Production equipment operation state monitoring and early warning method and device based on digital twin model

Abstract

This invention discloses a method and device for monitoring and early warning of the operating status of production equipment based on a digital twin model, relating to the field of production equipment status monitoring technology. The method includes: first, constructing a digital twin model corresponding to the production equipment, and collecting sample data of the production equipment under various faults through the digital twin model; then, constructing a production equipment fault early warning model using a deep learning model, and using GEGWO to enable the production equipment fault early warning model to learn from the sample data, obtaining a pre-trained production equipment fault early warning model, avoiding getting trapped in local optima and improving training effect; next, running the digital twin model, and analyzing real-time production equipment operating data through the pre-trained production equipment fault early warning model, and visually warning the corresponding production equipment in the digital twin model when the fault monitoring result is an abnormal category, thus achieving early fault warning and shortening fault response time.

Description

Technical Field

[0001] This invention relates to the field of production equipment condition monitoring technology, and more specifically, to a method and device for monitoring and early warning of production equipment operation status based on a digital twin model. Background Technology

[0002] In modern industrial production environments, the operational reliability of production equipment directly impacts product quality, production efficiency, and production safety. Traditional equipment condition monitoring methods primarily rely on installing sensors at critical equipment components to collect physical signals such as vibration and temperature. Alarms are triggered when these signals exceed preset thresholds. However, this threshold-based approach has significant shortcomings: firstly, signal values ​​at a single point in time are insufficient to reflect the evolutionary trend of equipment failures; secondly, different failure types may exhibit similar abnormal signals, leading to frequent false alarms or missed alarms. In recent years, the rise of digital twin technology has provided a new approach to equipment condition monitoring. By constructing a digital mirror image of the equipment, real-time data interaction between the physical equipment and the virtual model can be achieved. However, existing digital twin-based monitoring methods still face two main problems: firstly, while digital twin models can reflect equipment operating data in real time, they lack the ability to perform in-depth analysis of this data and predict failures; secondly, deep learning models used for fault identification are prone to getting trapped in local optima during parameter optimization, resulting in poor model training performance and consequently affecting the accuracy of fault warnings. Summary of the Invention

[0003] The present invention provides a method and device for monitoring and early warning of the operating status of production equipment based on a digital twin model, which aims to solve the problems of lack of ability to predict faults and / or easy getting trapped in local optima in the existing technology.

[0004] The first aspect of this invention provides a method for monitoring and early warning of the operating status of production equipment based on a digital twin model, comprising:

[0005] Construct a digital twin model corresponding to the production equipment, and establish a mapping relationship between the digital twin model and the production equipment so that the data of the digital twin model and the production equipment are synchronized in real time;

[0006] The digital twin model is used to collect sample data of the production equipment under various faults; the sample data refers to the sample production equipment operation data and its corresponding fault types at multiple consecutive sampling time points;

[0007] A production equipment fault early warning model is constructed using a deep learning model, and GEGWO is used to enable the production equipment fault early warning model to learn from the sample data to obtain a pre-trained production equipment fault early warning model.

[0008] Run the digital twin model, and analyze the real-time production equipment operation data collected by the digital twin model through the pre-trained production equipment fault early warning model to determine the fault monitoring results;

[0009] If the fault monitoring result is an abnormal category, the corresponding production equipment will be visualized and given an early warning in the digital twin model.

[0010] Furthermore, the digital twin model is used to collect sample data of the production equipment under various fault conditions, including:

[0011] The production equipment operation data is collected through the digital twin model; the production equipment operation data includes vibration, temperature, current, and voltage.

[0012] For any given fault, the fault type is collected. At the same time, based on a preset data sampling frequency, the operating data of the production equipment at N sampling time points before the fault occurs are collected to obtain the sample production equipment operating data and its corresponding fault type; N is the number of preset data sampling points.

[0013] Furthermore, a production equipment fault early warning model is constructed using a deep learning model, including: constructing a production equipment fault early warning model using a CNN-LSTM model.

[0014] Furthermore, GEGWO is used to enable the production equipment fault early warning model to learn from the sample data, thereby obtaining a pre-trained production equipment fault early warning model, including:

[0015] Initialize the gray wolf pack;

[0016] The sample data was used to divide the gray wolf pack into Wolf, Wolf, Wolves and ordinary gray wolves;

[0017] Regarding the above Wolf, wolves and Wolves employ an exploration strategy based on gradient direction correction. Wolf, wolves and The wolf updates itself and obtains the updated version. Wolf, wolves and Wolf;

[0018] For ordinary gray wolves, a hierarchical weighted encirclement strategy is used to control the ordinary gray wolves to track prey, and the ordinary gray wolves after tracking the prey are obtained.

[0019] After tracking its prey, the ordinary gray wolf, and the updated... Wolf, wolves and The wolves reunited as a gray wolf pack;

[0020] For any individual gray wolf in the gray wolf pack, a probability distribution exploration strategy is used to perform a global greedy update on the individual gray wolf, resulting in the gray wolf individual after the global greedy update.

[0021] Determine the training progress; the training progress refers to whether the training is completed or not.

[0022] If the training progress is incomplete, based on the gray wolf individuals after the global greedy update, return to the step of dividing the gray wolf group and proceed to the next training.

[0023] When the training progress is complete, the global optimal parameters of the production equipment fault early warning model are determined by the gray wolf individuals after global greedy update, and the pre-trained production equipment fault early warning model is obtained.

[0024] Furthermore, the sample data is used to divide the gray wolf pack into... Wolf, Wolf, Wolves and common gray wolves, including:

[0025] The sample production equipment operation data is used as the input data for the production equipment fault early warning model to obtain the predicted output data of the production equipment fault early warning model;

[0026] The fault types corresponding to the sample production equipment operation data are constructed into expected output data, and the loss function value is obtained based on the expected output data and the predicted output data.

[0027] The gray wolf individuals with the smallest, second smallest, and third smallest loss function values ​​are respectively considered as... Wolf, wolves and The remaining gray wolves were treated as ordinary gray wolves.

[0028] Furthermore, an exploration strategy based on gradient direction correction is adopted to... Wolf, wolves and The wolf updates itself and obtains the updated version. Wolf, wolves and Wolves, including:

[0029] Generate random gray wolf individuals with each dimension parameter randomly set to 0.1 or -0.1. ;

[0030] Based on the random gray wolf individuals The gradient direction information is obtained as follows:

[0031] ;

[0032] in, express Wolf, wolves or Wolf, Indicates gradient direction information. express The fitness of the solution space. express The fitness of the solution space is 1 / (loss function value + 0.0001).

[0033] The number of training iterations has been determined. The gradient update control coefficients are obtained based on the number of training iterations:

[0034] ;

[0035] in, This represents the gradient update control coefficient. This represents the base coefficient, and is set to 0.01. This represents the change in the coefficient, and is set to 0.005. Represents pi (π). This indicates the maximum number of training iterations preset.

[0036] Based on the gradient direction information and the gradient update control coefficients, obtain the updated... Wolf, wolves and Wolves are:

[0037] ;

[0038] in, Indicates the updated version Wolf, wolves or Wolf, Represents a symbolic function. This represents a single gray wolf whose elements are all 1.

[0039] Furthermore, a hierarchical weighted encirclement strategy is employed to control ordinary gray wolves tracking prey, resulting in the following ordinary gray wolves after tracking prey:

[0040] according to Wolf, wolves and The wolf obtains the first, second, and third guiding information as follows:

[0041] ;

[0042] ;

[0043] ;

[0044] in, This represents the m-th ordinary gray wolf. express The corresponding first guidance information, express The corresponding second guidance information, This indicates the third guiding information. express Wolf, express Wolf, express Wolf, This represents the linear dynamic coefficient, which decreases linearly from 2 to 0 as the training process progresses; Represents the first random number between (0,1). This represents the second random number between (0,1). This represents a third random number between (0,1). This represents the fourth random number between (0, 1);

[0045] A hierarchical weighting system is used to enable ordinary gray wolves to track prey based on the first, second, and third guidance information. The ordinary gray wolf after tracking the prey is obtained as follows:

[0046] ;

[0047] in, This indicates the m-th ordinary gray wolf after tracking its prey. This represents the first-level weight, and is set to the range (0, 0.5). This represents the second-level weight, and is set to the range (0, 0.3). This represents the third-level weight, and is set to the range (0, 0.2).

[0048] Furthermore, a probability distribution exploration strategy is used to perform a global greedy update on the individual gray wolves, resulting in the following gray wolf individuals after the global greedy update:

[0049] During the first training process, the mean vector and standard deviation vector are initialized; each element in the mean vector is the mean of all gray wolf individuals in the same dimension; each element in the standard deviation vector is the standard deviation of all gray wolf individuals in the same dimension.

[0050] In the second and subsequent training processes, the updated mean vector and updated standard deviation vector from the previous training process are used to obtain the perturbation individuals corresponding to the individual gray wolf:

[0051] ;

[0052] in, Let represent the d-th element of the k-th gray wolf individual, where k = 1, 2, ..., K, K represents the total number of gray wolves, and d = 1, 2, ..., D, D represents the total number of elements among the gray wolves. This represents the d-th dimension element of the perturbed individual. Represents the Gaussian error function. This represents the d-th element of the standard deviation vector after the last training iteration. This represents the d-th element of the mean vector after the last training iteration;

[0053] Among the gray wolf individuals and their corresponding perturbation individuals, the individuals with smaller loss function values ​​are designated as gray wolf individuals after global greedy updates, while the individuals with larger loss function values ​​are designated as inferior individuals.

[0054] The inferior individuals are used for updating the mean vector and standard deviation vector, and the update rule is as follows:

[0055] ;

[0056] ;

[0057] in, Let represent the d-th element of the updated mean vector. This represents the d-th element of the updated standard deviation vector. This represents the d-th element of the mean vector before the update. This represents the d-th element of the standard deviation vector before the update. Let represent the d-th element of the individual gray wolf after the global greedy update. The d-th element represents a substandard individual.

[0058] Furthermore, if the fault monitoring result is an abnormal category, the corresponding production equipment will be visualized and given an early warning in the digital twin model, including:

[0059] If the fault monitoring result is abnormal, the production equipment model will be rendered with different colors, and a flashing or highlighting effect will be superimposed based on the fault monitoring result in the digital twin model.

[0060] A second aspect of the present invention provides a production equipment operation status monitoring and early warning device based on a digital twin model, comprising:

[0061] The digital twin model construction module is used to construct a digital twin model corresponding to the production equipment and establish a mapping relationship between the digital twin model and the production equipment so that the data of the digital twin model and the production equipment are synchronized in real time.

[0062] The sample data acquisition module is used to collect sample data of the production equipment under various faults through the digital twin model; the sample data refers to the production equipment operation data and its corresponding fault types at multiple consecutive sampling time points;

[0063] The fault warning model training module is used to construct a production equipment fault warning model using a deep learning model, and to use GEGWO to enable the production equipment fault warning model to learn the sample data, thereby obtaining a pre-trained production equipment fault warning model.

[0064] The real-time online monitoring module is used to run the digital twin model and analyze the real-time production equipment operation data collected by the digital twin model through the pre-trained production equipment fault early warning model to determine the fault monitoring results.

[0065] The monitoring result visualization module is used to visualize and warn the corresponding production equipment in the digital twin model when the fault monitoring result is an abnormal category.

[0066] Beneficial effects:

[0067] This invention provides a method and apparatus for monitoring and early warning of the operating status of production equipment based on a digital twin model. The method includes: first, constructing a digital twin model corresponding to the production equipment and collecting sample data of the production equipment under various faults through the digital twin model; then, constructing a fault early warning model of the production equipment using a deep learning model and using GEGWO to train the fault early warning model of the production equipment to learn from the sample data, thereby obtaining a pre-trained fault early warning model of the production equipment, avoiding getting trapped in local optima and improving the training effect; next, running the digital twin model and analyzing the real-time operating data of the production equipment collected by the digital twin model through the pre-trained fault early warning model of the production equipment to determine the fault monitoring result; finally, when the fault monitoring result is an abnormal category, visually warning the corresponding production equipment in the digital twin model, realizing early warning of faults and shortening the fault response time. Attached Figure Description

[0068] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0069] Figure 1 This is a schematic diagram of a method for monitoring and early warning of the operating status of production equipment based on a digital twin model, proposed in an embodiment of the present invention.

[0070] Figure 2 This is a schematic diagram of a production equipment operation status monitoring and early warning device based on a digital twin model, according to an embodiment of the present invention.

[0071] Figure labeling: 201-Digital twin model construction module, 202-Sample data acquisition module, 203-Fault early warning model training module, 204-Real-time online monitoring module, 205-Monitoring result visualization module. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] like Figure 1 As shown, this embodiment of the invention provides a method for monitoring and early warning of the operating status of production equipment based on a digital twin model, including:

[0074] S101. Construct a digital twin model corresponding to the production equipment, and establish a mapping relationship between the digital twin model and the production equipment so that the data of the digital twin model and the production equipment are synchronized in real time.

[0075] For example, a virtual model with the same geometry as the physical machine tool can be created in a digital twin platform, including the machine tool body, spindle system, feed system, and other major components. Virtual sensors are then placed at corresponding locations on the virtual model to receive data collected by actual sensors on the physical machine tool. A data mapping relationship between the digital twin model and the physical machine tool is established via the OPC UA communication protocol, with a data synchronization period of 50 milliseconds to ensure that the digital twin model can reflect the real-time operating status of the physical machine tool.

[0076] S102. Collect sample data of the production equipment under various faults through the digital twin model; the sample data refers to the sample production equipment operation data and its corresponding fault types at multiple consecutive sampling time points.

[0077] In some possible embodiments, the digital twin model is used to collect sample data of the production equipment under various fault conditions, including:

[0078] The production equipment operation data is collected through the digital twin model; the production equipment operation data includes vibration, temperature, current, and voltage.

[0079] For any given fault, the fault type is collected. At the same time, based on a preset data sampling frequency, the operating data of the production equipment at N sampling time points before the fault occurs are collected to obtain the sample production equipment operating data and its corresponding fault type; N is the number of preset data sampling points.

[0080] For example, during the continuous operation of the machine tool, various fault events are recorded. This embodiment collects data on four typical faults: bearing wear, imbalance, misalignment, and foundation loosening, with 50 independent events collected for each fault. For each fault event, operating data within 2 seconds prior to the fault occurrence is collected at a sampling frequency of 100Hz, resulting in N = 200 sampling points. Data in four dimensions is collected at each sampling point: vibration amplitude, temperature, current, and voltage, thus each sample is a 200×4 two-dimensional matrix. This matrix is ​​used as the sample production equipment operating data, and the corresponding fault type is encoded as a one-hot vector as the expected output data, together forming the training sample set. It is worth noting that sample production equipment operating data under normal operating conditions should also be collected, and the corresponding label should be set to "normal type." Using data from both normal and faulty operating conditions during training allows the production equipment fault warning model to distinguish whether the production equipment is operating normally.

[0081] S103. A production equipment fault early warning model is constructed using a deep learning model, and GEGWO is used to enable the production equipment fault early warning model to learn the sample data, thereby obtaining a pre-trained production equipment fault early warning model.

[0082] In some possible embodiments, a production equipment fault early warning model is constructed using a deep learning model, including: constructing a production equipment fault early warning model using a CNN-LSTM (Convolutional Neural Network-Long Short-Term Memory Network) model.

[0083] For example, a CNN-LSTM combined model can be used, where CNN is used to extract local correlation features between data from various sensors, and LSTM is used to learn temporal dependencies over 200 time steps.

[0084] S104. Run the digital twin model and analyze the real-time production equipment operation data collected by the digital twin model through the pre-trained production equipment fault early warning model to determine the fault monitoring results;

[0085] The pre-trained production equipment fault early warning model is deployed into the digital twin system. The digital twin model synchronizes the real-time production equipment operation data of the physical machine tool in real time. Every time 200 consecutive sampling points are collected (i.e., a 2-second data window), the data matrix corresponding to the 200×4 real-time production equipment operation data is input into the fault early warning model for analysis.

[0086] S105. If the fault monitoring result is an abnormal category, then the corresponding production equipment will be visualized and given an early warning in the digital twin model.

[0087] In some possible embodiments, if the fault monitoring result is an abnormal category, the corresponding production equipment will be visualized and given an early warning in the digital twin model, including:

[0088] If the fault monitoring result is abnormal, the production equipment model will be rendered with different colors, and a flashing or highlighting effect will be superimposed based on the fault monitoring result in the digital twin model.

[0089] When the model output is an abnormal category, the digital twin platform immediately activates a visual early warning mechanism. For example, when a bearing wear fault is detected, the machine tool spindle component in the digital twin interface is rendered orange-yellow, and / or has a flashing effect at a frequency of 3 times per second; when a misalignment fault is detected, the coupling component is rendered red and / or has a bright halo overlaid; for faults of higher severity, a detailed fault information panel will automatically pop up, displaying the fault type, the time of occurrence, and the suggested handling measures corresponding to the preset fault type.

[0090] This invention uses a digital twin model to collect equipment operation data from multiple consecutive sampling time points before a failure occurs. Compared with traditional methods that rely solely on single-point thresholds, this approach captures the complete evolution of equipment from normal operation to failure. The time-series-based sample construction method allows deep learning models to learn pre-failure features, thus issuing warnings before the failure fully occurs and enabling predictive maintenance.

[0091] Optionally, after each preset model update cycle, the pre-trained production equipment fault early warning model can be updated using real-world data and faults to ensure that the production equipment fault early warning model can adapt to real-world production scenarios and guarantee recognition accuracy.

[0092] In some possible embodiments, GEGWO (Global Enhanced Grey Wolf Optimizer) is used to enable the production equipment fault early warning model to learn the sample data, thereby obtaining a pre-trained production equipment fault early warning model, including:

[0093] Initialize the gray wolf population; for example, the hyperparameters of the production equipment fault early warning model can be randomly generated within their upper and lower limits and formed into a vector, thereby obtaining individual gray wolves. Repeatedly obtaining multiple gray wolf individuals constitutes a gray wolf population.

[0094] The sample data was used to divide the gray wolf pack into Wolf, Wolf, Wolves and ordinary gray wolves;

[0095] Regarding the above Wolf, wolves and Wolves employ an exploration strategy based on gradient direction correction. Wolf, wolves and The wolf updates itself and obtains the updated version. Wolf, wolves and Wolf;

[0096] For ordinary gray wolves, a hierarchical weighted encirclement strategy is used to control the ordinary gray wolves to track prey, and the ordinary gray wolves after tracking the prey are obtained.

[0097] After tracking its prey, the ordinary gray wolf, and the updated... Wolf, wolves and The wolves reunited as a gray wolf pack;

[0098] For any individual gray wolf in the gray wolf pack, a probability distribution exploration strategy is used to perform a global greedy update on the individual gray wolf, resulting in the gray wolf individual after the global greedy update.

[0099] Determine the training progress; the training progress is defined as training completed or training not completed; for example, if the number of training sessions is equal to or greater than the preset maximum number of training sessions, the training is determined to be completed, otherwise the training is determined to be incomplete.

[0100] If the training progress is incomplete, based on the gray wolf individuals after the global greedy update, return to the step of dividing the gray wolf group and proceed to the next training.

[0101] When the training progress is complete, the globally optimal parameters of the production equipment fault warning model are determined using the gray wolf individuals after global greedy updates, thus obtaining the pre-trained production equipment fault warning model. For example, the globally optimal gray wolf with the minimum loss function value can be obtained using the globally optimal gray wolf individuals after global greedy updates, and the elements in the globally optimal gray wolf can be used as the globally optimal parameters of the production equipment fault warning model.

[0102] In some possible embodiments, the sample data is used to divide the gray wolf pack into Wolf, Wolf, Wolves and common gray wolves, including:

[0103] The sample production equipment operation data is used as the input data for the production equipment fault early warning model to obtain the predicted output data of the production equipment fault early warning model;

[0104] The fault types corresponding to the sample production equipment operation data are constructed into expected output data, and the loss function value is obtained based on the expected output data and the predicted output data.

[0105] The gray wolf individuals with the smallest, second smallest, and third smallest loss function values ​​are respectively considered as... Wolf, wolves and The remaining gray wolves were treated as ordinary gray wolves.

[0106] In some possible embodiments, an exploration strategy based on gradient orientation correction is employed. Wolf, wolves and The wolf updates itself and obtains the updated version. Wolf, wolves and Wolves, including:

[0107] Generate random gray wolf individuals with each dimension parameter randomly set to 0.1 or -0.1. ;

[0108] Based on the random gray wolf individuals The gradient direction information is obtained as follows:

[0109] ;

[0110] in, express Wolf, wolves or Wolf, Indicates gradient direction information. express The fitness of the solution space. express The fitness of the solution space is 1 / (loss function value + 0.0001).

[0111] The number of training iterations has been determined. The gradient update control coefficients are obtained based on the number of training iterations:

[0112] ;

[0113] in, This represents the gradient update control coefficient. This represents the base coefficient, and is set to 0.01. This represents the change in the coefficient, and is set to 0.005. Represents pi (π). This indicates the maximum number of training iterations preset.

[0114] Based on the gradient direction information and the gradient update control coefficients, obtain the updated... Wolf, wolves and Wolves are:

[0115] ;

[0116] in, Indicates the updated version Wolf, wolves or Wolf, Represents a symbolic function. This represents a single gray wolf whose elements are all 1.

[0117] In the traditional gray wolf optimization algorithm, the optimal individual is updated only through random perturbation, resulting in a slow convergence speed. This invention employs an exploration strategy based on gradient direction correction to update the α, β, and δ wolves in the gray wolf optimization algorithm. By introducing gradient direction information, the gradient of the fitness function is approximated, enabling superior individuals to move along the direction of fitness improvement, thus ensuring the renewal of the gray wolf population.

[0118] In some possible embodiments, a hierarchical weighted encirclement strategy is used to control the ordinary gray wolf to track prey, resulting in the ordinary gray wolf after tracking the prey, including:

[0119] according to Wolf, wolves and The wolf obtains the first, second, and third guiding information as follows:

[0120] ;

[0121] ;

[0122] ;

[0123] in, This represents the m-th ordinary gray wolf. express The corresponding first guidance information, express The corresponding second guidance information, This indicates the third guiding information. express Wolf, express Wolf, express Wolf, This represents the linear dynamic coefficient, which decreases linearly from 2 to 0 as the training process progresses; Represents the first random number between (0,1). This represents the second random number between (0,1). This represents a third random number between (0,1). This represents the fourth random number between (0, 1);

[0124] A hierarchical weighting system is used to enable ordinary gray wolves to track prey based on the first, second, and third guidance information. The ordinary gray wolf after tracking the prey is obtained as follows:

[0125] ;

[0126] in, This indicates the m-th ordinary gray wolf after tracking its prey. This represents the first-level weight, and is set to the range (0, 0.5). This represents the second-level weight, and is set to the range (0, 0.3). This represents the third-level weight, and is set to the range (0, 0.2).

[0127] In traditional algorithms, the guidance weights for the three wolves (α, β, and δ) on ordinary gray wolves are the same, ignoring the quality differences between individuals of different levels. This invention assigns different guidance weights to the α, β, and δ individuals based on their quality, with the α wolf having the highest weight, followed by the β wolf, and the δ wolf the lowest. At the same time, it retains some information about the ordinary gray wolves themselves. This hierarchical design ensures the population's tendency to cluster towards the optimal solution region while avoiding blind convergence caused by over-reliance on a single guidance source, thus effectively balancing local and global exploration.

[0128] In some possible embodiments, a probability distribution exploration strategy is used to perform a global greedy update on individual gray wolves, resulting in gray wolf individuals after the global greedy update, including:

[0129] During the first training process, the mean vector and standard deviation vector are initialized; each element in the mean vector is the mean of all gray wolf individuals in the same dimension; each element in the standard deviation vector is the standard deviation of all gray wolf individuals in the same dimension.

[0130] In the second and subsequent training processes, the updated mean vector and updated standard deviation vector from the previous training process are used to obtain the perturbation individuals corresponding to the individual gray wolf:

[0131] ;

[0132] in, Let represent the d-th element of the k-th gray wolf individual, where k = 1, 2, ..., K, K represents the total number of gray wolves, and d = 1, 2, ..., D, D represents the total number of elements among the gray wolves. This represents the d-th dimension element of the perturbed individual. Represents the Gaussian error function. This represents the d-th element of the standard deviation vector after the last training iteration. This represents the d-th element of the mean vector after the last training iteration;

[0133] Among the gray wolf individuals and their corresponding perturbation individuals, the individuals with smaller loss function values ​​are designated as gray wolf individuals after global greedy updates, while the individuals with larger loss function values ​​are designated as inferior individuals.

[0134] The inferior individuals are used for updating the mean vector and standard deviation vector, and the update rule is as follows:

[0135] ;

[0136] ;

[0137] in, Let represent the d-th element of the updated mean vector. This represents the d-th element of the updated standard deviation vector. This represents the d-th element of the mean vector before the update. This represents the d-th element of the standard deviation vector before the update. Let represent the d-th element of the individual gray wolf after the global greedy update. The d-th element represents a substandard individual.

[0138] The probability distribution exploration strategy introduced in this invention describes the distribution characteristics of a gray wolf pack by constructing two statistical measures: mean and standard deviation, and generates perturbation individuals based on these. Unlike the completely random perturbation method in traditional algorithms, the perturbation individuals generated by this strategy retain the statistical information of the pack distribution, making the exploration process more efficient. Simultaneously, a greedy selection mechanism retains superior individuals while continuously updating the distribution parameters using inferior individuals, forming an adaptive adjustment mechanism for the search range and avoiding the problem of getting trapped in local optima in existing technologies.

[0139] This invention uses GEGWO to enable the production equipment fault early warning model to learn from the sample data, which can effectively avoid the problem of easily getting trapped in local optima in the prior art, while ensuring the training speed and finding the optimal hyperparameter combination of the production equipment fault early warning model in the solution space.

[0140] like Figure 2 As shown, this embodiment of the invention also provides a production equipment operation status monitoring and early warning device based on a digital twin model, comprising:

[0141] The digital twin model construction module 201 is used to construct a digital twin model corresponding to the production equipment and establish a mapping relationship between the digital twin model and the production equipment so that the data of the digital twin model and the production equipment are synchronized in real time.

[0142] The sample data acquisition module 202 is used to acquire sample data of the production equipment under various faults through the digital twin model; the sample data refers to the production equipment operation data and its corresponding fault types at multiple consecutive sampling time points.

[0143] The fault warning model training module 203 is used to construct a production equipment fault warning model using a deep learning model, and to use GEGWO to enable the production equipment fault warning model to learn the sample data, thereby obtaining a pre-trained production equipment fault warning model.

[0144] The real-time online monitoring module 203 is used to run the digital twin model and analyze the real-time production equipment operation data collected by the digital twin model through the pre-trained production equipment fault early warning model to determine the fault monitoring results.

[0145] The monitoring result visualization module 204 is used to visualize and warn the corresponding production equipment in the digital twin model when the fault monitoring result is an abnormal category.

[0146] The production equipment operation status monitoring and early warning device based on a digital twin model provided in this invention can execute the method described in any of the above embodiments. Its principle and beneficial effects are similar and will not be repeated here.

[0147] This invention also provides a computer-readable storage medium storing a computer program / instructions thereon, which, when executed by a processor, implements the steps in the method disclosed in this invention.

[0148] This invention also provides a computer program product that, when run on an electronic device, causes a processor to execute the steps in the method disclosed in this invention.

[0149] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0150] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, electronic devices, and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0151] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0152] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0153] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0154] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0155] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for monitoring and early warning of the operating status of production equipment based on a digital twin model, characterized in that, include: Construct a digital twin model corresponding to the production equipment, and establish a mapping relationship between the digital twin model and the production equipment so that the data of the digital twin model and the production equipment are synchronized in real time; The digital twin model is used to collect sample data of the production equipment under various faults; the sample data refers to the sample production equipment operation data and its corresponding fault types at multiple consecutive sampling time points; A production equipment fault early warning model is constructed using a deep learning model, and GEGWO is used to enable the production equipment fault early warning model to learn from the sample data to obtain a pre-trained production equipment fault early warning model. Run the digital twin model, and analyze the real-time production equipment operation data collected by the digital twin model through the pre-trained production equipment fault early warning model to determine the fault monitoring results; If the fault monitoring result is an abnormal category, the corresponding production equipment will be visualized and given an early warning in the digital twin model.

2. The method for monitoring and early warning of production equipment operation status based on a digital twin model according to claim 1, characterized in that, The digital twin model is used to collect sample data of the production equipment under various fault conditions, including: The production equipment operation data is collected through the digital twin model; the production equipment operation data includes vibration, temperature, current, and voltage. For any given fault, the fault type is collected. At the same time, based on a preset data sampling frequency, the operating data of the production equipment at N sampling time points before the fault occurs are collected to obtain the sample production equipment operating data and its corresponding fault type; N is the number of preset data sampling points.

3. The method for monitoring and early warning of production equipment operation status based on a digital twin model according to claim 1, characterized in that, A production equipment fault early warning model is constructed using a deep learning model, including: constructing a production equipment fault early warning model using a CNN-LSTM model.

4. The method for monitoring and early warning of production equipment operation status based on a digital twin model according to claim 1, characterized in that, The GEGWO algorithm is used to train the production equipment fault early warning model on the sample data, resulting in a pre-trained production equipment fault early warning model, including: Initialize the gray wolf pack; The sample data was used to divide the gray wolf pack into Wolf, Wolf, Wolves and ordinary gray wolves; Regarding the above Wolf, wolves and Wolves employ an exploration strategy based on gradient direction correction. Wolf, wolves and The wolf updates itself and obtains the updated version. Wolf, wolves and Wolf; For ordinary gray wolves, a hierarchical weighted encirclement strategy is used to control the ordinary gray wolves to track prey, and the ordinary gray wolves after tracking the prey are obtained. After tracking its prey, the ordinary gray wolf, and the updated... Wolf, wolves and The wolves reunited as a gray wolf pack; For any individual gray wolf in the gray wolf pack, a probability distribution exploration strategy is used to perform a global greedy update on the individual gray wolf, resulting in the gray wolf individual after the global greedy update. Determine the training progress; the training progress refers to whether the training is completed or not. If the training progress is incomplete, based on the gray wolf individuals after the global greedy update, return to the step of dividing the gray wolf group and proceed to the next training. When the training progress is complete, the global optimal parameters of the production equipment fault early warning model are determined by the gray wolf individuals after global greedy update, and the pre-trained production equipment fault early warning model is obtained.

5. The method for monitoring and early warning of production equipment operation status based on a digital twin model according to claim 4, characterized in that, The sample data was used to divide the gray wolf pack into Wolf, Wolf, Wolves and common gray wolves, including: The sample production equipment operation data is used as the input data for the production equipment fault early warning model to obtain the predicted output data of the production equipment fault early warning model; The fault types corresponding to the sample production equipment operation data are constructed into expected output data, and the loss function value is obtained based on the expected output data and the predicted output data. The gray wolf individuals with the smallest, second smallest, and third smallest loss function values ​​are respectively considered as... Wolf, wolves and The remaining gray wolves were treated as ordinary gray wolves.

6. The method for monitoring and early warning of production equipment operation status based on a digital twin model according to claim 4, characterized in that, An exploration strategy based on gradient orientation correction is adopted for... Wolf, wolves and The wolf updates itself and obtains the updated version. Wolf, wolves and Wolves, including: Generate random gray wolf individuals with each dimension parameter randomly set to 0.1 or -0.

1. ; Based on the random gray wolf individuals The gradient direction information is obtained as follows: ； in, express Wolf, wolves or Wolf, Indicates gradient direction information. express The fitness of the solution space. express The fitness of the solution space is 1 / (loss function value + 0.0001). The number of training iterations has been determined. The gradient update control coefficients are obtained based on the number of training iterations: ； in, This represents the gradient update control coefficient. This represents the base coefficient, and is set to 0.

01. This represents the change in the coefficient, and is set to 0.

005. Represents pi (π). This indicates the maximum number of training iterations preset. Based on the gradient direction information and the gradient update control coefficients, obtain the updated... Wolf, wolves and Wolves are: ； in, Indicates the updated version Wolf, wolves or Wolf, Represents a symbolic function. This represents a single gray wolf whose elements are all 1.

7. The method for monitoring and early warning of production equipment operation status based on a digital twin model according to claim 6, characterized in that, A hierarchical weighted encirclement strategy is used to control ordinary gray wolves tracking prey. The ordinary gray wolves that have tracked the prey include: according to Wolf, wolves and The wolf obtains the first, second, and third guiding information as follows: ； ； ； in, This represents the m-th ordinary gray wolf. express The corresponding first guidance information, express The corresponding second guidance information, This indicates the third guiding information. express Wolf, express Wolf, express Wolf, This represents the linear dynamic coefficient, which decreases linearly from 2 to 0 as the training process progresses; Represents the first random number between (0,1). This represents the second random number between (0,1). This represents a third random number between (0,1). This represents the fourth random number between (0, 1); A hierarchical weighting system is used to enable ordinary gray wolves to track prey based on the first, second, and third guidance information. The ordinary gray wolf after tracking the prey is obtained as follows: ； in, This indicates the m-th ordinary gray wolf after tracking its prey. This represents the first-level weight, and is set to the range (0, 0.5). This represents the second-level weight, and is set to the range (0, 0.3). This represents the third-level weight, and is set to the range (0, 0.2).

8. The method for monitoring and early warning of production equipment operation status based on a digital twin model according to claim 7, characterized in that, A probability distribution exploration strategy is used to perform a global greedy update on the individual gray wolves, resulting in the following gray wolf individuals after the global greedy update: During the first training process, the mean vector and standard deviation vector are initialized; each element in the mean vector is the mean of all gray wolf individuals in the same dimension; each element in the standard deviation vector is the standard deviation of all gray wolf individuals in the same dimension. In the second and subsequent training processes, the updated mean vector and updated standard deviation vector from the previous training process are used to obtain the perturbation individuals corresponding to the individual gray wolf: ； in, Let represent the d-th element of the k-th gray wolf individual, where k = 1, 2, ..., K, K represents the total number of gray wolves, and d = 1, 2, ..., D, D represents the total number of elements among the gray wolves. This represents the d-th dimension element of the perturbed individual. Represents the Gaussian error function. This represents the d-th element of the standard deviation vector after the last training iteration. This represents the d-th element of the mean vector after the last training iteration; Among the gray wolf individuals and their corresponding perturbation individuals, the individuals with smaller loss function values ​​are designated as gray wolf individuals after global greedy updates, while the individuals with larger loss function values ​​are designated as inferior individuals. The inferior individuals are used for updating the mean vector and standard deviation vector, and the update rule is as follows: ； ； in, Let represent the d-th element of the updated mean vector. This represents the d-th element of the updated standard deviation vector. This represents the d-th element of the mean vector before the update. This represents the d-th element of the standard deviation vector before the update. Let represent the d-th element of the individual gray wolf after the global greedy update. The d-th element represents a substandard individual.

9. The method for monitoring and early warning of the operating status of production equipment based on a digital twin model according to claim 1, characterized in that, If the fault monitoring result is abnormal, the corresponding production equipment will be visualized and given an early warning in the digital twin model, including: If the fault monitoring result is abnormal, the production equipment model will be rendered with different colors, and a flashing or highlighting effect will be superimposed based on the fault monitoring result in the digital twin model.

10. A production equipment operation status monitoring and early warning device based on a digital twin model, characterized in that, include: The digital twin model construction module is used to construct a digital twin model corresponding to the production equipment and establish a mapping relationship between the digital twin model and the production equipment so that the data of the digital twin model and the production equipment are synchronized in real time. The sample data acquisition module is used to collect sample data of the production equipment under various faults through the digital twin model; the sample data refers to the production equipment operation data and its corresponding fault types at multiple consecutive sampling time points; The fault warning model training module is used to construct a production equipment fault warning model using a deep learning model, and to use GEGWO to enable the production equipment fault warning model to learn the sample data, thereby obtaining a pre-trained production equipment fault warning model. The real-time online monitoring module is used to run the digital twin model and analyze the real-time production equipment operation data collected by the digital twin model through the pre-trained production equipment fault early warning model to determine the fault monitoring results. The monitoring result visualization module is used to visualize and warn the corresponding production equipment in the digital twin model when the fault monitoring result is an abnormal category.

Images