Equipment description file generation method and device based on AI model
By using an AI model based on sequence-to-sequence models, device description files conforming to the GSDML standard are automatically generated, solving the inefficiency and error problems caused by relying on manual generation in existing technologies, and achieving efficient and accurate file generation.
Patent Information
- Application Number
- CN202511798644.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-02
- Publication Date
- 2026-01-13
AI Technical Summary
In existing technologies, the generation of device description files relies on the professional skills of engineers, which is inefficient and prone to errors.
Employing an AI model based on sequence-to-sequence models, integrating an encoder, attention mechanism, and decoder, and through progressive training and a target loss function, it automatically generates device description files that conform to the GSDML standard.
It enables intelligent and automatic generation of GSDML files, reducing errors, improving generation efficiency, and ensuring the accuracy and consistency of file format and content.
Smart Images
Figure CN121326286A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial automation technology, and in particular to a method and apparatus for generating equipment description files based on AI models. Background Technology
[0002] With the increasing level of industrial automation, Profinet has become the mainstream industrial Ethernet protocol globally, widely used in various industrial control systems. Device description files (GSDML (General Station Description Markup Language)) are the key carrier for interaction between Profinet devices and engineering tools (such as configuration software and programming software), responsible for conveying information such as device capabilities, parameters, and diagnostics. They are the foundation for achieving automatic device configuration, parameterization, and communication. Currently, when generating device description files, engineers manually write XML files directly according to the GSDML specification. This method relies heavily on the engineer's expertise, is inefficient, and prone to errors. Summary of the Invention
[0003] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a device description file generation method and apparatus based on an AI model, which solves the technical problems of the prior art in generating device description files that rely on the professional skills of engineers, are inefficient, and are prone to errors.
[0004] To achieve the above objectives, the main technical solutions adopted by the present invention include: The first aspect of this invention provides a method for generating device description files based on an AI model.
[0005] The device description file generation method based on an AI model proposed in this embodiment of the invention includes: Based on the sequence-to-sequence model, an AI model integrating an encoder, attention mechanism, and decoder is constructed; wherein, the AI model is used to generate standard files conforming to the GSDML universal site description markup language; The AI model is trained using files conforming to the GSDML standard as training samples, enabling the AI model to learn the structural hierarchy, semantic mapping relationships, and format compliance requirements of GSDML files, thereby obtaining the trained target AI model. The structured data of the device to be described is input into the target AI model, and the corresponding GSDML file is automatically generated.
[0006] In some instances, training the AI model using files conforming to the GSDML standard as training samples includes: The AI model is trained using a progressive training strategy and a target loss function, with files conforming to the GSDML standard as training samples. The progressive training strategy includes, in order of increasing complexity, a basic training stage, an intermediate training stage, an advanced training stage, and an expert training stage. The target loss function is a weighted sum of the cross-entropy loss of label prediction and the structured penalty loss; In some instances, the AI model integrates an encoder, an attention mechanism, and a decoder, including: A bidirectional long short-term memory network is used as an encoder to encode the input device parameters; wherein, the device parameters include numerical parameters and text parameters; An attention mechanism is employed to dynamically focus on key parameter information output by the encoder during the decoding process; A long short-term memory network integrating a pointer generation network is used as the decoder; wherein, the pointer generation network is used to copy fixed parameter values in the input sequence; A CRF (Conditional Random Field) layer is connected to the decoder output to constrain the transition rules of the output tags and ensure that the generated XML structure is valid.
[0007] In some instances, the training of the AI model based on a progressive training strategy and a target loss function, using files conforming to the GSDML standard as training samples, includes: introducing adversarial training during the training process, by injecting a preset proportion of noisy data into the training samples to improve the robustness of the AI model to abnormal inputs.
[0008] In some instances, the basic training phase is defined as follows: Using single-module device files as training samples, the device identifier generation function of the AI model is trained, so that the AI model has the ability to generate the basic structure of GSDML files. The intermediate training phase is defined as follows: Using device files with no more than a first predetermined number of sub-modules as training samples, the module structure generation function of the AI model is trained, so that the AI model has the ability to generate GSDML file module structures. The advanced training phase is defined as follows: Using device samples that support the PROFIsafe secure communication protocol as training samples, the AI model is trained to generate security parameters, enabling the AI model to generate GSDML files that conform to the IEC 61784-3-3 standard. The expert training phase is defined as follows: The AI model is trained using a second predetermined number or more sub-modules and complex device samples with a tag sequence length exceeding 5000 as training samples.
[0009] In some instances, the automatic generation of the corresponding GSDML file includes: The target AI model parses the input structured device data and automatically maps the structured device data to the corresponding elements in the GSDML file.
[0010] In some instances, the method includes: The generated GSDML files are verified for compliance using compliance checking tools. The checks include whether the file format conforms to the GSDML V2.44 specification and whether the file content conforms to relevant industry standards. The target AI model is iteratively optimized based on the verification results; wherein, the iterative optimization of the target AI model includes: When compliance verification finds errors or non-compliance, the verification results will be used as feedback information. Based on the feedback information, the model is iteratively optimized by adjusting the parameters of the AI model or updating the model's training data to improve the accuracy and consistency of the subsequently generated GSDML files.
[0011] A second aspect of this invention provides a device description file generation apparatus based on an AI model, characterized in that it includes: The model building unit is used to build an AI model integrating an encoder, attention mechanism, and decoder based on a sequence-to-sequence model; wherein, the AI model is used to generate a standard file conforming to the GSDML general site description markup language; The model training unit is used to train the AI model based on files conforming to the GSDML standard as training samples, so that the AI model learns the structural hierarchy, semantic mapping relationship and format compliance requirements of GSDML files, and obtains the trained target AI model. The file generation unit is used to input the structured data of the device to be described into the target AI model and automatically generate the corresponding GSDML file.
[0012] A third aspect of the present invention provides a computer-readable storage medium, characterized in that it stores an AI-based device description file generation program thereon, which, when executed by a processor, implements the AI-based device description file generation method described in the first aspect above.
[0013] A fourth aspect of the present invention provides an electronic device, characterized in that it includes a memory, a processor, and an AI model-based device description file generation program stored in the memory and executable on the processor. When the processor executes the AI model-based device description file generation program, it implements the AI model-based device description file generation method described in the first aspect above.
[0014] This invention discloses a device description file generation method based on an AI model, comprising: constructing an AI model integrating an encoder, attention mechanism, and decoder based on a sequence-to-sequence model; wherein the AI model is used to generate a standard file conforming to the GSDML universal site description markup language; training the AI model using the GSDML-compliant file as training samples, enabling the AI model to learn the structural hierarchy, semantic mapping relationships, and format compliance requirements of the GSDML file, thereby obtaining a trained target AI model; and inputting the structured data of the device to be described into the target AI model to automatically generate the corresponding GSDML file. This application uses a GSDML-compliant file as a training sample to train the AI model, enabling the AI model to learn the structural hierarchy, semantic mapping relationships, and format compliance requirements of the GSDML file, thereby obtaining a trained target AI model; and inputting the structured data of the device to be described into the target AI model to automatically generate the corresponding GSDML file, thus achieving intelligent and automatic generation of GSDML files, reducing errors while improving efficiency. Attached Figure Description
[0015] Figure 1 A flowchart of a device description file generation method based on an AI model is provided in an embodiment of the present invention; Figure 2 A flowchart for generating a device description file based on an AI model is provided in this embodiment of the invention; Figure 3 This is a schematic diagram of a device description file generation device based on an AI model, provided in an embodiment of the present invention. Detailed Implementation
[0016] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
[0017] The device description file generation method based on an AI model proposed in this invention addresses the problems of low efficiency and susceptibility to errors caused by reliance on engineers' professional skills in device description file generation. It trains the AI model using files conforming to the GSDML standard as training samples, enabling the AI model to learn the structural hierarchy, semantic mapping relationships, and format compliance requirements of GSDML files, resulting in a trained target AI model. The structured data of the device to be described is then input into the target AI model, automatically generating the corresponding GSDML file. This achieves intelligent and automatic GSDML file generation, reducing errors while improving efficiency.
[0018] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be understood more clearly and thoroughly, and that the scope of the present invention can be fully conveyed to those skilled in the art.
[0019] Figure 1 A flowchart illustrating a method for generating device description files based on an AI model, as provided in existing technologies. Figure 1 As shown in the embodiment of the present invention, the method for generating device description files based on AI models includes: Step 100: Based on the sequence-to-sequence model, construct an AI model that integrates an encoder, attention mechanism, and decoder; wherein, the AI model is used to generate standard files that conform to the GSDML general site description markup language; Step 110: Using files conforming to the GSDML standard as training samples, train the AI model so that the AI model learns the structural hierarchy, semantic mapping relationship and format compliance requirements of GSDML files, and obtains the trained target AI model. Step 120: Input the structured data of the device to be described into the target AI model, and automatically generate the corresponding GSDML file.
[0020] In this exemplary embodiment, the device description file (GSDML) refers to the device description markup language conforming to the ISO 15745-4 standard. It is used to define the network interfaces, modules, sub-modules, IO data, parameters, diagnostics and other information of the Profinet device. It is a key file for engineering tools to identify, configure and parameterize the Profinet device.
[0021] Sequence-to-sequence models are deep learning models that transform input sequences into output sequences, and are suitable for tasks involving the automatic generation of structured data.
[0022] In this application, the GSDML file may be GSDML V2.44. GSDML V2.44 refers to the latest version of the GSDML specification conforming to the ISO 15745-4 standard, which specifies the format, structure, and element requirements of the device description file.
[0023] In this exemplary embodiment, before model training, various technical information of the target Profinet device is first collected, including but not limited to: device model, vendor ID, device model ID, device name, information text, network interface (DAP) configuration, module / submodule list, IO data structure (input / output data type, bit decomposition, channel mapping), parameter records (writable parameters, PROFIsafe parameters), graphic resources, multilingual text, authentication level (such as ConformanceClass, ApplicationClass, SecurityClass), extended functions (such as TSN support, PROFIenergy, PROFIsafe, etc.), and other raw data.
[0024] Then, the collected raw data is cleaned to remove duplicate and invalid information, and the data format is standardized (such as unifying data types, label naming, units, etc.) to form a structured dataset, which is convenient for subsequent AI model training and use.
[0025] In this application, a GSDML-compliant file is used as a training sample to train the AI model, enabling the AI model to learn the structural hierarchy, semantic mapping relationship, and format compliance requirements of the GSDML file, thus obtaining the trained target AI model. The structured data of the device to be described is input into the target AI model, and the corresponding GSDML file is automatically generated, thereby achieving intelligent and automatic generation of GSDML files, reducing errors and improving efficiency.
[0026] In some instances, training the AI model using files conforming to the GSDML standard as training samples includes: The AI model is trained using a progressive training strategy and a target loss function, with files conforming to the GSDML standard as training samples. The progressive training strategy includes, in order of increasing complexity, a basic training stage, an intermediate training stage, an advanced training stage, and an expert training stage. The target loss function is a weighted sum of the cross-entropy loss of label prediction and the structured penalty loss.
[0027] In this exemplary embodiment, the target loss function is ℒ Total ； Among them, ℒ Total = ℒCE +λ*ℒ Struct； ℒ CE Predict cross-entropy loss for labels; ℒ CE = ;in, The probability that the model predicts the correct label at position t; This represents the total length of the XML sequence.
[0028] The role of the label prediction cross-entropy loss is to drive the model to accurately predict the XML label at each position (e.g., : <module> 、 <parmeter>wait).
[0029] in, <module>It defines the device's "hardware structure" and pluggable units.
[0030] <parmeter>Define the device's "configurable parameters".
[0031] ℒ Struct For structured penalty loss; ℒ Struct =α*CountError+β*PathError; among them, CountError indicates an error in the number of tags. PathError is the number of illegal nested paths; Where α=1.0, β=2.0.
[0032] λ is the penalty weight coefficient, where λ = 0.8 * (1 + 0.02 * epoch); The initial value λ = 0.8, and the number of epochs increases by 1.6% after each training round; The purpose of the structured penalty loss is to make the model pay more attention to structural compliance in the later stages of training (e.g., λ=0.8*(1+0.02*50)=1.6 in the 50th round).
[0033] In some instances, the AI model integrates an encoder, an attention mechanism, and a decoder, including: A bidirectional long short-term memory network is used as an encoder to encode the input device parameters, which include numerical parameters and text parameters. An attention mechanism is employed to dynamically focus on key parameter information output by the encoder during the decoding process; An LSTM with an integrated pointer generation network is used as the decoder, which is used to copy fixed parameter values in the input sequence; A Conditional Random Field (CRF) layer is connected to the decoder output to constrain the transition rules of the output tags and ensure that the generated XML structure is valid.
[0034] In this exemplary embodiment, a suitable deep learning model for the automatic generation of structured data, such as a sequence-to-sequence (Seq2Seq) model, is selected to construct an AI model for GSDML file generation.
[0035] A GSDML file generation model is constructed using a sequence-to-sequence (Seq2Seq) model. The specific construction process is as follows: Encoder design: 1. A bidirectional LSTM (Long Short-Term Memory) network structure is adopted, with the input dimension being the device parameter feature dimension; 2. Numerical encoding of numerical parameters (such as VendorID, number of modules); 3. Vectorize text-type parameters (such as InfoText) using word embedding techniques; 4. The output is a context vector that incorporates bidirectional semantics.
[0036] Attention mechanism: The Bahdanau attention mechanism is used to dynamically focus on key parameters; key parameters may include device address, watchdog time, model-specific parameters, etc.
[0037] The decoder automatically associates device parameters when generating a specific tag.
[0038] Decoder Design: A decoder is constructed by combining an LSTM structure with a pointer generation network. Pointer networks directly copy fixed parameter values (such as VendorID); The output layer connects to a CRF (Conditional Random Field) to ensure the validity of the XML structure.
[0039] Specifically, when the AI model generates GSDML files and encounters places where it needs to fill in device-specific parameter values (such as VendorID=0x002A, DeviceID=0x1234, security codes, etc.), the pointer generation network allows the model to directly "copy and paste" this original value from the input data. This avoids the model having to guess and generate these strictly accurate, error-free fixed values or specific identifiers.
[0040] In some instances, the training of the AI model based on a progressive training strategy and a target loss function, using files conforming to the GSDML standard as training samples, includes: introducing adversarial training during the training process, by injecting a preset proportion of noisy data into the training samples to improve the robustness of the AI model to abnormal inputs.
[0041] In this exemplary embodiment, 5% noise data is injected (e.g., required fields are removed) to improve the model's robustness to abnormal inputs.
[0042] In some instances, the basic training phase is defined as follows: Using single-module device files as training samples, the device identifier generation function of the AI model is trained, so that the AI model has the ability to generate the basic structure of GSDML files. The intermediate training phase is defined as follows: Using device files with no more than a first predetermined number of sub-modules as training samples, the module structure generation function of the AI model is trained, so that the AI model has the ability to generate GSDML file module structures. The advanced training phase is defined as follows: Using device samples that support the PROFIsafe secure communication protocol as training samples, the AI model is trained to generate security parameters, enabling the AI model to generate GSDML files that conform to the IEC 61784-3-3 standard. The expert training phase is defined as follows: The AI model is trained using a second predetermined number or more sub-modules and complex device samples with a tag sequence length exceeding 5000 as training samples.
[0043] In this exemplary embodiment, the first predetermined quantity can be between 5 and 10, and the specific quantity can be set as needed; it is not a limitation here. The second predetermined quantity can be between 50 and 60, and the specific quantity can be set as needed; it is not a limitation here.
[0044] In some instances, the automatic generation of the corresponding GSDML file includes: The target AI model parses the input structured device data and automatically maps the structured device data to the corresponding elements in the GSDML file.
[0045] In this exemplary embodiment, Map device identity information to <deviceidentity>Under the element <vendorid> 、 <deviceid>Field; Map module information to <applicationprocess>Below <modulelist> 、 <submodulelist>element; Map IO data and parameter records to <iodata> 、 <recorddatalist>element; Map authentication information and language support to the corresponding authentication mechanisms and language support elements.
[0046] In some instances, the method includes: The generated GSDML files are verified for compliance using compliance checking tools. The checks include whether the file format conforms to the GSDML V2.44 specification and whether the file content conforms to relevant industry standards. The target AI model is iteratively optimized based on the verification results; wherein, the iterative optimization of the target AI model includes: When compliance verification finds errors or non-compliance, the verification results will be used as feedback information. Based on the feedback information, the model is iteratively optimized by adjusting the parameters of the AI model or updating the model's training data to improve the accuracy and consistency of the subsequently generated GSDML files.
[0047] In this exemplary embodiment, the generated results are quantitatively evaluated based on predefined evaluation metrics, which include at least XML format accuracy, key parameter accuracy, number of nested thousand tags, and inference speed.
[0048] In this exemplary embodiment, the generated GSDML file is verified by a compliance check tool (such as GSDChecker) to check whether it conforms to the format specifications of GSDML V2.44 (such as label integrity, attribute validity, reasonableness of nesting relationships, etc.) and industry standards (such as the accuracy of device capabilities, compliance of parameter settings, etc.).
[0049] For example, check if the XML format accuracy is greater than 99.95%, the critical parameter accuracy is greater than 0.98, the nesting error count per thousand tags is less than 1, and the inference speed is less than 3 seconds, etc. If all the above indicators are met, the generated GSDML file is determined to conform to the GSDML V2.44 format specification.
[0050] Figure 2 This is a flowchart illustrating a device description file generation process based on an AI model, provided as an embodiment of the present invention. Figure 2 As shown, the device description file generation process based on the AI model includes: Step 20: Begin; Step 21, Data Acquisition; Step 22: Determine if the data is complete and valid; Step 23: If the data is incomplete or invalid, perform data cleaning and standardization. Step 24: If the data is complete and valid, then build the AI model; Step 25: Determine if the model training has met the standards. Step 26: If the model training meets the requirements, a GSDML file will be automatically generated. Step 27: If the model training does not meet the requirements, adjust the model parameters; Step 28, End.
[0051] The device description file generation method based on the AI model of this application automatically generates GSDML files through artificial intelligence models, which greatly shortens the file generation time. Especially for complex devices (devices containing multiple modules and advanced functions), the description file can be generated in a short time, thus improving work efficiency.
[0052] The device description file generation method based on AI model in this application avoids errors caused by manual writing or traditional tool generation by leveraging the learning capabilities of deep learning models, ensuring that the generated GSDML file fully conforms to the GSDML V2.44 format specification, and improving the accuracy of engineering tools in identifying and configuring devices.
[0053] The device description file generation method based on the AI model in this application unifies the GSDML file generation logic of different devices, ensures the uniformity of file format and content, and facilitates subsequent maintenance and management.
[0054] The device description file generation method based on AI models in this application can quickly adapt to new GSDML standards and new functions (such as Time Sensitive Networking (TSN), Functional Safety (PROFIsafe), and Energy Saving Mode (PROFIenergy). It only requires updating the training data or model and does not require a lot of manual adjustments, thus enhancing the continuous development capability of the technology.
[0055] In this exemplary embodiment, the method includes: An AI-based device description file generation method is applied to the lifecycle management scenario of PROFINET devices; the PROFINET device lifecycle management scenario includes configuration and integration phases; The configuration and integration phase includes: When using PLC programming or network configuration software, new PROFINET devices need to be added to the project hardware configuration.
[0056] The GSDML file provides all the necessary information about the device, enabling engineering tools to identify it as a valid PROFINET device.
[0057] The project hardware configuration includes correctly displaying the device name, model, order number, and image in the hardware catalog or device library.
[0058] The project hardware configuration includes understanding the modules and sub-modules supported by the device (for example, an I / O device may have sub-modules such as digital inputs, analog outputs, and diagnostics).
[0059] Know which configuration parameters (parameterized data) the device requires, as well as the data type, range, and default value of these parameters.
[0060] The project hardware configuration includes understanding the communication relationships supported by the device (such as input data length, output data length, supported RT / IRT levels, supported alarm types, etc.).
[0061] The project's hardware configuration includes providing a user-friendly configuration interface (based on information generated in GSDML to create a configuration dialog box).
[0062] In summary, GSDML is applicable to any situation where information about PROFINET devices needs to be "told" to engineering tools or PROFINET controllers, covering the entire process of device selection, configuration, integration, startup, operation monitoring, and diagnostics.
[0063] It is the key technological foundation for enabling PROFINET device operation and plug-and-play functionality. Without the correct GSDML file, engineering tools cannot recognize and configure the PROFINET device.
[0064] This invention provides a device description file generation apparatus based on an AI model. Figure 3 This is a schematic diagram of a device description file generation apparatus based on an AI model, provided as an embodiment of the present invention. Figure 3 As shown, the device description file generation apparatus based on an AI model includes: The model building unit 30 is used to build an AI model integrating an encoder, attention mechanism, and decoder based on a sequence-to-sequence model; wherein, the AI model is used to generate a standard file conforming to the GSDML general station description markup language; The model training unit 31 is used to train the AI model based on files conforming to the GSDML standard as training samples, so that the AI model learns the structural hierarchy, semantic mapping relationship and format compliance requirements of GSDML files, and obtains the trained target AI model. The file generation unit 32 is used to input the structured data of the device to be described into the target AI model and automatically generate the corresponding GSDML file.
[0065] In this exemplary embodiment, the device description file (GSDML) refers to the device description markup language conforming to the ISO 15745-4 standard. It is used to define the network interfaces, modules, sub-modules, IO data, parameters, diagnostics and other information of the Profinet device. It is a key file for engineering tools to identify, configure and parameterize the Profinet device.
[0066] Sequence-to-sequence models are deep learning models that transform input sequences into output sequences, and are suitable for tasks involving the automatic generation of structured data.
[0067] In this application, the GSDML file may be GSDML V2.44. GSDML V2.44 refers to the latest version of the GSDML specification conforming to the ISO 15745-4 standard, which specifies the format, structure, and element requirements of the device description file.
[0068] In this exemplary embodiment, before model training, various technical information of the target Profinet device is collected, including but not limited to: device model, vendor ID, device ID, device name, information text, network interface (DAP) configuration, module / submodule list, IO data structure (input / output data type, bit decomposition, channel mapping), parameter records (writable parameters, PROFIsafe parameters), graphic resources, multilingual text, authentication level (such as ConformanceClass, ApplicationClass, SecurityClass), extended functions (such as TSN support, PROFIenergy, PROFIsafe, etc.), and other raw data.
[0069] The collected raw data is cleaned to remove duplicate and invalid information; the data format is standardized (e.g., uniform data types, label naming, units, etc.) to form a structured dataset and to build files that conform to the GSDML standard, so as to facilitate subsequent AI model training and use.
[0070] In this application, a GSDML-compliant file is used as a training sample to train the AI model, enabling the AI model to learn the structural hierarchy, semantic mapping relationship, and format compliance requirements of the GSDML file, thus obtaining the trained target AI model. The structured data of the device to be described is input into the target AI model, and the corresponding GSDML file is automatically generated, thereby achieving intelligent and automatic generation of GSDML files, reducing errors and improving efficiency.
[0071] Since the systems / devices described in the above embodiments of the present invention are systems / devices used to implement the methods of the above embodiments of the present invention, those skilled in the art can understand the specific structure and modifications of the systems / devices based on the methods described in the above embodiments of the present invention, and therefore will not be repeated here. All systems / devices used in the methods of the above embodiments of the present invention fall within the scope of protection of the present invention.
[0072] This invention provides a computer-readable storage medium, characterized in that it stores an AI-based device description file generation program thereon, which, when executed by a processor, implements the AI-based device description file generation method described in the above embodiments.
[0073] This invention provides an electronic device, characterized in that it includes a memory, a processor, and an AI-based device description file generation program stored in the memory and executable on the processor. When the processor executes the AI-based device description file generation program, it implements the AI-based device description file generation method described in the above embodiments.
[0074] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0075] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.
[0076] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first and second features are in direct contact, or that they are in indirect contact through an intermediate medium. Furthermore, "above," "over," and "on top" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "beneath" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.
[0077] In the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0078] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make modifications, alterations, substitutions and variations to the above embodiments within the scope of the present invention.< / recorddatalist> < / iodata> < / submodulelist> < / modulelist> < / applicationprocess> < / deviceid> < / vendorid> < / deviceidentity> < / parmeter> < / module> < / parmeter> < / module>
Claims
1. A method for generating device description files based on AI models, characterized in that, include: Based on the sequence-to-sequence model, an AI model integrating an encoder, attention mechanism, and decoder is constructed; wherein, the AI model is used to generate standard files conforming to the GSDML universal site description markup language; The AI model is trained using files that conform to the GSDML standard as training samples, so that the AI model learns the structural hierarchy, semantic mapping relationship and format compliance requirements of GSDML files, and obtains the trained target AI model. The structured data of the device to be described is input into the target AI model, and the corresponding GSDML file is automatically generated.
2. The method for generating device description files based on AI models according to claim 1, characterized in that, The training of the AI model using files conforming to the GSDML standard as training samples includes: The AI model is trained using a progressive training strategy and a target loss function, with files conforming to the GSDML standard as training samples. The progressive training strategy includes, in order of increasing complexity, a basic training stage, an intermediate training stage, an advanced training stage, and an expert training stage. The target loss function is a weighted sum of the cross-entropy loss of label prediction and the structured penalty loss.
3. The method for generating device description files based on AI models according to claim 1, characterized in that, The AI model integrating the encoder, attention mechanism, and decoder includes: A bidirectional long short-term memory network is used as an encoder to encode the input device parameters; wherein, the device parameters include numerical parameters and text parameters; An attention mechanism is employed to dynamically focus on key parameter information output by the encoder during the decoding process; A long short-term memory network integrating a pointer generation network is used as the decoder; wherein, the pointer generation network is used to copy fixed parameter values in the input sequence; A Conditional Random Field (CRF) layer is connected to the decoder output to constrain the transition rules of the output tags and ensure that the generated XML structure is valid.
4. The method for generating device description files based on AI models according to claim 2, characterized in that, The AI model is trained using a progressive training strategy and target loss function, with files conforming to the GSDML standard as training samples. This includes introducing adversarial training during the training process, by injecting a preset proportion of noisy data into the training samples to improve the robustness of the AI model to abnormal inputs.
5. The method for generating device description files based on AI models according to claim 2, characterized in that, The basic training phase is defined as follows: Using single-module device files as training samples, the device identifier generation function of the AI model is trained, so that the AI model has the ability to generate the basic structure of GSDML files. The intermediate training phase is defined as follows: Using device files with no more than a first predetermined number of sub-modules as training samples, the module structure generation function of the AI model is trained, so that the AI model has the ability to generate GSDML file module structures. The advanced training phase is defined as follows: Using device samples that support the PROFIsafe secure communication protocol as training samples, the AI model is trained to generate security parameters, enabling the AI model to generate GSDML files that conform to the IEC 61784-3-3 standard. The expert training phase is defined as follows: The AI model is trained using a second predetermined number or more sub-modules and complex device samples with a tag sequence length exceeding 5000 as training samples.
6. The method for generating device description files based on AI models according to claim 1, characterized in that, The automatic generation of the corresponding GSDML file includes: The target AI model parses the input structured device data and automatically maps the structured device data to the corresponding elements in the GSDML file.
7. The method for generating device description files based on AI models according to claim 1, characterized in that, The method includes: The generated GSDML files are verified for compliance using compliance checking tools. The checks include whether the file format conforms to the GSDML V2.44 specification and whether the file content conforms to relevant industry standards. The target AI model is iteratively optimized based on the verification results; wherein, the iterative optimization of the target AI model includes: When compliance verification finds errors or non-compliance, the verification results will be used as feedback information. Based on the feedback information, the model is iteratively optimized by adjusting the parameters of the AI model or updating the model's training data to improve the accuracy and consistency of the subsequently generated GSDML files.
8. A device description file generation apparatus based on an AI model, characterized in that, include: The model building unit is used to build an AI model integrating an encoder, attention mechanism, and decoder based on a sequence-to-sequence model; wherein, the AI model is used to generate a standard file conforming to the GSDML general site description markup language; The model training unit is used to train the AI model based on files conforming to the GSDML standard as training samples, so that the AI model learns the structural hierarchy, semantic mapping relationship and format compliance requirements of GSDML files, and obtains the trained target AI model. The file generation unit is used to input the structured data of the device to be described into the target AI model and automatically generate the corresponding GSDML file.
9. A computer-readable storage medium, characterized in that, It stores an AI-based device description file generation program, which, when executed by a processor, implements the AI-based device description file generation method according to any one of claims 1-7.
10. An electronic device, characterized in that, The device description file generation method based on the AI model is included, comprising a memory, a processor, and an AI model-based device description file generation program stored on the memory and executable on the processor. When the processor executes the AI model-based device description file generation program, it implements the AI model-based device description file generation method according to any one of claims 1-7.