Disturbance prediction method and system for digital twin workshop

Abstract

The invention discloses a disturbance prediction method and system for a digital twin workshop, and the method comprises the steps: obtaining the data of a physical workshop, and building a corresponding workshop digital twin model; extracting multi-modal features of the physical workshop data and the workshop digital twinborn model to obtain physical workshop data features and workshop digital twinborn model features; constructing a disturbance feature library by using the physical workshop data features, the workshop digital twin model features and a historical feature database; setting a disturbance threshold value according to a standard disturbance characteristic case in the disturbance characteristic library; inputting the physical workshop data features and the workshop digital twin model features into a pre-constructed disturbance prediction model, and outputting a disturbance occurrence probability; and comparing the disturbance occurrence probability with the disturbance threshold value to obtain a disturbance prediction result.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial intelligent simulation, and in particular to a disturbance prediction method and system for a digital twin workshop. Background Art

[0002] There are many uncertain factors in the workshop production process, which will cause various disturbances, which have a great impact on the normal production and processing of the workshop, resulting in reduced production efficiency, delayed delivery and supply chain disruptions. The application of advanced manufacturing technology helps the workshop to monitor the production process in real time and predict the time of disturbance. At present, the research on disturbances is mainly divided into two categories, namely prediction before the disturbance occurs, and response and rescheduling after the disturbance occurs. Responding and scheduling after the disturbance occurs will cause the production plan to be delayed and the production efficiency to be reduced. Therefore, the realization of disturbance prediction can ensure the efficient operation of workshop production.

[0003] A digital twin workshop refers to a virtual workshop created using digital twin technology, which realizes real-time synchronization and bidirectional mapping between the physical workshop and the virtual workshop through sensors, the Internet of Things and other technologies. The digital twin workshop can reflect the status of the actual workshop in real time, including equipment operation, production progress, environmental parameters, etc. Therefore, most of the data generated by the digital twin workshop is real-time data, and the existing technology does not disclose how to make full use of this data to mine hidden information and realize real-time prediction of potential disturbances. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a disturbance prediction method and system for a digital twin workshop, which can more comprehensively reflect the workshop status and thus improve the accuracy of disturbance prediction.

[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0006] On the one hand, the present invention provides a disturbance prediction method for a digital twin workshop, comprising:

[0007] Obtain physical workshop data and build a corresponding workshop digital twin model;

[0008] Extracting multimodal features of the physical workshop data and the workshop digital twin model to obtain physical workshop data features and workshop digital twin model features;

[0009] A disturbance feature library is constructed by using the physical workshop data features, the workshop digital twin model features and the historical feature database; a disturbance threshold is set according to standard disturbance feature cases in the disturbance feature library;

[0010] Inputting the physical workshop data features and the workshop digital twin model features into a pre-built disturbance prediction model, and outputting the probability of disturbance occurrence;

[0011] Comparing the disturbance occurrence probability with the disturbance threshold to obtain a disturbance prediction result;

[0012] Wherein, the construction of the disturbance prediction model includes:

[0013] Get the original CNN-GRU-Attention model;

[0014] A memory network layer is added between the input layer and the GRU layer of the original CNN-GRU-Attention model, which is parallel to the CNN layer, to obtain the disturbance prediction model.

[0015] Optionally, also include:

[0016] According to the disturbance prediction results, the digital twin workshop model and the physical workshop data, and the standard disturbance features of the physical workshop data and the disturbance feature library are updated.

[0017] Optionally, before extracting the features of the physical workshop data, the method further includes:

[0018] The physical workshop data is preprocessed by data de-redundancy, data smoothing and data noise reduction.

[0019] Optionally, the extracting multimodal features of the physical workshop data includes:

[0020] For the one-dimensional time series data of physical workshop data, time domain features, frequency domain features and time-frequency domain features are extracted respectively;

[0021] For the multi-source heterogeneous data of physical workshop data, it is divided into multiple one-dimensional data groups, and the features of each one-dimensional data group are extracted using a single channel. The features extracted from each one-dimensional data group are composed into a feature matrix, and the feature matrix is ​​gray-processed to obtain a feature gray-scale image. The feature gray-scale image is used as the multimodal feature of the multivariate heterogeneous data.

[0022] Optionally, the extracting of multimodal features of the workshop digital twin model includes extracting working time features, process features and equipment features of the workshop digital twin model.

[0023] Optionally, the disturbance feature library is a three-layer analytical disturbance feature library of "disturbance position-disturbance type-disturbance feature".

[0024] Optionally, the processing steps of the disturbance prediction model include:

[0025] Inputting the physical workshop data features and the workshop digital twin model features into the CNN layer for secondary feature extraction to obtain low-dimensional compressed features;

[0026] Inputting the physical workshop data feature into the memory network layer, retrieving historical disturbance data related to the physical workshop data feature, and obtaining the historical disturbance feature;

[0027] Inputting the low-dimensional compression features and the historical disturbance features into the GRU layer and the Attention layer in sequence, performing feature sequence processing, feature weight assignment and feature fusion in sequence, and obtaining fused features;

[0028] The fused features are sequentially input into the fully connected layer and the classifier layer for concatenation and classification to obtain the probability of disturbance occurrence.

[0029] Optionally, the training step of the disturbance prediction model includes:

[0030] Generate a perturbation dataset similar to the real perturbation distribution using a generative adversarial network;

[0031] The real disturbance data and the false disturbance data are mixed as a new training data set, and the disturbance prediction model is trained using the new training data set.

[0032] In a second aspect, the present invention provides a disturbance prediction system for a digital twin workshop, comprising:

[0033] Data acquisition module, which acquires physical workshop data and builds the corresponding workshop digital twin model;

[0034] A feature extraction module extracts multimodal features of the physical workshop data and the workshop digital twin model to obtain physical workshop data features and workshop digital twin model features;

[0035] A threshold setting module, which constructs a disturbance feature library by using the physical workshop data features, the workshop digital twin model features and the historical feature database; and sets a disturbance threshold according to a standard disturbance feature case in the disturbance feature library;

[0036] Inputting the physical workshop data features and the workshop digital twin model features into a pre-built disturbance prediction model, and outputting the probability of disturbance occurrence;

[0037] Wherein, the construction of the disturbance prediction model includes:

[0038] Get the original CNN-GRU-Attention model;

[0039] Adding a memory network layer parallel to the CNN layer between the input layer and the GRU layer of the original CNN-GRU-Attention model to obtain the disturbance prediction model;

[0040] The disturbance occurrence probability is compared with the disturbance threshold to obtain a disturbance prediction result.

[0041] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the disturbance prediction method for the digital twin workshop described in the first aspect are implemented.

[0042] Beneficial Effects

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The present invention utilizes the multimodal characteristics of physical workshop data and workshop digital twin models to predict disturbances, which can more comprehensively reflect the workshop status and thus improve the accuracy of disturbance prediction; introduces a memory network to store and retrieve historical disturbance data, and provides historical information related to the current workshop status, thereby helping the model to better understand the state changes of the workshop and improve the prediction accuracy of the model; uses a generative adversarial network to generate false disturbance data with a distribution similar to that of real disturbance data, expands the training data set, thereby solving the data imbalance problem and improving the generalization ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Shown is a schematic flow chart of a disturbance prediction method for a digital twin workshop of the present invention in one embodiment;

[0046] Figure 2 FIG. 1 is a flow chart of a multimodal feature extraction method for physical workshop data in an embodiment of the present invention;

[0047] Figure 3 FIG. 1 is a flow chart of a multimodal feature extraction method for a digital twin model of a workshop according to an embodiment of the present invention;

[0048] Figure 4 FIG. 1 is a schematic diagram showing the principle of the attention mechanism of the present invention in one embodiment;

[0049] Figure 5 Shown is a schematic diagram of the structure of a disturbance prediction model in one embodiment of the present invention;

[0050] Figure 6 Shown is a schematic structural diagram of the hardware equipment of the milling cutter wear disturbance prediction experimental platform in one embodiment of the present invention. DETAILED DESCRIPTION

[0051] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0052] The term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0053] Example 1

[0054] like Figure 1 As shown, the present invention introduces a disturbance prediction method for a digital twin workshop, which specifically includes the following steps:

[0055] Acquire physical workshop data and build the corresponding workshop digital twin model.

[0056] Physical workshop data includes data on equipment, personnel, materials, production lines, etc. in the workshop. The equipment operation status data and workshop production parameter data are uploaded to the next layer for data processing mainly through data collection methods. Data processing mainly includes three parts: data storage, data preprocessing and feature extraction. Specifically, they include:

[0057] First, after receiving multi-source heterogeneous data at the physical workshop level, each sensing device stores it in a shared database, providing data support for building a workshop digital twin model and disturbance prediction;

[0058] Secondly, preprocessing operations of data redundancy removal, data smoothing, and data noise reduction are performed on the physical workshop data;

[0059] Finally, the multimodal features of the preprocessed physical workshop data and the workshop digital twin model are extracted to obtain the physical workshop data features and the workshop digital twin model features; on the one hand, multimodal feature extraction is performed on the workshop operation status data and the workshop production parameter data to provide available data for disturbance prediction; on the other hand, multimodal feature extraction is performed on the workshop digital twin model to complete the state feature extraction of the digital twin workshop disturbance.

[0060] in,

[0061] like Figure 2 As shown, there are multiple types of data in the physical workshop, and different types of data require different feature extraction methods. The multimodal features of the physical workshop data are extracted including:

[0062] For the one-dimensional time series data of the physical workshop data, the time domain features, frequency domain features and time-frequency domain features are extracted respectively, and the obtained feature data are input into the convolutional neural network (CNN) in the disturbance prediction model for further feature extraction.

[0063] For the two-dimensional image data of the physical workshop data, since the CNN in the disturbance prediction model has the function of image feature extraction, the image data can be directly input into the CNN in the disturbance prediction model for feature extraction.

[0064] For the multi-source heterogeneous data of physical workshop data, it is divided into multiple one-dimensional data groups, and the features of each one-dimensional data group are extracted using a single channel. The features extracted from each one-dimensional data group are composed into a feature matrix. Since the feature matrix of the constructed multi-source heterogeneous data has a large amount of data, the feature matrix is ​​gray-processed to obtain a feature gray-scale image. The obtained feature gray-scale image is used as the multimodal feature of the multivariate heterogeneous data and is input into the CNN in the disturbance prediction model for feature extraction.

[0065] like Figure 3 As shown, the multimodal features of the digital twin model of the workshop are extracted, including the working time features, process features and equipment features of the digital twin model of the workshop. In the workshop production process, the processing time of the workpiece will be affected due to changes in processing time, machine failures and other reasons. Therefore, the production status of the workshop is characterized by extracting the processing time of each link of the workpiece. When there is a deviation between the actual processing process and the theoretical processing process, it may cause a disturbance of inconsistent processes. In the digital twin workshop, the workstation will issue an error instruction because the workpiece has not arrived or has arrived incorrectly. Therefore, by extracting the theoretical and actual information of the process, the difference between the two is used to reflect the disturbance caused by the workshop order. During the processing of the equipment, the status of the parts will change with the processing time. By extracting relevant feature data, the equipment status is predicted and monitored. For example, by collecting data such as the wear bandwidth of the flank face and the stress distribution of the flank face of the virtual model of the milling cutter, the disturbance prediction of the milling cutter is realized.

[0066] The disturbance feature library is constructed by using the physical workshop data features, the workshop digital twin model features and the historical feature database. The establishment of the disturbance feature library first requires sorting out and classifying the workshop disturbances and clarifying the manifestation of disturbance features. Based on the type of data collected in the workshop and the problems that may arise in the production process, the disturbance events are mainly divided into equipment, task and storage categories. According to the location of the workshop disturbance, disturbance characteristics and other information, the disturbances are classified and a three-layer analytical disturbance feature library of "disturbance location-disturbance type-disturbance characteristics" is established.

[0067] A disturbance threshold is set according to a standard disturbance feature case in the disturbance feature library.

[0068] The physical workshop data features and the workshop digital twin model features are input into a pre-built disturbance prediction model to output the probability of disturbance occurrence.

[0069] Wherein, the construction of the disturbance prediction model includes:

[0070] Get the original CNN-GRU-Attention model;

[0071] A memory network layer is added between the input layer and the GRU layer of the original CNN-GRU-Attention model, which is parallel to the CNN layer, to obtain the disturbance prediction model.

[0072] A perturbation data set similar to the real perturbation distribution is generated by using a generative adversarial network; the real perturbation data and the false perturbation data are mixed as a new training data set, and the perturbation prediction model is trained using the new training data set.

[0073] like Figure 5 The figure shows a schematic diagram of the structure of the disturbance prediction model. In this embodiment, the disturbance prediction model is a convolutional neural network (CNN-GRU-Attention) including a memory network and a gated recurrent unit, and uses an attention mechanism to dynamically allocate the weights of input features.

[0074] like Figure 4 The following is a schematic diagram of the principle of the attention mechanism. The weighting of the attention mechanism includes three steps:

[0075] Step 1: Calculate the similarity between the Query and the Key. By using different functions and calculation methods, the similarity between the Query and a Key is calculated. The method in this embodiment includes:

[0076] Calculate the dot product of the two vectors, cosine similarity, or use other neural networks to evaluate the similarity. The formula is expressed as:

[0077] Dot product:

[0078] ;

[0079] Cosine Similarity:

[0080] ;

[0081] Feedforward Artificial Neural Network (MLP, Multilayer Perceptron):

[0082] .

[0083] Step 2: Normalize the scores. The calculated similarities are normalized through the Softmax function to make them into probability distributions, ensuring that the sum of the weights is 1. The weight coefficient corresponding to the Value is calculated, and the formula is expressed as:

[0084] ;

[0085] In the formula, Indicates the length of the input data.

[0086] Step 3: Perform weighted summation on the weight and value. Use the attention weight normalized by the Softmax function to perform weighted summation on the value to obtain the final attention value, expressed as:

[0087] .

[0088] The disturbance prediction model consists of five parts: input layer, CNN feature extraction layer, memory network layer, GRU-Attention layer, fully connected layer and output layer.

[0089] The processing steps of the disturbance prediction model include:

[0090] Input layer: input the physical workshop data features and the workshop digital twin model features into the CNN feature extraction layer; input the physical workshop data features into the memory network layer.

[0091] CNN layer: performs secondary feature extraction on the physical workshop data features and the workshop digital twin model features to obtain low-dimensional compressed features; through multiple convolution and pooling operations, the input data is represented as low-dimensional compressed data and input into the GRU-Attention layer.

[0092] Memory network layer: retrieve historical disturbance data related to the physical workshop data characteristics to obtain historical disturbance characteristics.

[0093] GRU-Attention layer: The low-dimensional compression features and the historical disturbance features are processed in sequence, feature weights are assigned, and features are fused to obtain fused features; the rules between feature data are fully learned, and weights are introduced to dynamically capture important information in the sequence, so as to better understand the long-term dependency of the workshop status and provide more accurate information for the prediction task. The processing flow of the GRU-Attention layer is as follows:

[0094] Step 1: Calculate the update gate and reset gate of GRU. According to the input of the current time step and the hidden state of the previous time step, calculate the update gate and reset gate The activation value is expressed as:

[0095] ;

[0096] ;

[0097] in, Represents the Sigmoid activation function, which is used to update and reset the gate operations and map the input value to between 0 and 1; Represents the weight matrix during the update process; Represents the time step The hidden state of Represents the input data of the sequence at time step t; represents the bias vector of the update process; The weight matrix representing the reset process; Represents the bias vector for the reset process.

[0098] Step 2: Update candidate hidden states. Use the reset gate and the input of the current time step to calculate the candidate hidden state for subsequent updates, expressed as:

[0099] ;

[0100] in, represents the hyperbolic tangent activation function; represents the hidden weight matrix; represents the hidden bias vector.

[0101] Step 3: Update the hidden state. Use the input gate and the candidate hidden state, combined with the hidden state of the previous time step, to calculate the hidden state of the current time step. , expressed as:

[0102] .

[0103] Step 4: After completing the processing of all time steps, the final hidden state As the output of GRU, we get a vector set , and input to the attention layer.

[0104] Step 5: In the attention layer, calculate the hidden state of each time step The corresponding attention score , expressed as:

[0105] ;

[0106] in, The weight matrix representing the attention; represents a learnable weight vector; Represents the bias vector.

[0107] Step 6: Use the Softmax activation function to normalize the attention score to obtain the attention weight , expressed as:

[0108] ;

[0109] in, Represents the number of all attention mechanisms. If it is a two-headed attention mechanism, then =2.

[0110] Step 7: Use the obtained attention weights to calculate the context attention vector , the attention vector Weighted sum of the input sequence.

[0111] .

[0112] Fully connected layer: concatenate the fused features, and connect a classifier after the fully connected layer to classify the obtained feature vector to obtain the probability of disturbance. Since the problem dealt with in this embodiment is a binary classification problem (disturbance event is 1, normal event is 0), the Sigmoid activation function is used to map the output value to a probability value between 0 and 1, which is the probability of disturbance occurrence. The Sigmoid activation function is expressed as:

[0113] .

[0114] Output layer: Set the perturbation threshold according to the standard perturbation feature cases in the constructed perturbation feature library .

[0115] Compare the disturbance occurrence probability with the disturbance threshold to obtain a disturbance prediction result. When , the output is 1 and it is judged as a disturbance event; otherwise, the output is 0 and it is judged as a normal event.

[0116] According to the disturbance prediction results, the digital twin workshop model and the physical workshop data, and the standard disturbance features of the physical workshop data and the disturbance feature library are updated.

[0117] Example 2

[0118] This embodiment introduces a disturbance prediction system for a digital twin workshop, which specifically includes:

[0119] Data acquisition module, which acquires physical workshop data and builds the corresponding workshop digital twin model;

[0120] A feature extraction module extracts multimodal features of the physical workshop data and the workshop digital twin model to obtain physical workshop data features and workshop digital twin model features;

[0121] A threshold setting module, which constructs a disturbance feature library by using the physical workshop data features, the workshop digital twin model features and the historical feature database; and sets a disturbance threshold according to a standard disturbance feature case in the disturbance feature library;

[0122] Inputting the physical workshop data features and the workshop digital twin model features into a pre-built disturbance prediction model, and outputting the probability of disturbance occurrence;

[0123] Wherein, the construction of the disturbance prediction model includes:

[0124] Get the original CNN-GRU-Attention model;

[0125] Adding a memory network layer parallel to the CNN layer between the input layer and the GRU layer of the original CNN-GRU-Attention model to obtain the disturbance prediction model;

[0126] The disturbance occurrence probability is compared with the disturbance threshold to obtain a disturbance prediction result.

[0127] Example 3

[0128] like Figure 6 As shown, this embodiment introduces a specific experimental design of a disturbance prediction method for a digital twin workshop, including the following steps:

[0129] The hardware equipment of the milling cutter wear disturbance prediction experimental platform mainly includes CNC milling machine, vibration sensor, rotary cutting dynamometer, tool appearance detector and data acquisition card, etc. Figure 6 As shown in the figure, the selected equipment models are as follows: the model of the CNC milling machine is DMU-70V; the model of the vibration sensor is Vtall-S203L-1 three-axis vibration sensor, which supports Modbus and RS485 communication protocols, and the sampling frequency is set to 20kHz; the rotary cutting dynamometer is a rotary multi-component cutting dynamometer model 9171A, which only needs to be installed on the spindle to measure the forces in the three axes of the milling cutter , the sampling frequency is set to 20kHz; the model of the tool appearance detector is YK-D1800B-2, which is used for tool wear measurement; the model of the data acquisition card is PCI5650-D0, and the sampling frequency is set to 20kHz; the milling cutter material is high-speed steel with a hardness of 63; the workpiece material is 45 steel.

[0130] According to the parameter range of the milling cutter and the milling machine, three cutting tests were designed. The specific experimental parameters are shown in Table 1. In the experimental processing, the down-milling processing method was adopted, the single tool path was 200 mm, and 200 tool passes were performed under each working condition. When each single-stage milling process was completed, it was numbered, and the machine tool vibration signal, milling force data, and tool flank wear status were measured and recorded.

[0131] Table 1 Experimental parameters of prop wear

[0132]

[0133] Tool wear stages are usually divided into initial wear, normal wear and severe wear. In actual processing, since the wear rate of the tool may be different under different working conditions, the division of the wear stage will also be different. The division of the wear stage is mainly based on the trend of the tool wear curve and the wear change rate. Therefore, this paper divides the wear stage according to the change trend of the tool back face wear curve and the change rate of the front and back wear values. First, the approximate segmentation area is determined by the wear curve, and then the position with the largest change rate of the front and back wear values ​​is selected as the final division stage position, and severe wear is defined as a disturbance event. The wear type and sample number under the three working conditions are shown in Table 2.

[0134] Table 2 Tool wear types and corresponding sample numbers under three working conditions

[0135]

[0136] The milling cutter wear disturbance prediction algorithm was run on a computer with Windows 11 system. The specific software and hardware parameters are shown in Table 3.

[0137] Table 3. Software and hardware parameters of the milling cutter wear disturbance prediction algorithm platform

[0138]

[0139] In order to verify the effectiveness of the disturbance prediction model proposed in this paper, the prediction results are compared with the CNN, CNN-LSTM and CNN-GRU models, as shown in Table 4. It can be seen from the table that under three different working conditions, the four performance indicators of the model in this embodiment all perform best. In particular, under working condition OC3, the performance of the model in this embodiment reached the best, with accuracy, correctness, recall and values ​​of 97.5%, 95.38%, 96.88% and 96.12% respectively. This corresponds to the case where the loss value is the lowest under working condition OC3. Compared with the CNN-GRU model without the attention mechanism, the prediction accuracy and correctness are improved by 0.84% ​​and 4.16% on average.

[0140] Table 4 Comparison of experimental results

[0141]

[0142] Example 3

[0143] This embodiment introduces a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the disturbance prediction method for the digital twin workshop described in Embodiment 1 or 3 are implemented.

[0144] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0145] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0146] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0147] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0148] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A disturbance prediction method for a digital twin workshop, characterized in that: include: Obtain physical workshop data and build a corresponding workshop digital twin model; Extracting multimodal features of the physical workshop data and the workshop digital twin model to obtain physical workshop data features and workshop digital twin model features; A disturbance feature library is constructed by using the physical workshop data features, the workshop digital twin model features and the historical feature database; a disturbance threshold is set according to standard disturbance feature cases in the disturbance feature library; Inputting the physical workshop data features and the workshop digital twin model features into a pre-built disturbance prediction model, and outputting the probability of disturbance occurrence; Comparing the disturbance occurrence probability with the disturbance threshold to obtain a disturbance prediction result; Wherein, the construction of the disturbance prediction model includes: Get the original CNN-GRU-Attention model; A memory network layer is added between the input layer and the GRU layer of the original CNN-GRU-Attention model, which is parallel to the CNN layer, to obtain the disturbance prediction model.

2. The disturbance prediction method of the digital twin workshop according to claim 1 is characterized in that: Also includes: According to the disturbance prediction results, the digital twin workshop model and the physical workshop data, and the standard disturbance features of the physical workshop data and the disturbance feature library are updated.

3. The disturbance prediction method of the digital twin workshop according to claim 1 is characterized in that: Before extracting the features of the physical workshop data, the method further includes: The physical workshop data is preprocessed by data de-redundancy, data smoothing and data noise reduction.

4. The disturbance prediction method of the digital twin workshop according to claim 1 is characterized in that: The extracting multimodal features of the physical workshop data includes: For the one-dimensional time series data of physical workshop data, time domain features, frequency domain features and time-frequency domain features are extracted respectively; For the multi-source heterogeneous data of physical workshop data, it is divided into multiple one-dimensional data groups, and the features of each one-dimensional data group are extracted using a single channel. The features extracted from each one-dimensional data group are composed into a feature matrix, and the feature matrix is ​​gray-processed to obtain a feature gray-scale image. The feature gray-scale image is used as the multimodal feature of the multivariate heterogeneous data.

5. The disturbance prediction method of the digital twin workshop according to claim 1 is characterized in that: The extracting of multimodal features of the workshop digital twin model includes extracting working time features, process features and equipment features of the workshop digital twin model.

6. The disturbance prediction method of the digital twin workshop according to claim 1 is characterized in that: The disturbance feature library is a three-layer analytical disturbance feature library of "disturbance position-disturbance type-disturbance feature".

7. The disturbance prediction method of the digital twin workshop according to claim 1 is characterized in that: The processing steps of the disturbance prediction model include: Inputting the physical workshop data features and the workshop digital twin model features into the CNN layer for secondary feature extraction to obtain low-dimensional compressed features; Inputting the physical workshop data feature into the memory network layer, retrieving historical disturbance data related to the physical workshop data feature, and obtaining the historical disturbance feature; Inputting the low-dimensional compression features and the historical disturbance features into the GRU layer and the Attention layer in sequence, performing feature sequence processing, feature weight assignment and feature fusion in sequence, and obtaining fused features; The fused features are sequentially input into the fully connected layer and the classifier layer, and are sequentially concatenated and classified to obtain the probability of disturbance occurrence.

8. The disturbance prediction method of the digital twin workshop according to claim 1 is characterized in that: The training steps of the disturbance prediction model include: Generate a perturbation dataset similar to the real perturbation distribution using a generative adversarial network; The real disturbance data and the false disturbance data are mixed as a new training data set, and the disturbance prediction model is trained using the new training data set.

9. A disturbance prediction system for a digital twin workshop, characterized in that: include: Data acquisition module, which acquires physical workshop data and builds the corresponding workshop digital twin model; A feature extraction module extracts multimodal features of the physical workshop data and the workshop digital twin model to obtain physical workshop data features and workshop digital twin model features; A threshold setting module, which constructs a disturbance feature library by using the physical workshop data features, the workshop digital twin model features and the historical feature database; and sets a disturbance threshold according to a standard disturbance feature case in the disturbance feature library; Inputting the physical workshop data features and the workshop digital twin model features into a pre-built disturbance prediction model, and outputting the probability of disturbance occurrence; Wherein, the construction of the disturbance prediction model includes: Get the original CNN-GRU-Attention model; Adding a memory network layer parallel to the CNN layer between the input layer and the GRU layer of the original CNN-GRU-Attention model to obtain the disturbance prediction model; The disturbance occurrence probability is compared with the disturbance threshold to obtain a disturbance prediction result.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the disturbance prediction method for a digital twin workshop described in any one of claims 1 to 8 are implemented.

Images