to each training sample according to the reconstructed data in the physical space corresponding to each of the virtual space data samples, and the physical space data samples corresponding to each of the virtual space data samples;

When the error index does not meet the model convergence condition, iterative optimization is performed on the parameters of the reconstructed model to obtain a trained reconstructed model.

In an optional implementation manner, according to the preset time sequence length, the current moment and the physical space data and virtual space data obtained before the current moment are intercepted to obtain the target physical space data and the target virtual space data, including :

Perform denoising on the physical space data and virtual space data obtained at the current moment and before the current moment to obtain denoised physical space data and virtual space data;

The denoised physical space data and virtual space data are intercepted according to the preset time sequence length to obtain target physical space data and target virtual space data.

In an optional implementation manner, the target virtual space data is input into a pre-trained reconstruction model, and the target virtual space data is reconstructed by using the reconstruction model to obtain the target virtual space Reconstructed data in physical space corresponding to spatial data, including:

Inputting the target virtual space data into the coding layer of the pre-trained reconstruction model to obtain a context vector corresponding to the target virtual space data;

Inputting the context vector and the target virtual space data into the decoding layer of the reconstruction model to obtain initial reconstruction data in the physical space corresponding to the target virtual space data;

The initial reconstruction data is input into the nonlinear projection layer of the reconstruction model to obtain reconstruction data in the physical space corresponding to the target virtual space data.

In an optional implementation manner, the determining an error index according to the reconstructed data and the target physical space data includes:

According to the reconstructed data and the target physical space data, calculate the shape difference and the distortion difference between the target virtual space data and the target physical space data;

The error index is calculated from the shape difference and the twist difference.

In an optional implementation manner, calculating the shape difference and distortion difference between the target virtual space data and the target physical space data according to the reconstructed data and the target physical space data includes:

The shape difference and distortion difference between the target virtual space data and the target physical space data are calculated by the following formulas:

表征目标虚拟空间数据对应的物理空间下的重构数据，

表征目标物理空间数据，

表征形状差异，

表征扭曲差异，

表征大于等于0的平滑指数，

表征目标虚拟空间数据对应的物理空间下的重构数据，与目标物理空间数据之间的规整路径，

表征目标虚拟空间数据对应的物理空间下的重构数据，与目标物理空间数据的规整开销矩阵。in,

Represents the reconstructed data in the physical space corresponding to the target virtual space data,

Characterize the target physical space data,

to characterize shape differences,

Representation Distortion Differences,

represents a smoothing exponent greater than or equal to 0,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular path between the target physical space data,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular cost matrix of the target physical space data.

In an optional implementation manner, the calculating the error index according to the shape difference and the distortion difference includes: calculating the error index by the following formula:

表征所述误差指数，

表征超参数，

表征所述形状差异，

表征所述扭曲差异。in,

characterizing the error index,

Characterization hyperparameters,

to characterize the shape difference,

Characterize the twist difference.

In a second aspect, the present application provides a device for correcting a digital twin model, the device comprising:

The acquisition module is used for intercepting the physical space data and virtual space data obtained at the current moment and before the current moment according to the preset time sequence length to obtain target physical space data and target virtual space data; the physical space data is: The operation data of each component on the industrial equipment, the virtual space data is the operation data of each component in the digital twin model corresponding to the industrial equipment;

The processing module is used to input the target virtual space data into the pre-trained reconstruction model, and use the reconstruction model to reconstruct the target virtual space data to obtain the physical data corresponding to the target virtual space data. Reconstructed data in space;

A correction module, configured to determine an error index according to the reconstructed data and the target physical space data, and correct the digital twin model when the error index satisfies a correction condition.

In a third aspect, the present application provides an electronic device, including a processor and a memory, where the memory stores a computer program that can be executed by the processor, and the processor can execute the computer program to implement any of the foregoing embodiments. one of the methods described.

In a fourth aspect, the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the method according to any one of the foregoing embodiments.

The digital twin model correction method, device, electronic device, and storage medium provided by the embodiments of the present application obtain target physical space data and target virtual space data in combination with timing information, and reconstruct the target virtual space data to obtain the target virtual space data. The reconstructed data in the physical space corresponding to the spatial data, and the error index is determined according to the reconstructed data and the target physical space data, thereby avoiding the error caused by the error calculation of the original data level only based on the virtual space data and the physical space data. The problem of low accuracy improves the error accuracy, and then the digital twin model can be corrected in time, which improves the accuracy of the digital twin model.

In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the preferred embodiments are exemplified below, and are described in detail as follows in conjunction with the accompanying drawings.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present application more clearly, the following drawings will briefly introduce the drawings that need to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore do not It should be regarded as a limitation of the scope, and for those of ordinary skill in the art, other related drawings can also be obtained according to these drawings without any creative effort.

Fig. 1 shows the schematic diagram of the time series autocorrelation of physical data and virtual data;

Fig. 2 shows the schematic diagram that does not consider the time series autocorrelation of physical data and virtual data;

FIG. 3 shows a schematic block diagram of an electronic device provided by an embodiment of the present application;

4 shows a schematic flowchart of a method for revising a digital twin model provided by an embodiment of the present application;

FIG. 5 shows another schematic flowchart of the method for revising a digital twin model provided by an embodiment of the present application;

Figure 6 shows a schematic diagram of a sliding window intercepting training samples;

FIG. 7 shows another schematic flowchart of the method for revising a digital twin model provided by an embodiment of the present application;

FIG. 8 shows another schematic flowchart of the method for revising a digital twin model provided by an embodiment of the present application;

FIG. 9 shows another schematic flowchart of the method for correcting the digital twin model provided by the embodiment of the present application;

FIG. 10 shows a functional block diagram of the device for correcting the digital twin model provided by the embodiment of the present application;

FIG. 11 shows another functional block diagram of the apparatus for correcting the digital twin model provided by the embodiment of the present application.

Icons: 100-electronic device; 110-memory; 120-processor; 130-communication module; 200-acquisition module; 210-processing module; 220-correction module; 230-model training module.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application generally described and illustrated in the drawings herein may be arranged and designed in a variety of different configurations.

Thus, the following detailed description of the embodiments of the application provided in the accompanying drawings is not intended to limit the scope of the application as claimed, but is merely representative of selected embodiments of the application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

It should be noted that relational terms such as the terms "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any relationship between these entities or operations. any such actual relationship or sequence exists. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device that includes a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in a process, method, article or apparatus that includes the element.

At present, digital twin technology can be used for fault diagnosis and predictive maintenance of industrial equipment. The process includes digital twin model modeling, digital twin model correction, and digital twin model application. The accuracy of the digital twin model is important for the normal operation of industrial equipment. Operation is of great significance, therefore, the revision of the digital twin model plays a crucial role.

In the prior art, the error between the virtual data of the digital twin model and the physical data of the industrial equipment can generally be calculated, so that when the error meets the preset condition, the error learning of the clustering is performed, and the virtual data corresponding to the digital twin model is extracted. The characteristic data of the physical space corresponding to the space and the industrial equipment is used to correct the digital twin model. It is understandable that in the prior art, only the virtual data of the digital twin model and the physical data of the industrial equipment are actually calculated at the original data level, that is, the virtual data of the digital twin model and the physical data of the industrial equipment are directly calculated. error, on this basis, the existing technology mainly has the following problems:

1. The time series data generated by the equipment during operation often has time series autocorrelation. With the change of time, the data will have a certain contextual correlation, such as showing the characteristics of monotonic changes. Referring to Figure 1, it can be seen that with the increase of time, the physical data and virtual data will show the characteristics of monotonous increase. Therefore, in the prior art, only the error calculation at the original data level is performed, and the monotonic variation characteristics of data over time are not considered. Please refer to FIG. 2, although the error variation between the virtual space data and the physical space data in FIG. Small, but with the change of time, the virtual data shows a trend of increasing and then decreasing, which obviously does not satisfy the autocorrelation on the time series. The existing technology obviously cannot take this situation into account when calculating the error, so it will As a result, the digital twin model cannot correct the problem in time.

2. When industrial equipment is running, there may also be a certain correlation between its various components and various operating states. For example, there is a certain dependence and restriction relationship between the positions of each axis of the robot, and there is a certain relationship between torque and current. dependencies. It can be understood that the correlation is mainly reflected in the fact that in an overall operation process, there may be a certain correlation between the various components at the same time, and there may also be a certain correlation between the various components at different times. The error calculation at the raw data level obviously does not take it into account.

To sum up, the error calculation at the original data level in the prior art cannot take into account the autocorrelation of the data in the time series and the correlation of each component and each device during the operation process, so there is a problem of poor error accuracy, which will lead to As a result, the digital twin model cannot be corrected based on the error in time, resulting in an inaccurate digital twin model.

Based on this, the embodiments of the present application provide a method for revising a digital twin model to solve the above problems.

Please refer to FIG. 3 , which is a schematic block diagram of an electronic device 100 . The electronic device 100 may be a terminal device, such as a PC terminal, a mobile terminal, and the like. The electronic device 100 also needs to be communicatively connected with the industrial device, so as to obtain the physical space data sent by the industrial device.

Optionally, the industrial equipment may be equipment engaged in industrial production, such as industrial robots, industrial machine tools, and the like.

In a possible implementation manner, the electronic device 100 may also be pre-set with a digital twin model corresponding to the industrial equipment, so that the simulated spatial data of the digital twin model can be directly obtained; in another possible implementation manner, The electronic device 100 may be connected in communication with other electronic devices provided with a digital twin model corresponding to the industrial device, so as to acquire simulated spatial data of the digital twin model from the other electronic devices.

The electronic device 100 includes a memory 110 , a processor 120 and a communication module 130 . The elements of the memory 110 , the processor 120 and the communication module 130 are directly or indirectly electrically connected to each other to realize data transmission or interaction. For example, these elements may be electrically connected to each other through one or more communication buses or signal lines.

The memory 110 is used for storing programs or data. The memory 110 may be, but is not limited to, a random access memory (Random Access Memory, RAM), a read only memory (Read Only Memory, ROM), a programmable read only memory (Programmable Read-Only Memory, PROM), which can be Erasable Read-Only Memory (ErasableProgrammable Read-Only Memory, EPROM), Electrical Erasable Programmable Read-Only Memory (EEPROM), etc.

The processor 120 is used to read/write data or programs stored in the memory, and perform corresponding functions.

The communication module 130 is configured to establish a communication connection between the server and other communication terminals through the network, and to send and receive data through the network.

It should be understood that the structure shown in FIG. 3 is only a schematic structural diagram of the electronic device 100 , and the electronic device 100 may further include more or less components than those shown in FIG. 3 , or have different components from those shown in FIG. 3 . Configuration. Each component shown in FIG. 3 can be implemented in hardware, software, or a combination thereof.

Embodiments of the present application further provide a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for revising the digital twin model provided by the embodiments of the present application can be implemented.

Next, the electronic device 100 is used as the main body of execution, and the method for revising the digital twin model provided by the embodiment of the present application will be described in detail in conjunction with the schematic flowchart. For details, please refer to FIG. 4 . A schematic flow chart, the method includes:

Step S20, according to the preset time sequence length, intercept the physical space data and virtual space data obtained at the current moment and before the current moment, and obtain target physical space data and target virtual space data;

Among them, the physical space data is the operation data of each component on the industrial equipment, and the virtual space data is the operation data of each component in the digital twin model corresponding to the industrial equipment;

Optionally, the physical space data may be data obtained from industrial equipment, including operating data of various components on the industrial equipment; the virtual space data may be data obtained from a digital twin model corresponding to the industrial equipment, including Operating data of individual components.

It can be understood that the physical space data is the actual data obtained during the actual operation of the industrial equipment, and the virtual space data is the virtual data obtained through the simulation operation of the digital twin model.

Optionally, the digital twin model can run synchronously with its corresponding industrial equipment. In this case, the physical space data and virtual space data obtained at the same time correspond.

In one possible implementation, the electronic device can acquire and store physical space data and virtual space data in real time, and then intercept them when errors need to be determined; in another possible implementation, the electronic device The device can collect physical space data and virtual space data from industrial equipment and virtual space data at preset time intervals.

Optionally, the preset time sequence length is the length of intercepting the physical space data and the virtual space data. For example, if the preset time sequence length is two minutes, the current time can be intercepted to within two minutes from the current time, The physical space data and virtual space data corresponding to each moment.

Optionally, the target physical space data is a collection of physical space data corresponding to each moment after the physical space data is intercepted according to a preset time sequence length. It can be understood that the target physical space data includes corresponding data at multiple moments. Physical space data, and the physical space data corresponding to each moment includes the operation data of each component of the industrial equipment at that moment.

Optionally, the target virtual space data is a collection of virtual space data corresponding to each moment after the virtual space data is intercepted according to a preset time sequence length. It can be understood that the target virtual space data includes corresponding data at multiple moments. Virtual space data, and the virtual space data corresponding to each moment includes the operation data of each component in the digital twin model at that moment.

Optionally, the electronic device may intercept the physical space data and virtual space data obtained at the current moment and before the current moment in a sliding window manner. For example, the physical data and virtual data obtained at the current time and before the current time are placed in a time series, and the target physical space data and the target virtual space data can be directly intercepted according to the preset time sequence length by means of a sliding window.

Step S21, input the target virtual space data into the pre-trained reconstruction model, use the reconstruction model to reconstruct the target virtual space data, and obtain the reconstruction data in the physical space corresponding to the target virtual space data;

Optionally, a pre-trained reconstruction model may be set in the electronic device, after obtaining the target virtual space data, the target virtual space data may be input into the reconstruction model, and the target virtual space may be reconstructed by using the reconstruction model. The spatial data is processed, and the reconstructed data in the physical space corresponding to the target virtual spatial data is finally obtained.

Optionally, the reconstructed data represents data obtained by reconstructing the physical space according to the target virtual space data. It can be understood that the reconstructed data is the data fed back from the target virtual space data to the physical space.

Step S22, determining an error index according to the reconstructed data and the target physical space data, and correcting the digital twin model when the error index satisfies the correction condition.

Optionally, the error index can characterize the error between the digital twin model and the industrial equipment, that is, feedback whether the digital twin model can reconstruct the physical space with high accuracy.

Optionally, the correction condition includes that the error index is greater than a first threshold, or the error index is less than a second threshold, wherein the first threshold is greater than the second threshold. It can be understood that the first threshold and the second threshold are the upper and lower thresholds of the error, respectively. In a possible implementation manner, the first threshold and the second threshold may be values determined in combination with loss function values according to actual requirements during the training process of the reconstructed model.

Optionally, the electronic device may correct the digital twin model by optimizing the digital twin model or retraining the digital twin model under the condition that the error index satisfies the correction condition. In a possible way, the correction of the digital twin model can be realized through artificial intelligence technologies such as deep learning and machine vision.

Optionally, if the industrial equipment is equipment with high precision requirements, the physical space data and virtual space data can be acquired in real time, and the error index can be calculated in real time to ensure timely correction of the digital twin model; if the industrial equipment is required for precision If the data is lower, a preset time period can be set in the electronic device, so that the electronic device can calculate the error index every preset time period.

The method for revising a digital twin model provided by the embodiment of the present application intercepts the physical space data and virtual space data obtained at the current moment and before the current moment according to a preset time sequence length to obtain target physical space data and target virtual space data, and then Use the pre-trained reconstruction model to reconstruct the target virtual space data, so as to obtain the reconstruction data in the physical space corresponding to the target virtual space data, and then determine the error index according to the reconstructed data and the target physical space data, thereby Determines whether to make corrections to the digital twin. The target physical space data and the target virtual space data are obtained in combination with the timing information, the reconstruction data in the physical space corresponding to the target virtual space data is obtained by reconstructing the target virtual space data, and the reconstruction data and the target physical space are obtained according to the reconstruction data and the target physical space. The spatial data determines the error index, thereby avoiding the problem of low error accuracy caused by the original data-level error calculation only based on virtual spatial data and physical spatial data, improving the error accuracy, and then correcting the digital twin model in time. Improved the accuracy of the digital twin model.

Optionally, in order to realize the acquisition of reconstruction data, the reconstruction model needs to be trained in advance. Specifically, on the basis of FIG. 4 , FIG. 5 is another schematic flowchart of the modification method of the digital twin model provided by the embodiment of the present application, Referring to Figure 5, this reconstructed model can be obtained by the following training steps:

Step S10, acquiring a plurality of training samples of preset time sequence lengths;

Wherein, each training sample includes a physical space data sample and a virtual space data sample corresponding to each moment within the preset time sequence length;

Optionally, the training sample is a sample for training the reconstructed model.

Optionally, the electronic device may first acquire physical space data corresponding to the industrial equipment and virtual space data corresponding to the twin model within a preset time period, and then acquire training samples according to a preset time sequence length. It can be understood that the physical space data is the operation data of each component on the industrial equipment, and the virtual space data is the operation data of each component in the digital twin model corresponding to the industrial equipment.

Optionally, the preset time period may be set in advance in the electronic device according to the operation cycle of the industrial equipment, for example, the preset time period may be the time length corresponding to one operation cycle of the industrial equipment, or The duration corresponding to the multiple.

Optionally, the electronic device may intercept the physical space data corresponding to the industrial equipment and the virtual space data corresponding to the twin model within the preset time period according to the preset time sequence length by means of a sliding window.

In an example, referring to FIG. 6 , the physical space data and the virtual space data can be corresponded at different times, and the physical space data and the virtual space data can be intercepted at the same time according to the preset time sequence length by means of a sliding window to obtain There are multiple windows, and the sample data corresponding to each window is the training sample of the preset time sequence length, and the training sample includes physical space data samples and virtual space data samples corresponding to each moment.

It can be understood that for a physical space data sample, it contains the physical space data corresponding to each moment in the preset time sequence length, and the physical space data corresponding to each moment includes the various components of the industrial equipment at that moment. For a virtual space data sample, it contains virtual space data corresponding to each moment in the preset time series length, and the virtual space data corresponding to each moment includes various components in the digital twin model Operational data at this moment.

Optionally, in order to ensure the accuracy of the training, the electronic device may also screen the physical space data and the virtual space data within the preset time period before intercepting the physical space data and the virtual space data within the preset time period, For example, denoising is performed on it, so as to intercept the physical space data and the virtual space data within the denoised preset time period.

Step S11, input all virtual space data samples into a pre-built reconstruction model, use the reconstruction model to reconstruct each virtual space data sample, and obtain reconstructed data in the physical space corresponding to each virtual space data sample;

Optionally, the reconstruction model may include an encoding layer, a decoding layer, and a nonlinear projection layer. After inputting all the virtual space data samples into the pre-built reconstruction model, the codes in the reconstruction model can be used in turn. The layer, the decoding layer, and the nonlinear projection layer process each virtual space data sample, thereby obtaining reconstructed data in the physical space corresponding to each virtual space data sample.

Specifically, the coding layer may be an LSTM (Long Short-Term Memory, artificial neural network with long short-term memory) unit, which is used to learn the whole of the virtual spatial data sample by taking the plurality of data in the virtual spatial data sample as a whole feature.

Optionally, the coding layer can process the virtual space data corresponding to each moment in the virtual space data sample, and update its hidden state, that is, combine the hidden state of the previous moment and the virtual space data of the current moment to obtain the following: A momentary hidden state. On this basis, the hidden state output by the encoder at time t can be expressed as the following formula:

表征编码器所输出的t时刻下的隐状态，

表征编码器所输出的t-1时刻下的隐状态，

表征t时刻对应的虚拟空间数据。in,

Represents the hidden state at time t output by the encoder,

Represents the hidden state at time t-1 output by the encoder,

Characterize the virtual space data corresponding to time t.

Optionally, by processing the virtual space data sample by the encoder, the context vector of the virtual space data sample can be obtained at the last moment corresponding to the virtual space data in the virtual space data sample, and the context vector can be expressed as The following formula:

表征上下文向量，

表征虚拟空间数据样本中的虚拟空间数据对应的最后一个时刻，

表征编码器所输出的

时刻下的隐状态，

表征

时刻对应的虚拟空间数据。in,

representing the context vector,

represents the last moment corresponding to the virtual space data in the virtual space data sample,

Characterize the output of the encoder

the hidden state of the moment,

representation

The virtual space data corresponding to the time.

It can be understood that after all the virtual space data samples are input into the coding layer, the context vector corresponding to each data sample can be obtained, and the context vector can reflect the virtual space data contained in the virtual space data sample at each moment. relationship between the whole.

Optionally, the decoding layer may also be an LSTM unit, which is used to reconstruct the data of the virtual space data sample in the physical space by combining the context vector and the virtual space data sample.

In an example, the decoding layer can use a teacher-forcing mode when decoding. In this mode, the hidden state of the previous moment, the virtual space data of the previous moment, and the context vector can be combined to obtain the hidden state of the next moment. . On this basis, the hidden state output by the decoder at time t can be expressed as the following formula:

表征解码器所输出的

时刻下的隐状态，

表征解码器所输出的

时刻下的隐状态，

表征

时刻对应的虚拟空间数据，

表征上下文向量。in,

Characterize what the decoder outputs

the hidden state of the moment,

Characterize what the decoder outputs

the hidden state of the moment,

representation

The virtual space data corresponding to the time,

Represents the context vector.

In another example, the decoder may not use the teacher-forcing mode when decoding. In this case, the hidden state at the previous moment and the reconstruction in the physical space corresponding to the virtual space data at the previous moment may be combined. data and context vector to obtain the hidden state at the next moment. On this basis, the hidden state output by the decoder at time t can be expressed as the following formula:

表征解码器所输出的

时刻的隐状态，

表征解码器所输出的

时刻下的隐状态，

表征

时刻下的虚拟空间数据对应的物理空间下的重构数据，

表征上下文向量。in,

Characterize what the decoder outputs

the hidden state of the moment,

Characterize what the decoder outputs

the hidden state of the moment,

representation

The reconstructed data in the physical space corresponding to the virtual space data at the moment,

Represents the context vector.

In this embodiment, after each virtual space data sample and its corresponding context vector are input into the decoding layer of the reconstruction model, through the processing of the decoding layer, the physical space corresponding to each virtual space data sample can finally be obtained The initial reconstruction data below. At this time, for any initial reconstruction data, its data dimension may be different from the data dimension of the corresponding virtual space data sample.

Optionally, in order to unify the data dimensions of the initial reconstructed data and its corresponding virtual spatial data samples, each initial reconstructed data can be input into the nonlinear projection layer for processing, thereby obtaining each virtual spatial data sample. Reconstructed data in the corresponding physical space. It can be understood that for any reconstructed data, it includes reconstructed data in the physical space corresponding to the virtual space data at each moment in the preset time series length, and the data dimension of the reconstructed data is also the same as the corresponding data. The data dimensions of the physical spatial data samples are the same.

Step S12, according to the reconstruction data in the physical space corresponding to each virtual space data sample, and the physical space data sample corresponding to each virtual space data sample, calculate the error index corresponding to each training sample;

Optionally, each virtual space data sample and its corresponding physical space data sample can be calculated according to the reconstructed data in the physical space corresponding to each virtual space data sample and the physical space data sample corresponding to each virtual space data sample. shape difference and warp difference between, and then calculate the error index corresponding to each training sample according to the shape difference and warp difference.

Optionally, shape differences and distortion differences can be calculated by constructing a regular path between the reconstructed data and the corresponding physical space sample data.

的规整矩阵，并将该重构数据与物理空间数据样本之间的规划路径的起点和终点，分别定义为该规整矩阵的左上角坐标（0,0），以及右下角坐标（n,n），并定义每次仅能沿着向右、向下或者向右下角方向移动。Specifically, for any physical space sample data, if the physical space sample data includes physical space data of n components at multiple times, it can be understood that the corresponding reconstruction data also includes n components at multiple times. On the basis of the data at this time, we can construct a

The regular matrix of , and the starting point and end point of the planned path between the reconstructed data and the physical space data samples are defined as the coordinates of the upper left corner (0, 0) and the coordinates of the lower right corner (n, n) of the regular matrix, respectively. , and define that it can only move in the right, down or bottom right direction at a time.

On this basis, the error index can be calculated by the following formula:

表征误差指数，

表征超参数，

表征形状差异，

表征扭曲差异。in,

Characterization Error Index,

Characterization hyperparameters,

to characterize shape differences,

Characterization of distortion differences.

Combining the Bellman equation on the basis of this formula, we can get:

表征虚拟空间数据样本对应的物理空间下的重构数据，

表征物理空间数据样本，

表征对虚拟空间数据样本对应的物理空间下的重构数据，以及物理空间数据样本进行动态时间规整计算，

表征形状差异，

表征大于等于0的平滑指数，

表征虚拟空间数据样本对应的物理空间下的重构数据，与物理空间数据样本之间的规整路径，

表征目标虚拟空间数据对应的物理空间下的重构数据，与目标物理空间数据的规整开销矩阵。in,

Represents the reconstructed data in the physical space corresponding to the virtual space data sample,

to characterize physical spatial data samples,

Represents the reconstructed data in the physical space corresponding to the virtual space data samples, and the dynamic time warping calculation of the physical space data samples,

to characterize shape differences,

represents a smoothing exponent greater than or equal to 0,

Represents the reconstructed data in the physical space corresponding to the virtual space data samples, and the regular path between the physical space data samples,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular cost matrix of the target physical space data.

Optionally, the twist difference can be calculated by calculating the difference between the optimal regular path and the penalty square matrix, then the twist difference can be obtained by the following formula:

表征扭曲差异，

表征虚拟空间数据样本对应的物理空间下的重构数据，与物理空间数据样本之间的最优规整路径，

是一个

的惩罚方阵，其中的每个元素值表征该元素位置上的点到该方阵对角线上的距离。in,

Representation Distortion Differences,

Characterize the optimal regular path between the reconstructed data in the physical space corresponding to the virtual space data sample and the physical space data sample,

Is an

The penalty square matrix of , in which each element value represents the distance from the point on the element's position to the diagonal of the square matrix.

方阵中的每个元素值可以通过以下公式获得：Optionally,

The value of each element in the square matrix can be obtained by the following formula:

表征

方阵中的行信息，

表征

方阵中的列信息，

表征

方阵中的元素位置。综上，结合该形状差异以及扭曲差异，该误差指数可通过以下公式最终确定：in,

representation

row information in a square matrix,

representation

column information in a square matrix,

representation

Element positions in a square matrix. To sum up, combining the shape difference and the distortion difference, the error index can be finally determined by the following formula:

In this embodiment, an error index needs to be calculated according to each reconstructed data and combined with its corresponding physical space data sample, so as to obtain an error index corresponding to each training sample.

Step S13, in the case that the error index does not meet the model convergence condition, iteratively optimize the parameters of the reconstructed model to obtain a trained reconstructed model.

Optionally, the convergence condition is a condition for determining to stop the iterative optimization of the reconstructed model. In a possible implementation manner, the convergence condition may be that the error index corresponding to each training sample is consistent in multiple trainings. reaches the preset number of times.

Optionally, the preset number of times is the number of times preset in the electronic device according to actual needs.

Optionally, if the error index does not meet the model convergence condition, the parameters of the reconstructed model can be iteratively optimized based on the error index corresponding to each sample, until the error index corresponding to each training sample is consistent in multiple training times. When the preset number of times is reached, the training is stopped and the trained reconstructed model is obtained.

Optionally, in the process of practical application, in order to improve the accuracy of the error, the physical space data and virtual space data obtained at the current moment and before the current moment can also be screened first, and then the screened physical space data and virtual space data can be screened. Interception of spatial data. Specifically, on the basis of FIG. 4 , FIG. 7 is another schematic flowchart of a method for correcting a digital twin model provided by an embodiment of the present application. Referring to FIG. 7 , the above step S20 can also be obtained by the following steps:

Step S20-1, performing denoising on the physical space data and virtual space data obtained at the current moment and before the current moment, to obtain the denoised physical space data and virtual space data;

Step S20-2, intercepting the denoised physical space data and virtual space data according to a preset time sequence length to obtain target physical space data and target virtual space data.

In this embodiment, the acquired physical space data and virtual space data can be denoised, so as to realize the cleaning of the physical space data and virtual space data obtained at the current moment and before the current moment, and then according to the preset time sequence length The denoised physical space data and virtual space data are intercepted to obtain target physical space data and target virtual space data.

Optionally, the pre-trained reconstruction model may include an encoding layer, a decoding layer and a nonlinear projection layer. On this basis, after inputting the target virtual space data into the pre-trained reconstruction model, the The encoding layer, the decoding layer and the nonlinear projection layer process the target virtual space data, and finally obtain the reconstructed data in the physical space corresponding to the target virtual space data.

Specifically, on the basis of FIG. 4 , FIG. 8 is another schematic flowchart of a method for correcting a digital twin model provided by an embodiment of the present application. Referring to FIG. 8 , the above step S21 can also be obtained by the following steps:

Step S21-1, input the target virtual space data into the coding layer of the pre-trained reconstruction model to obtain the context vector corresponding to the target virtual space data;

Optionally, the target virtual space data is input into the pre-trained reconstruction model, the target virtual space data can be processed through the coding layer of the reconstruction model first, combined with the hidden state of the previous moment and the hidden state of the previous moment. Virtual space data and context vector, obtain the hidden state at the next moment, and finally obtain the context vector corresponding to the target virtual space data, and the context vector can reflect the virtual space data contained in the target virtual space data at each moment. relationship between the whole.

Step S21-2, input the context vector and the target virtual space data into the decoding layer of the reconstruction model, and obtain the initial reconstruction data in the physical space corresponding to the target virtual space data;

Optionally, the obtained context vector and target virtual space data can be input into the decoding layer of the reconstruction model, and the hidden state of the next moment can be obtained by combining the hidden state of the previous moment, the virtual space data of the previous moment, and the context vector. , or combine the hidden state of the previous moment, the reconstructed data in the physical space corresponding to the virtual space data of the previous moment, and the context vector to obtain the hidden state of the next moment, so as to finally obtain the physical space corresponding to the target virtual space data. the initial reconstruction data.

In step S21-3, the initial reconstruction data is input into the nonlinear projection layer of the reconstruction model to obtain reconstruction data in the physical space corresponding to the target virtual space data.

In this embodiment, the nonlinear projection layer can be used to process the initial reconstructed data, so as to obtain reconstructed data in the physical space corresponding to the target virtual space data with the same dimension as the target virtual space data.

It can be understood that the reconstructed data at this time is the set of reconstructed data in the physical space of the virtual space data corresponding to each moment in the target virtual space data.

The method for correcting the digital twin model provided by the embodiment of the present application uses the encoding layer, the decoding layer and the nonlinear projection layer to take the virtual space data corresponding to each moment in the target virtual space data as a whole, determine its overall characteristics and determine the overall characteristics according to the The target virtual space data is reconstructed to obtain the reconstructed data in the physical space corresponding to the target virtual space data, taking into account the autocorrelation of the virtual space data in the time series, as well as the correlation between various components and various operating states. On the basis of the nature of the reconstruction model, the data in the physical space corresponding to the target virtual space data can be reconstructed as much as possible through the reconstruction model, so the accuracy of the error index can be improved when the error index is determined according to the reconstructed data and the target physical space data. .

Optionally, on the basis of FIG. 4 , FIG. 9 is another schematic flowchart of a method for revising a digital twin model provided by an embodiment of the present application. Referring to FIG. 9 , in the above step S22, according to the reconstructed data and The target physical space data determines the error index, which can also be achieved by the following steps:

Step S22-1, according to the reconstructed data and the target physical space data, calculate the shape difference and the distortion difference between the target virtual space data and the target physical space data;

Optionally, the shape difference and twist difference can be calculated by the following formulas:

表征目标虚拟空间数据对应的物理空间下的重构数据，

表征目标物理空间数据，

表征形状差异，

表征扭曲差异，

表征大于等于0的平滑指数，

表征目标虚拟空间数据对应的物理空间下的重构数据，与目标物理空间数据之间的规整路径，

表征目标虚拟空间数据对应的物理空间下的重构数据，与目标物理空间数据的规整开销矩阵。in,

Represents the reconstructed data in the physical space corresponding to the target virtual space data,

Characterize the target physical space data,

to characterize shape differences,

Representation Distortion Differences,

represents a smoothing exponent greater than or equal to 0,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular path between the target physical space data,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular cost matrix of the target physical space data.

In step S22-2, an error index is calculated according to the shape difference and the distortion difference.

In this embodiment, the error index can be calculated by the following formula:

表征误差指数，

表征超参数，

表征形状差异，

表征扭曲差异。in,

Characterization Error Index,

Characterization hyperparameters,

to characterize shape differences,

Characterization of distortion differences.

Optionally, in the case that the shape difference and the distortion difference are represented by the above formulas, the error index can also be obtained by the following formula:

It can be understood that, after the error index is obtained by calculation, it can be determined whether to correct the digital twin model according to the error index and the correction conditions.

In order to perform the corresponding steps in the foregoing embodiments and various possible manners, an implementation manner of an apparatus for correcting a digital twin model is given below. Further, please refer to FIG. 10 , which is a functional block diagram of the apparatus for correcting a digital twin model provided by an embodiment of the present application. It should be noted that the basic principle and the technical effect of the device for correcting the digital twin model provided in this embodiment are the same as those in the above-mentioned embodiment. Corresponding content in the examples. The device for correcting the digital twin model includes: an acquisition module 200 , a processing module 210 , and a correction module 220 .

The obtaining module 200 is used for intercepting the physical space data and virtual space data obtained at the current moment and before the current moment according to a preset time sequence length to obtain target physical space data and target virtual space data; the physical space data is industrial equipment The operation data of each component above, and the virtual space data is the operation data of each component in the digital twin model corresponding to the industrial equipment;

It can be understood that the obtaining module 200 can be used to perform the above step S20;

The processing module 210 is configured to input the target virtual space data into the pre-trained reconstruction model, use the reconstruction model to reconstruct the target virtual space data, and obtain the reconstruction data in the physical space corresponding to the target virtual space data ;

It can be understood that the processing module 210 can be used to perform the above step S21;

The correction module 220 is configured to determine an error index according to the reconstructed data and the target physical space data, and correct the digital twin model when the error index satisfies the correction condition.

It can be understood that the correction module 220 can be used to perform the above step S22.

Optionally, FIG. 11 is another functional block diagram of the apparatus for correcting a digital twin model provided by an embodiment of the present application. Please refer to FIG. 11 , the apparatus for correcting a digital twin model may further include a model training module 230 .

The model training module 230 is used to obtain a plurality of training samples of preset time series length; each training sample includes physical space data samples and virtual space data samples corresponding to each moment in the preset time series length; The sample is input into a pre-built reconstruction model, and the reconstruction model is used to reconstruct each virtual space data sample, and the reconstruction data in the physical space corresponding to each virtual space data sample is obtained; The reconstructed data in the physical space, and the physical space data samples corresponding to each virtual space data sample, calculate the error index corresponding to each training sample; if the error index does not meet the model convergence conditions, the reconstruction model The parameters are iteratively optimized to obtain a trained reconstructed model.

It can be understood that the model training module 230 can be used to perform the above steps S10 to S13.

Optionally, the obtaining module 200 is further configured to perform denoising on the physical space data and virtual space data obtained at the current moment and before the current moment, and obtain the denoised physical space data and virtual space data; Length intercepts the denoised physical space data and virtual space data to obtain target physical space data and target virtual space data.

It can be understood that the obtaining module 200 can also be used to perform the above steps S20-1 to S20-2.

Optionally, the processing module 210 is further configured to input the target virtual space data into the coding layer of the pre-trained reconstruction model to obtain a context vector corresponding to the target virtual space data; input the context vector and the target virtual space data into the reconfiguration model. The decoding layer of the reconstruction model is used to obtain the initial reconstruction data in the physical space corresponding to the target virtual space data; the initial reconstruction data is input into the nonlinear projection layer of the reconstruction model to obtain the reconstruction data in the physical space corresponding to the target virtual space data. structure data.

It can be understood that the processing module 210 can also be used to execute the above steps S21-1 to S21-3.

Optionally, the correction module 220 is also used to calculate the shape difference and the distortion difference between the target virtual space data and the target physical space data according to the reconstructed data and the target physical space data; according to the shape difference and the distortion difference, calculate the error. index.

It can be understood that the correction module 220 can be used to execute the above steps S22-1 to S22-2.

Optionally, the correction module 220 is further configured to calculate the shape difference and the distortion difference between the target virtual space data and the target physical space data by the following formula:

表征目标虚拟空间数据对应的物理空间下的重构数据，

表征目标物理空间数据，

表征形状差异，

表征扭曲差异，

表征大于等于0的平滑指数，

表征目标虚拟空间数据对应的物理空间下的重构数据，与目标物理空间数据之间的规整路径，

表征目标虚拟空间数据对应的物理空间下的重构数据，与目标物理空间数据的规整开销矩阵。in,

Represents the reconstructed data in the physical space corresponding to the target virtual space data,

Characterize the target physical space data,

to characterize shape differences,

Representation Distortion Differences,

represents a smoothing exponent greater than or equal to 0,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular path between the target physical space data,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular cost matrix of the target physical space data.

Optionally, the correction module 220 is further configured to calculate the error index by the following formula:

表征误差指数，

表征超参数，

表征形状差异，

表征扭曲差异。in,

Characterization Error Index,

Characterization hyperparameters,

to characterize shape differences,

Characterization of distortion differences.

The apparatus for correcting the digital twin model provided by the embodiment of the present application, through the acquisition module intercepts the physical space data and virtual space data obtained at the current moment and before the current moment according to the preset time sequence length, and obtains the target physical space data and the target virtual space data. Spatial data; physical space data is the operation data of each component on the industrial equipment, and virtual space data is the operation data of each component in the digital twin model corresponding to the industrial equipment; the target virtual space data is input into the pre-trained data through the processing module. In the construction model, the reconstruction model is used to reconstruct the target virtual space data, and the reconstruction data in the physical space corresponding to the target virtual space data is obtained; When the error index satisfies the correction conditions, the digital twin model is corrected. The target physical space data and the target virtual space data are obtained in combination with the timing information, the reconstruction data in the physical space corresponding to the target virtual space data is obtained by reconstructing the target virtual space data, and the reconstruction data and the target physical space are obtained according to the reconstruction data and the target physical space. The spatial data determines the error index, thereby avoiding the problem of low error accuracy caused by the original data-level error calculation only based on virtual spatial data and physical spatial data, improving the error accuracy, and then correcting the digital twin model in time. Improved the accuracy of the digital twin model.

Optionally, the above-mentioned modules may be stored in the memory shown in FIG. 3 in the form of software or firmware (Firmware) or solidified in the operating system (Operating System, OS) of the electronic device, and can be executed by the processor in FIG. 3 . . Meanwhile, data required to execute the above-mentioned modules, codes of programs, and the like may be stored in the memory.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may also be implemented in other manners. The apparatus embodiments described above are merely illustrative, for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality and possible implementations of apparatuses, methods and computer program products according to various embodiments of the present application. operate. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code that contains one or more functions for implementing the specified logical function(s) executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It is also noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented in dedicated hardware-based systems that perform the specified functions or actions , or can be implemented in a combination of dedicated hardware and computer instructions.

In addition, each functional module in each embodiment of the present application may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .

The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims (10)

1. a correction method of digital twin model, is characterized in that, described method comprises: According to the preset time sequence length, intercept the physical space data and virtual space data obtained at the current moment and before the current moment to obtain target physical space data and target virtual space data; the physical space data is each component on the industrial equipment operation data, the virtual space data is the operation data of each component in the digital twin model corresponding to the industrial equipment; Input the target virtual space data into the pre-trained reconstruction model, use the reconstruction model to reconstruct the target virtual space data, and obtain the reconstruction in the physical space corresponding to the target virtual space data data; An error index is determined according to the reconstructed data and the target physical space data, and when the error index satisfies a correction condition, the digital twin model is corrected. 2. method according to claim 1, is characterized in that, described reconstruction model is obtained by following steps training: Acquiring a plurality of training samples of preset time series lengths; each of the training samples includes physical space data samples and virtual space data samples corresponding to each moment in the preset time series length; Input all virtual space data samples into a pre-built reconstruction model, use the reconstruction model to reconstruct each virtual space data sample, and obtain the weight in the physical space corresponding to each virtual space data sample. structure data; Calculate the error index corresponding to each training sample according to the reconstructed data in the physical space corresponding to each of the virtual space data samples, and the physical space data samples corresponding to each of the virtual space data samples; When the error index does not meet the model convergence condition, iterative optimization is performed on the parameters of the reconstructed model to obtain a trained reconstructed model. 3. The method according to claim 1, wherein, according to the preset time sequence length, the current moment and the physical space data and virtual space data obtained before the current moment are intercepted to obtain target physical space data and target virtual space data, including: Perform denoising on the physical space data and virtual space data obtained at the current moment and before the current moment to obtain denoised physical space data and virtual space data; The denoised physical space data and virtual space data are intercepted according to the preset time sequence length to obtain target physical space data and target virtual space data. 4. The method according to claim 1, wherein the target virtual space data is input into a pre-trained reconstruction model, and the target virtual space data is reconstructed by using the reconstruction model. to obtain the reconstructed data in the physical space corresponding to the target virtual space data, including: Inputting the target virtual space data into the coding layer of the pre-trained reconstruction model to obtain a context vector corresponding to the target virtual space data; Inputting the context vector and the target virtual space data into the decoding layer of the reconstruction model to obtain initial reconstruction data in the physical space corresponding to the target virtual space data; The initial reconstruction data is input into the nonlinear projection layer of the reconstruction model to obtain reconstruction data in the physical space corresponding to the target virtual space data. 5. The method according to claim 1, wherein the determining an error index according to the reconstructed data and the target physical space data comprises: According to the reconstructed data and the target physical space data, calculate the shape difference and the distortion difference between the target virtual space data and the target physical space data; The error index is calculated from the shape difference and the twist difference. 6 . The method according to claim 5 , wherein calculating the shape difference between the target virtual space data and the target physical space data according to the reconstructed data and the target physical space data. 7 . and warping differences, including: The shape difference and distortion difference between the target virtual space data and the target physical space data are calculated by the following formulas:

in,

Represents the reconstructed data in the physical space corresponding to the target virtual space data,

Characterize the target physical space data,

to characterize shape differences,

Representation Distortion Differences,

represents a smoothing exponent greater than or equal to 0,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular path between the target physical space data,

Represents the reconstructed data in the physical space corresponding to the target virtual space data, and the regular cost matrix of the target physical space data. 7. The method according to claim 5, wherein calculating the error index according to the shape difference and the distortion difference comprises: The error index is calculated by the following formula:

in,

characterizing the error index,

Characterization hyperparameters,

to characterize the shape difference,

Characterize the twist difference. 8. A correction device for a digital twin model, wherein the device comprises: The acquisition module is used for intercepting the physical space data and virtual space data obtained at the current moment and before the current moment according to a preset time sequence length to obtain target physical space data and target virtual space data; the physical space data is The operation data of each component on the industrial equipment, the virtual space data is the operation data of each component in the digital twin model corresponding to the industrial equipment; The processing module is used to input the target virtual space data into the pre-trained reconstruction model, and use the reconstruction model to reconstruct the target virtual space data to obtain the physical data corresponding to the target virtual space data. Reconstructed data in space; A correction module, configured to determine an error index according to the reconstructed data and the target physical space data, and correct the digital twin model when the error index satisfies a correction condition. 9. An electronic device, comprising a processor and a memory, wherein the memory stores a computer program executable by the processor, and the processor can execute the computer program to implement claims 1-7 The method of any one. 10. A computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the method according to any one of claims 1-7 is implemented.

Images