Operation and maintenance knowledge base construction and alarm linkage method and system based on AI translation

Abstract

The invention provides an operation and maintenance knowledge base construction and alarm linkage method and system based on AI translation, and relates to the technical field of intelligent operation and maintaining.The method comprises the steps that equipment alarm information conforming to an original language is obtained, an AI translation engine preloaded with an operation and maintenance terminology library is called, semantic understanding and language conversion are conducted on the equipment alarm information, and the equipment alarm information is obtained; target semantics conforming to a target language are generated, and the operation and maintenance terminology library is pre-established and stores equipment operation and maintenance terminology; based on the target semantics, retrieving a matched operation and maintenance processing scheme from an equipment operation and maintenance knowledge base, and performing multi-language association display; obtaining operation feedback information based on the operation and maintenance processing scheme; and updating the equipment operation and maintenance knowledge base and / or the operation and maintenance professional term base based on the operation feedback information. According to the invention, the terminology translation of equipment operation and maintenance is accurate, and the operation and maintenance efficiency after equipment alarm can be improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent equipment operation and maintenance technology, and more specifically, to a method and system for constructing an operation and maintenance knowledge base and linking alarms based on AI translation. Background Technology

[0002] As industrial globalization accelerates, multinational corporations face increasingly prominent challenges in equipment operation and maintenance, requiring multilingual collaboration.

[0003] In related technologies, equipment operation and maintenance solutions can be summarized into three categories: First, traditional information management systems without multilingual capabilities, relying on manual translation and prone to errors; second, general knowledge platforms using rule-based translation, which result in stilted translations, low accuracy of professional terminology, and a gradual decline in translation accuracy as the number of equipment types increases, making it difficult to adapt to the operation and maintenance needs of new equipment; third, single alarm systems, which are inefficient and risky for operation and maintenance personnel. These solutions are insufficient to meet the needs of efficient and accurate equipment operation and maintenance support in globalized industrial scenarios. Summary of the Invention

[0004] The problem this invention aims to solve is at least one of the following issues in the operation and maintenance of industrial equipment: inaccurate translation of technical terms, low efficiency in alarm handling, and inability to adapt to the operation and maintenance needs of new equipment.

[0005] To address the aforementioned problems, in a first aspect, this invention provides a method for constructing an operations and maintenance knowledge base and linking alarms based on AI translation, comprising: The system obtains device alarm information that conforms to the original language, calls an AI translation engine preloaded with an operation and maintenance terminology database, performs semantic understanding and language conversion on the device alarm information, and generates target semantics that conforms to the target language. The operation and maintenance terminology database is pre-established and stores device operation and maintenance terms. Based on the target semantics, matching operation and maintenance solutions are retrieved from the equipment operation and maintenance knowledge base and displayed in multiple languages. The equipment operation and maintenance knowledge base is pre-established and stores operation and maintenance information for the entire life cycle of the equipment. Obtain operation feedback information based on the aforementioned operation and maintenance processing solution; Based on the operation feedback information, update the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology database.

[0006] This invention provides an AI-based method for constructing an operations and maintenance (O&M) knowledge base and linking alarms. By building an AI translation engine that integrates an O&M terminology database and performing semantic understanding and language conversion on device alarm information, it obtains target semantics consistent with the target language. This achieves terminology matching and automatic translation of device alarm information, meeting the accurate translation needs of O&M scenarios involving language conversion (such as Chinese-English conversion). Based on the target semantics, it retrieves matching O&M solutions from the device O&M knowledge base and displays them in multiple languages, establishing a real-time linkage mechanism between translation and alarms. This enables synchronous retrieval and display of alarm information and O&M solutions, shortening the alarm processing cycle in multilingual environments and improving alarm processing efficiency. Furthermore, by updating the device O&M knowledge base and / or O&M terminology database based on operational feedback information, it establishes a feedback-based dynamic database update mechanism. This drives iterative optimization of the translation model, continuously improving the industry adaptability and terminology accuracy of the translation engine, and enabling it to adapt to the O&M needs of new equipment. Finally, it achieves full automation of multilingual retrieval and display based on the knowledge base, lowering the language barrier for O&M personnel and thus improving the overall efficiency of cross-border O&M collaboration.

[0007] Secondly, embodiments of the present invention provide an AI-translated operation and maintenance knowledge base construction and alarm linkage system, applying the AI-translated operation and maintenance knowledge base construction and alarm linkage method as described in any of the above claims, including: The alarm information acquisition and translation module is used to: acquire device alarm information that conforms to the original language, call an AI translation engine with a pre-loaded operation and maintenance terminology library, perform semantic understanding and language conversion on the device alarm information, and generate target semantics that conforms to the target language. The operation and maintenance terminology library is a pre-established library that stores device operation and maintenance terms. The solution matching and display module is used to: retrieve matching operation and maintenance solutions from the equipment operation and maintenance knowledge base based on the target semantics and display them in multiple languages, wherein the equipment operation and maintenance knowledge base is pre-established and stores equipment operation and maintenance information throughout its entire life cycle; The feedback acquisition module is used to: acquire operation feedback information based on the operation and maintenance processing plan; The knowledge update module is used to update the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology base based on the operation feedback information.

[0008] Thirdly, the present invention provides an electronic device, including a memory and a processor; The memory is used to store computer programs; The processor is configured to, when executing the computer program, implement the AI ​​translation-based operation and maintenance knowledge base construction and alarm linkage method as described in the first aspect.

[0009] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the AI ​​translation-based operation and maintenance knowledge base construction and alarm linkage method as described in the first aspect.

[0010] The AI ​​translation-based operation and maintenance knowledge base construction and alarm linkage system, electronic device, and computer-readable storage medium provided by this invention have the same beneficial effects as the AI ​​translation-based operation and maintenance knowledge base construction and alarm linkage method compared to the prior art, and will not be repeated here. Attached Figure Description

[0011] Figure 1 A flowchart illustrating the method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation in an embodiment of the present invention is shown. Figure 2 A flowchart of the alarm linkage processing in an embodiment of the present invention is shown; Figure 3 A schematic diagram illustrating the translation process implemented by the AI ​​translation engine in an embodiment of the present invention is shown; Figure 4 This invention illustrates the overall architecture diagram of the AI ​​translation-based operation and maintenance knowledge base construction and alarm linkage method in an embodiment of the present invention. Figure 5 This diagram illustrates the structure of the AI ​​translation-based operation and maintenance knowledge base construction and alarm linkage system in an embodiment of the present invention. Figure 6 A schematic diagram of the structure of an electronic device according to an embodiment of the present invention is shown. Detailed Implementation

[0012] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0013] It should be noted that relational terms such as "first" and "second" in this invention are used merely 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 apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, as well as elements inherent to such a process, method, article, or apparatus. 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 apparatus that includes said element.

[0014] The terminology used in the embodiments of the present invention will be explained and described below.

[0015] Equipment Domain Knowledge Base: A professional knowledge base focusing on the operation and maintenance information of industrial equipment throughout its entire life cycle, integrating parameter standards, operating procedures, and alarm handling solutions; Alarm linkage: When a device triggers an alarm, the system automatically associates the corresponding solution in the knowledge base and displays it synchronously. AI Translation Engine: A core component that enables automatic translation, proofreading, and adaptation of equipment operation and maintenance professional terminology in Chinese and English based on deep learning models (such as the Transformer architecture), supporting contextual semantic understanding and industry terminology database updates; Operation and Maintenance Terminology Database: A standardized terminology collection built for the industrial equipment field, including Chinese-English translations and definitions of professional terms such as instrument parameters, fault types, and handling procedures; Translation confidence: A reliability index of the AI ​​translation engine output (value 0-100%), calculated based on the probability distribution of the translation results by the model (e.g., if the probability of "sensor wiring" is 0.98 and the probability of "sensor line" is 0.02, then the confidence is 98%), used to determine whether the translation results need to be manually proofread; Transformer architecture: A deep learning model architecture based on "self-attention mechanism" that can simultaneously capture the contextual associations of different words in the text (such as the semantic binding between "overheating" and "alarm"), which is the core foundation of the AI ​​translation engine of this invention, improving the accuracy of professional terminology translation by more than 30% compared with traditional models. BERT model (Bidirectional Encoder Representations from Transformers): A bidirectional language model based on the Transformer architecture, which understands the semantics of text through "bidirectional context learning" (such as the different meanings of "shell" in the scenarios of "device shell" and "program shell"). This invention adopts an industrial customized BERT-Industrial model (fine-tuned with 50,000+ equipment operation and maintenance corpora) to adapt to professional text translation in the equipment field. RNN (Recurrent Neural Network): A deep learning model for processing sequential data. It captures the dependencies between texts through "temporal memory units", but it suffers from the problem of "vanishing gradients in long texts". Its accuracy is lower than that of the Transformer architecture in translating long documents for equipment operation and maintenance (such as operating specifications). LSTM (Long Short-Term Memory): An improved model of RNN, which solves the gradient vanishing problem of long texts through "forget gate, input gate, and output gate". It can handle medium-length text translation, but its ability to capture contextual semantics is weaker than that of Transformer (such as difficulty in recognizing the professional collocation relationship between "calibration" and "offset"). In this invention, it is used as a lightweight alternative to AI translation engines. Equipment lifecycle operation and maintenance: Covers the entire operation and maintenance information management of equipment from factory delivery, installation, commissioning, operation, maintenance to scrapping. The knowledge base stores the corresponding technical documents for each stage (such as installation manuals, calibration specifications, and troubleshooting guides), and realizes multi-language support for all stages.

[0016] Reference Figure 1 As shown in the figure, this embodiment of the invention proposes a method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation; The method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation includes: S100: Obtain device alarm information that conforms to the original language, call an AI translation engine with a pre-loaded maintenance terminology library, perform semantic understanding and language conversion on the device alarm information, and generate target semantics that conforms to the target language. The maintenance terminology library is pre-established and stores device maintenance terms.

[0017] Specifically, by means of monitoring industrial control networks (e.g., subscribing to topics published by the device alarm center via the MQTT protocol), real-time device alarm information described in original languages ​​such as English is obtained. Subsequently, an AI translation engine preloaded with an operation and maintenance terminology library is invoked for processing. This translation engine first performs term recognition and matching on the device alarm information based on the operation and maintenance terminology library (e.g., identifying "overheating alarm" as a domain term), and then combines its contextual understanding ability (e.g., the self-attention mechanism of the Transformer in deep neural networks) to generate target semantic information that conforms to, for example, Chinese expression habits, such as accurately translating "overheating alarm in motor" as "motor overheating alarm".

[0018] S200: Based on the target semantics, retrieve matching operation and maintenance solutions from the equipment operation and maintenance knowledge base and display them in multiple languages. The equipment operation and maintenance knowledge base is pre-established and stores operation and maintenance information for the entire life cycle of the equipment.

[0019] Specifically, based on the generated target semantics (such as "motor over-temperature alarm"), key search terms (such as "motor" and "over-temperature alarm") are extracted and combined in a pre-built equipment operation and maintenance knowledge base. The operation and maintenance solutions entries in the equipment operation and maintenance knowledge base have been stored in multiple languages ​​(such as Chinese and English) according to equipment type and information category. After matching the corresponding Chinese processing solution (such as "1. Check the cooling fan; 2. Measure the motor winding temperature"), the corresponding operation and maintenance processing solution can be displayed in a multilingual view on the user interface (such as the alarm page of the PRIDE knowledge base APP, which is the implementation carrier of this invention and is an application for industrial equipment operation and maintenance personnel). For example, it can display the original alarm in English and the corresponding English solution, and simultaneously display the translated Chinese alarm information and its Chinese solution for operation and maintenance personnel to choose from.

[0020] S300: Obtain operation feedback information based on the operation and maintenance processing scheme.

[0021] Specifically, after completing the on-site handling according to the presented solution, the maintenance personnel can enter the actual results of this operation, i.e., operation feedback information, through the feedback submission portal of the user interface. The operation feedback information may include the fault confirmation status (such as "resolved"), details of the actual measures taken (such as "the cooling fan bearing was found to be stuck, and has been cleaned and lubricated"), and suggestions for correcting the presented solution or translation content (such as pointing out that the more professional translation of "over-temperature alarm" in this equipment scenario should be "overheating alarm").

[0022] S400: Based on the operation feedback information, update the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology base.

[0023] Specifically, the operation feedback information is parsed and the database is updated according to its content type. For example, for information containing new operation and maintenance experience, it is converted into multilingual (such as Chinese and English) data and stored in the equipment operation and maintenance knowledge base; and / or, for correction information containing terminology translation, the operation and maintenance professional terminology database is updated accordingly (such as establishing a standard correspondence between "overheating alarm" and "overheating alarm"). Finally, using these newly added or corrected multilingual data, the AI ​​translation engine loaded with the operation and maintenance professional terminology database will also have its translation capabilities improved, enhancing the professionalism and accuracy of the translation.

[0024] In practical application, this embodiment constructs an AI translation engine loaded with an O&M terminology database and performs semantic understanding and language conversion on device alarm information to obtain target semantics that conforms to the target language. This achieves terminology matching and automatic translation of device alarm information, meeting the accurate translation needs of device alarm information in language conversion O&M scenarios (such as Chinese-English conversion scenarios). By retrieving matching O&M solutions from the device O&M knowledge base based on the target semantics and displaying them in multiple languages, a real-time linkage mechanism between translation and alarms is established. This enables synchronous retrieval and display of alarm information and O&M solutions, shortening the alarm processing cycle in multilingual environments and improving alarm processing efficiency. By updating the device O&M knowledge base and / or O&M terminology database based on operation feedback information, a feedback-based dynamic database update mechanism is established. This drives iterative optimization of the translation model, continuously improving the industry adaptability and terminology accuracy of the translation engine, and enabling it to adapt to the O&M needs of new equipment. Finally, it achieves full automation of multilingual retrieval and display based on the knowledge base, reducing the language barrier for O&M personnel and thus improving the overall efficiency of cross-border O&M collaboration.

[0025] Compared with related technologies, the present invention achieves improved performance, as shown in Table 1 below.

[0026] Table 1

[0027] It should be noted that the original language refers to the default language of the alarm signal (corresponding to the device alarm information) directly obtained from the device. This is usually determined by the factory settings of the device control system. For example, in global industrial scenarios, many device control systems (especially imported equipment from multinational corporations) default to outputting English alarm signals. Of course, the original language can also be Chinese, German, Japanese, or a combination of languages. The target language refers to the language converted for the convenience of a specific group of maintenance personnel. It is usually consistent with the operating language of the maintenance team. For example, for a Chinese-speaking maintenance team, the target language is usually set to Chinese. The core of this invention lies in achieving intelligent conversion between the two languages, the direction of which can be configured according to actual needs. When the default output of the device alarm is Chinese, the target language can also be set to English or other languages ​​to meet the collaboration needs of maintenance personnel with different language backgrounds. In specific embodiments of this invention, for clarity, English is mainly used as the original language and Chinese as the target language for illustrative description. However, it should be understood that this does not constitute any limitation on the scope of protection of this invention.

[0028] like Figure 2As shown, as an optional embodiment of the present invention, the steps of obtaining device alarm information in the original language, invoking an AI translation engine pre-loaded with an operation and maintenance professional term library, performing semantic understanding and language conversion on the device alarm information, and generating a target semantics in the target language include: Invoke an AI translation engine pre-loaded with the operation and maintenance professional term library, and based on the device alarm information, match corresponding professional terms from the operation and maintenance professional term library to obtain a term matching result; Specifically, when invoking an AI translation engine pre-loaded with an operation and maintenance professional term library, after the AI translation engine receives the alarm information in the original language (such as the English alarm "overheating alarm detected in pump"), it immediately compares it with the term library, identifies and matches the professional terms included therein. For example, it accurately identifies that "overheating alarm" is an integral professional term and matches its standard Chinese translation "超温报警" from the library, rather than translating "overheating" and "alarm" separately in isolation. The output is thus an exact matching result between the source term and the target term, providing a key constraint for subsequent in-depth translation.

[0029] When constructing the operation and maintenance professional terms, it includes more than 3000 core terms (such as "sensor failure", "poor heat dissipation", "PLC programming"), with standard Chinese-English counterparts, industry interpretations, and usage scenarios marked, and pre-loaded into the AI translation engine to ensure the accuracy of term translation.

[0030] Taking the term matching result as a constraint condition, perform context semantic association analysis on the device alarm information, generate a translation result to be evaluated in the target language and conduct a reliability assessment to obtain a first confidence level; Specifically, taking the term matching result obtained in the previous step as a constraint, drive a deep neural network to perform context semantic association analysis on the alarm information. For example, on the premise that "overheating alarm" corresponds to "超温报警" and "pump" corresponds to "泵", the model will focus on analyzing the semantic relationship formed by "detected in" and these core terms, thereby generating a translation result to be evaluated such as "泵内检测到超温报警" that conforms to Chinese expression habits and has accurate terms. Subsequently, based on the output probability distribution of the model for this translation, automatically calculate a quantified reliability score, that is, the first confidence level. According to the first confidence level, perform corresponding processing procedures on the translation result to be evaluated to obtain the target semantics.

[0031] Specifically, based on the calculated first confidence level, the subsequent processing path is automatically selected. If the confidence level is higher than a preset threshold (e.g., 90%), the evaluation result is directly adopted as the final target semantics. If the confidence level is lower than the threshold (e.g., the confidence level may be low for the translation of polysemous words like "shell"), a manual correction process is automatically triggered. The evaluation result is pushed to an expert for review, and the manually confirmed correction result is used as the final reliable target semantic output, thereby ensuring the accuracy of the final output semantics in all scenarios.

[0032] In practical applications, this embodiment deeply embeds domain knowledge into the translation process. While ensuring the absolute accuracy of the translation of professional terms, it utilizes neural networks to achieve smooth conversion of contextual semantics and automatically sorts and processes data through a confidence mechanism. Ultimately, it achieves optimal synergy in professionalism, accuracy, and processing efficiency in the cross-language conversion of industrial alarm information.

[0033] like Figure 3 As shown, in an optional embodiment of the present invention, the AI ​​translation engine employs a deep neural network model, and the step of performing a corresponding processing procedure on the translation result to be evaluated based on the first confidence level to obtain the target semantics includes: Compare the first confidence level with a preset threshold; Specifically, the translation result to be evaluated output by the AI ​​translation engine and its calculated first confidence level are compared with a preset threshold.

[0034] If the first confidence level is greater than or equal to the preset threshold, the translation result to be evaluated is directly determined as the target semantics; Specifically, for high-quality translation results with a confidence level reaching or exceeding a preset threshold, they are directly identified as usable target semantics and used in subsequent knowledge base retrieval and display processes. For example, with a preset threshold of 90%, when AI translates "Check sensor wiring for looseness" and outputs "Check if the sensor wiring is loose," if the model calculates a confidence level of 95% (higher than the threshold) based on the overall output probability distribution of the translation, the result is automatically adopted as the final target semantic information. The aforementioned "Check if the sensor wiring is loose" with a confidence level of 95% will be directly stored in the knowledge base and displayed to the user without manual intervention, thus ensuring the processing efficiency of high-frequency, standard alarm scenarios.

[0035] If the first confidence level is less than the preset threshold, a manual correction process is executed, the translation result to be evaluated is pushed to the manual review end for correction, the corrected translation result is obtained, and the corrected translation result is used as the target semantics.

[0036] Specifically, for translation results with a confidence level lower than a preset threshold, they are automatically marked as low-quality outputs and an artificial correction process is triggered. For example, when an AI translates the professional phrase "shell pressure test", due to the polysemy of the word "shell", a translation method with a confidence level of only 70% may be output, such as "outer shell pressure test". At this time, the result to be evaluated and its low-confidence status are pushed to the expert review terminal. After being reviewed by domain experts, it may be corrected to the more accurate "housing pressure test". Finally, the corrected translation result produced through human-machine collaboration is used as the official target semantics, thus ensuring absolute accuracy in complex or ambiguous scenarios.

[0037] When this embodiment is applied in practice, through the comparison of the confidence threshold, automatic hierarchical control of the quality of translation results is achieved. While ensuring the translation efficiency of high-confidence standard terms, artificial correction is intelligently triggered for low-confidence difficult cases, and ultimately the translation accuracy can be improved.

[0038] As an optional embodiment of the present invention, the deep neural network model adopts a Transformer architecture and is used for: Through the self-attention mechanism of the Transformer architecture, the device alarm information is encoded to extract the global semantic features of the device alarm information and form an encoded feature representation. Specifically, the model first deeply encodes the input device alarm information through the self-attention mechanism of the Transformer. For example, for the alarm information "overheating alarm detected in motor after continuous operation", the self-attention mechanism can simultaneously analyze the mutual associations between all words in the sentence (such as "overheating", "alarm", "motor", "continuous operation"), capture semantic dependencies across any distance (such as the fault association between "overheating" and "motor", the temporal causal relationship between "continuous operation" and "detected"), thereby generating an encoded feature representation that integrates the global semantics of the entire sentence, laying a foundation for accurately understanding the complete scenario of the alarm.

[0039] Based on the encoded feature representation, through the multi-head attention mechanism of the Transformer architecture, combined with the term matching results obtained by matching from the operation and maintenance professional term library, context semantic association analysis is performed on the device alarm information to identify the professional term correspondence relationship of the device alarm information in the operation and maintenance scenario. Specifically, based on the encoded features, further perform domain-oriented context semantic analysis through the multi-head attention mechanism and in combination with the term knowledge retrieved in real time from the operation and maintenance professional term library. For example, when the term library provides the rule that "overheating alarm" should be preferably translated as "超温报警" and "motor" should be translated as "电机" in the transmission scenario, the multi-head attention of the model will use this domain knowledge as a strong constraint to guide the model to pay more attention to the binding relationship between "overheating" and "alarm" as a fixed term, and the professional scenario where "motor" and the context "continuous operation" together refer to "电机过热", so as to accurately identify the corresponding relationship of professional terms that conform to industrial operation and maintenance specifications in the alarm information and effectively avoid inaccurate translations such as "过热警报" and "马达" that may be generated by a general translation model.

[0040] Based on the identified corresponding relationship of the professional terms and the encoded feature representation, generate the translation result to be evaluated that conforms to the target language through the decoder of the Transformer architecture.

[0041] Specifically, based on the accurate semantic understanding (encoded features) and term constraints (corresponding relationship of professional terms) obtained in the first two steps, the model gradually generates a translation result that conforms to the grammar and professional habits of the target language through the decoder of the Transformer. For example, combining the global semantics (the motor fails after continuous operation) and the term correspondence relationship ("overheating alarm" - "超温报警", "motor" - "电机"), the decoder will generate a professional Chinese expression such as "电机持续运行后检测到超温报警" instead of the literal translation "在持续运行后在马达中检测到过热警报", ensuring that the translation result is both faithful to the original text and conforms to the professional expression specifications in the field of equipment operation and maintenance.

[0042] When applied in practice, the Transformer architecture deep neural network adopted in this embodiment realizes the accurate capture of the global semantics of the alarm information through the self-attention mechanism, and combines the real-time guidance of the professional term library to ensure the accurate correspondence and translation of domain terms in the context, and finally generates a translation result that conforms to the target language specifications and meets the professional requirements of industrial operation and maintenance, significantly improving the accuracy, professionalism and reliability of information transmission in the cross-language operation and maintenance scenario.

[0043] Optionally, the AI translation engine in the present invention can adopt two different deep learning architecture solutions in specific implementation. The core features of the two solutions are compared as follows: I. RNN-LSTM architecture solution (alternative solution) Structural layer features; The architecture employs a "temporal chain structure" as its core computational framework. This architecture addresses the "vanishing gradient problem in long texts" of traditional RNNs through a gating mechanism of "forget gate, input gate, and output gate." However, it still needs to strictly adhere to the temporal order of the text, i.e., process each word unit sequentially according to the order in which the words appear. The inherent drawback of this structure is that it cannot process text data in parallel, resulting in low efficiency when processing long documents such as equipment operation specifications. At the same time, its ability to capture contextual semantics is limited to the association between adjacent words, making it difficult to accurately capture the common long-distance semantic dependencies in equipment operation and maintenance texts, such as the professional collocation relationship between "calibration" and "offset."

[0044] Training layer features: During the model training phase, this architecture must be trained word by word in the order of the text, resulting in a long overall training cycle. It also exhibits poor stability when training on long text data, with model parameters easily converging to local optima rather than global optima. More importantly, this architecture faces challenges in domain adaptability: fine-tuning is difficult, and even when trained using specialized corpora from the equipment maintenance field, it struggles to quickly and effectively adapt to the specificities of industrial terminology, resulting in relatively poor domain adaptability.

[0045] II. Transformer Architecture Solution (Preferred Solution) Structural layer characteristics: The core improvement of this solution lies in completely abandoning the temporal chain structure and innovatively introducing a "self-attention mechanism." This mechanism allows the model to process all input words in parallel without relying on any temporal order, thus improving the processing speed of long texts related to equipment maintenance by more than 50% compared to the RNN-LSTM architecture. Another key advantage of the self-attention mechanism is its ability to accurately capture the semantic relationships between words at any distance in the text. For example, it can accurately identify the professional semantic binding of the word "shell" with "equipment" and "casing" in the context of equipment casing, rather than its general meaning. In addition, by introducing "multi-head attention," the model can analyze semantic information in parallel from different dimensions (such as part of speech, industry attributes, and contextual roles), thereby significantly improving the deep semantic understanding of professional texts related to equipment maintenance.

[0046] Training layer features: At the training level, this architecture supports efficient parallel training, allowing all words in the text to participate in the training process simultaneously. This reduces the training time to about one-third that of RNN-LSTM for the same amount of data. Through a self-attention mechanism, the model achieves bidirectional context learning, enabling it to simultaneously capture both positive and negative semantic relationships in the text (e.g., fully understanding the mutual indicative relationship between "overheating" and "alarm"). This architecture exhibits excellent domain adaptability, making it particularly suitable for domain fine-tuning. For example, the BERT-Industrial model (based on the Transformer architecture) specifically used in this invention only requires fine-tuning with a relatively small amount (e.g., more than 50,000 entries) of multilingual (e.g., Chinese-English bilingual) corpus in the equipment operation and maintenance field to quickly and accurately adapt to the industrial terminology system, achieving high-accuracy translation of professional texts.

[0047] In the field of industrial equipment operation and maintenance, which demands stringent requirements for translation accuracy, professionalism, and processing efficiency, the Transformer-based solution, with its efficient parallel processing and powerful long-range semantic capture capabilities at the structural layer, as well as its fast convergence and excellent domain fine-tuning characteristics at the training layer, has been established as the core and preferred implementation scheme of this invention. In contrast, the RNN-LSTM-based solution, limited by its inherent sequence processing characteristics and weaker domain adaptability, is only considered as a lightweight alternative in specific resource-constrained scenarios (such as some edge computing devices).

[0048] like Figure 3 As shown, in an optional embodiment of the present invention, after obtaining the corrected translation result, the method further includes: The equipment operation and maintenance knowledge base is updated based on the corrected translation results; Specifically, after obtaining the corrected translation result confirmed by human review, the corresponding entry in the equipment operation and maintenance knowledge base will be updated immediately. For example, when it is confirmed by human review that the best translation of "Check sensor wiring for looseness" should be "Check whether the sensor wiring is loose (use torque wrench)" rather than the simple version output by the original AI, the Chinese field of the alarm solution in the knowledge base will be directly updated to this more accurate and operational version, ensuring that all users who query it subsequently will see the corrected translation result.

[0049] Add the first terminology comparison relationship and contextual information contained in the corrected translation results to the operation and maintenance professional terminology database; Specifically, key term correspondence relationships and their specific contexts are extracted from the correction results. For example, during the correction process, it may be determined that "shell" should be translated as "housing" instead of the general translation "outer shell" in a specific equipment document. Not only is the correspondence "shell - housing" recorded, but the context in which it appears (such as "equipment shell pressuretest / equipment housing pressure test") is also associated, and this enhanced term entry is added to the operation and maintenance professional term library to provide more accurate guidance for subsequent translations.

[0050] Adjust the parameters of the deep neural network model according to at least one of the first term correspondence relationship and the obtained user translation correction suggestion information.

[0051] Specifically, the aforementioned high-quality data, that is, the corrected translation results, including the corrected accurate translation pairs and the newly added or optimized term correspondence relationships, will be used as training samples to fine-tune the deep neural network model of the AI translation engine (such as a customized industrial series model based on the Transformer architecture). For example, using a large number of correct examples of "shell - housing" in specific contexts to retrain the model to optimize its parameters, so that when encountering similar contexts in subsequent translations, it can more accurately select the professional translation method, achieving continuous and automated improvement of translation accuracy.

[0052] When this embodiment is applied in practice, by constructing a closed-loop optimization mechanism of "manual correction → knowledge base correction → term library enhancement → model iteration", the corrected translation results are transformed into improved content of the knowledge base, not only immediately improving the accuracy and practicality of the knowledge base content, but also continuously enriching the domain term system, and driving the translation model to perform adaptive learning, ultimately achieving coordinated reinforcement in three dimensions: professional accuracy, domain adaptability, and self-evolution ability.

[0053] Such as Figure 2 and Figure 4 As shown, as an optional embodiment of the present invention, the retrieving a matching operation and maintenance processing solution from the equipment operation and maintenance knowledge base based on the target semantics and performing multilingual association display includes: Analyze the keywords in the target semantics that conform to the target language, and query in the equipment operation and maintenance knowledge base based on the keywords to match the corresponding operation and maintenance processing solution that conforms to the target language and the original language; Specifically, extract keywords from the target semantics generated by AI translation. For example, parse keywords such as "sensor wiring" and "looseness" from the Chinese target semantics "Check whether the sensor wiring is loose". Based on these keywords, retrieve matching multilingual operation and maintenance processing solutions in the device operation and maintenance knowledge base. The solution entries in the knowledge base are quickly located through pre-stored Chinese-English association fields. For example, the corresponding English original text "Check sensor wiring for looseness" can be retrieved by associating with Chinese keywords.

[0054] Based on the user display interface, create a synchronous display view that includes at least a first display area and a second display area and display a language switch button for the user to select; wherein, the first display area is used to display the device alarm information that conforms to the original language and the corresponding operation and maintenance processing solution, and the second display area is used to display the content that conforms to the target semantics and the corresponding operation and maintenance processing solution.

[0055] Specifically, create a dual-region synchronous display view on the user interface; the first region displays the original alarm information and the matching original language solutions. For example, display the English alarm "overheating alarm" and its corresponding processing solutions "Check sensor wiring for looseness; Clean heat sink"; the second region synchronously displays the translated Chinese alarm "Over-temperature alarm" and its Chinese processing solutions "Check whether the sensor wiring is loose; Clean the heat sink". The interface also provides a language switch button, allowing the user to select the main display language according to the operation; when the user clicks the language switch button, dynamically adjust the display content and layout of the dual regions. If the user selects Chinese as the main language, the Chinese region occupies the dominant visual position, and the English region is used as a reference for comparison.

[0056] It should be noted that after displaying the operation and maintenance processing solutions, after the operation and maintenance personnel complete the alarm handling, submit the processing result (such as "The sensor wiring is loose and has been tightened") on the APP side, that is, obtain the operation feedback information based on the operation and maintenance processing solution, and automatically translate the operation feedback information into English through AI and store it in the knowledge base, while updating the term base and AI model training data, and finally realize the integrated linkage process of "alarm trigger - AI translation - solution display - processing feedback".

[0057] In practical applications, this embodiment achieves accurate matching of maintenance solutions in at least Chinese and English through intelligent semantic parsing, and presents content in at least multiple languages ​​(such as Chinese and English) intuitively through a dual-region synchronous display interface. Combined with dynamic language switching, this enables Chinese maintenance personnel to directly understand English device alarm information and obtain corresponding solutions, greatly improving the efficiency of information processing and operational accuracy in maintenance scenarios, while reducing the risk of misoperation; and laying the foundation for processing operation feedback information.

[0058] like Figure 2 , Figure 3 and Figure 4 As shown, in an optional embodiment of the present invention, the AI ​​translation engine employs a deep neural network model, and updating the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology database based on the operation feedback information includes: The operation feedback information is parsed to identify its information category; Specifically, the system automatically analyzes the operational feedback information submitted by maintenance personnel after handling alarms to intelligently identify its content category.

[0059] If the operation feedback information is identified to contain new equipment fault handling experience, the processing experience is converted into the operation and maintenance handling scheme that at least conforms to the original language and the target language through the AI ​​translation engine, and the operation and maintenance handling scheme is added as a new entry to the equipment operation and maintenance knowledge base. Specifically, when the analysis identifies that the feedback contains new fault handling experience, such as the maintenance personnel adding "a special tightening method for loose sensor wiring", the AI ​​translation engine will be called to automatically translate this Chinese experience into English, generate a standardized maintenance handling solution in both Chinese and English, and store it as a new entry in the equipment maintenance knowledge base.

[0060] If the operation feedback information includes suggestions for correcting existing terminology translations or adding new terms, then a corresponding standard terminology comparison relationship is established based on the suggestions for correcting or adding new terms, and the operation and maintenance professional terminology database is updated based on the standard terminology comparison relationship. Specifically, when the system identifies feedback containing suggestions for terminology correction or new terms, such as a user pointing out that "overheating alarm" should be optimized to "overheating alarm", the system will establish or update the standard terminology comparison between "overheating alarm" and "overheating alarm" accordingly, and update the operation and maintenance professional terminology database accordingly.

[0061] Based on the newly added entries and the standard terminology comparison relationship, the parameters of the deep neural network of the AI ​​translation engine are adjusted.

[0062] Specifically, the newly added multilingual (e.g., Chinese-English bilingual) operation and maintenance solution entries and the updated standard terminology comparison relationships are used as high-quality domain training data to fine-tune the parameters of the deep neural network (e.g., the Transformer architecture model) of the AI ​​translation engine, enabling the translation engine to continuously improve translation accuracy based on feedback from real-world scenarios.

[0063] In practical applications, this embodiment automatically transforms new fault handling experience into multilingual (e.g., Chinese-English bilingual) knowledge entries to expand the solution library. At the same time, it updates domain terminology standards in real time based on terminology correction suggestions and uses this high-quality data to fine-tune the parameters of the translation engine. This enables the entire system to continuously learn from real operation and maintenance scenarios and improve itself, significantly enhancing the accuracy, timeliness, and adaptability of cross-language operation and maintenance support.

[0064] like Figure 4 As shown, as an optional embodiment of the present invention, the pre-establishment of a device operation and maintenance knowledge base includes: Collect equipment operation and maintenance data, which includes at least equipment technical parameters, historical operation records, and alarm handling solutions; Specifically, a knowledge base is established by collecting data from multiple sources throughout the entire lifecycle of the equipment. This includes collecting core operation and maintenance information such as equipment technical parameters, historical operation records, and alarm handling solutions. For example, the range and accuracy parameters of instruments are collected from the DCS control system, standardized equipment start-up and shutdown procedures are extracted from the operation and maintenance logs, and typical fault causes and solutions are summarized from alarm history to ensure the professionalism and completeness of the information sources.

[0065] The equipment operation and maintenance data is structured and classified according to equipment type and information category to generate standardized data entries; wherein each standardized data entry contains at least a content field that conforms to the original language, and a reserved field for filling in the target language translation content; Specifically, the collected multi-source heterogeneous data is standardized through a two-level architecture of "device type - information category". For example, the "Temp-001" device is classified as "instrument device - alarm scheme", and standardized data entries in JSON format containing device ID, device type, information category and English content fields are generated to provide a unified and standardized input data structure for subsequent AI translation.

[0066] The AI ​​translation engine is invoked to translate the content fields in the standardized data entries that conform to the original language, generating corresponding target language translation content, and generating corresponding translation confidence based on the translation process; Specifically, an AI translation engine pre-loaded with a professional term library is automatically called to perform intelligent translation on the English content in the data entries for domain adaptation. For example, "Check sensor wiring for looseness" is accurately translated into "检查传感器接线是否松动" instead of the general "检查传感器线路是否松动". At the same time, the confidence level of this translation is calculated based on the output probability distribution of the Transformer model. For example, if the output probability is 0.98, the confidence level is 98%.

[0067] Associate and store the content fields in the original language, the translated content in the target language, and the translation confidence level to construct the device operation and maintenance knowledge base.

[0068] Specifically, finally, the English original text, the Chinese translation generated by AI (including at least two languages), and the calculated translation confidence level are associated and stored. For example, "overheating alarm" and "超温报警" and a confidence level of 98% are stored in the same knowledge entry, and hierarchical management is implemented according to the confidence level: content with a high confidence level (such as ≥95%) is directly published for use, and content with a low confidence level is marked for manual proofreading, thereby constructing a high-quality and trustworthy device operation and maintenance knowledge base with multilingual (such as Chinese-English bilingual) comparison and translation quality labels.

[0069] As Figure 4 shown, generally speaking, a three-step method of "information collection - structured arrangement - multilingual storage" is adopted to construct the device operation and maintenance knowledge base: Information collection layer: Connect to the device management system and DCS control system through interfaces to collect device technical parameters (such as instrument range, accuracy), operation specifications (such as start-stop procedures, calibration processes), and alarm handling solutions (such as fault causes, solution steps); at the same time, support maintenance personnel to manually enter new device information to ensure information integrity; Arrange the information according to a two-level classification structure of "device type - information category" (such as "instrument device - technical parameters", "transmission device - alarm handling"), and store it in a standardized JSON format to provide structured input for AI translation. The example is as follows: {“deviceId”:“Temp-001”, “deviceType”:“Instrument Equipment”, “infoType”:“Alarm Scheme”, “alarmType”:“overheating alarm”, “enContent”:“Check sensor wiring for looseness;Clean heat sink”, “zhContent”:“”, / / “” indicates that the AI-translated Chinese storage field “confidence”:0, / / Translation confidence field “updateTime”:“2025-11-05 10:30:00”}.

[0070] Multi-language storage layer: Adopts a dual-version storage mechanism such as "main language (English) + AI translation language (Chinese)", with each information entry associated with English and Chinese counterparts and AI translation confidence level (e.g., ≥95% is high confidence and can be used directly; <95% is marked for manual proofreading) to ensure consistency of multi-language versions.

[0071] In practical application, this embodiment uses "information collection - structured organization - multilingual storage" to build a device domain knowledge base. This transforms the originally scattered, multi-source, and unstructured device operation and maintenance information into unified, accurate, and directly usable structured knowledge assets for cross-language intelligent retrieval and decision-making. This not only significantly improves the query efficiency and accuracy of operation and maintenance knowledge in a multilingual environment, but also achieves automated control of translation quality through a confidence mechanism, providing a foundation for subsequent alarm linkage and intelligent decision-making.

[0072] As an optional embodiment of the present invention, the AI ​​translation engine adopts a deep neural network model, and the method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation further includes: Regularly extract new terminology entries from equipment manufacturers' technical manuals and industry standard documents; Specifically, regularly crawling equipment manufacturer manuals and industry standard documents (such as the ISO industrial equipment terminology standard) and extracting newly added professional terms establishes a knowledge acquisition channel that keeps pace with the development of industrial equipment technology, enabling continuous absorption of the latest technical vocabulary in the industry.

[0073] The newly added terminology entries are pushed to the expert review end, where the expert review end determines their standard terminology comparison relationship and interpretation, and generates standardized terminology data. Specifically, the automatically extracted term entries are submitted to an artificial review interface composed of domain experts. Based on industry norms and usage scenarios, the experts standardize these terms - determining their precise translations in different contexts, establishing the standard Chinese-English correspondence, supplementing professional interpretations and applicable conditions, thereby transforming the original term materials into standardized term data with complete structure and accurate definitions.

[0074] Add the said standardized term data to the operation and maintenance professional term library; Specifically, the standardized term data confirmed by expert review is systematically integrated into the operation and maintenance professional term library, realizing the dynamic incremental update of the term library.

[0075] Based on the said standardized term data, adjust the parameters of the deep neural network of the AI translation engine.

[0076] Specifically, use the newly added standardized term data as high-quality training samples to conduct targeted fine-tuning and parameter optimization on the deep neural network model of the AI translation engine to ensure the translation accuracy of new terms.

[0077] When this embodiment is applied in practice, by constructing a term library iteration mechanism of "external knowledge acquisition → expert review and standardization → dynamic update of term library → collaborative training and update of model", it not only realizes the self-evolution and intelligent iteration of the industrial term system, but also enables the AI translation engine to not only query new terms during translation, but more importantly, fundamentally improve the model's semantic understanding ability and generation accuracy of new terms, achieving the co-evolution of knowledge base iteration and the improvement of AI model capabilities, and maintaining term accuracy and semantic understanding ability.

[0078] As Figure 3 shown, combining the above embodiments, it can be summarized that the AI translation engine adopts a four-stage workflow of "preloading term library - context semantic translation - confidence verification - model iteration": Preloading of operation and maintenance professional term library: Construct a special term library for the equipment field, including more than 3,000 core terms (such as "sensor failure", "poor heat dissipation", "PLC programming"), mark the standard Chinese-English correspondence, industry interpretations and usage scenarios, and preload them into the AI translation engine to ensure the translation accuracy of terms; Context semantic translation: Based on the AI model (the fine-tuned BERT model) of the Transformer architecture, input structured equipment information (such as the English alarm plan "Check sensor wiring for looseness"), the model combines the term library and context semantics to output the Chinese translation result ("检查传感器接线是否松动") to avoid translation errors of isolated words; Translation confidence verification: Set a confidence threshold (e.g., 90%), and the AI ​​will automatically calculate the confidence level of the translation result. Confidence level ≥ 90%: Store directly in the knowledge base for use as the Chinese version; Confidence level <90%: Mark as "Pending Proofing", push to the operation and maintenance expert end, manually proofread and update to the knowledge base, and feed the proofreading results back to the AI ​​model for training; Model iteration and optimization: Regularly collect manual proofreading data and user feedback (such as translation correction suggestions in the use of Chinese interface) to fine-tune the AI ​​translation model, and update the operation and maintenance professional terminology library (such as adding emerging terms such as "AI diagnostic module") to continuously improve the translation accuracy (target accuracy ≥ 98%).

[0079] like Figure 5 As shown, this embodiment of the invention provides an AI-translated operation and maintenance knowledge base construction and alarm linkage system 200, which applies the AI-translated operation and maintenance knowledge base construction and alarm linkage method described in the above embodiment, including: The alarm information acquisition and translation module 210 is used to: acquire device alarm information that conforms to the original language, call an AI translation engine that is preloaded with an operation and maintenance professional terminology library, perform semantic understanding and language conversion on the device alarm information, and generate target semantics that conforms to the target language, wherein the operation and maintenance professional terminology library is pre-established and stores device operation and maintenance terms. The solution matching and display module 220 is used to: retrieve matching operation and maintenance solutions from the equipment operation and maintenance knowledge base based on the target semantics and display them in multiple languages, wherein the equipment operation and maintenance knowledge base is pre-established and stores equipment operation and maintenance information throughout its entire life cycle; Feedback acquisition module 230 is used to: acquire operation feedback information based on the operation and maintenance processing scheme; The knowledge update module 240 is used to update the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology base based on the operation feedback information.

[0080] The specific implementation method of this embodiment can be referred to the corresponding implementation method described above, and will not be described again here.

[0081] like Figure 6 As shown in the figure, an electronic device 300 provided in this embodiment of the invention includes a memory 310 and a processor 320; the memory 310 is used to store a computer program; the processor 320 is used to implement the above-described method for constructing an operation and maintenance knowledge base and alarm linkage based on AI translation when executing the computer program.

[0082] Alternatively, an electronic device 300 includes a memory 310 and a processor 320 coupled to the memory 310; the memory 310 is configured to store a computer program; and the processor 320 is configured to perform the following operations when the computer program is executed: The system obtains device alarm information that conforms to the original language, calls an AI translation engine preloaded with an operation and maintenance terminology database, performs semantic understanding and language conversion on the device alarm information, and generates target semantics that conforms to the target language. The operation and maintenance terminology database is pre-established and stores device operation and maintenance terms. Based on the target semantics, matching operation and maintenance solutions are retrieved from the equipment operation and maintenance knowledge base and displayed in multiple languages. The equipment operation and maintenance knowledge base is pre-established and stores operation and maintenance information for the entire life cycle of the equipment. Obtain operation feedback information based on the aforementioned operation and maintenance processing solution; Based on the operation feedback information, update the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology database.

[0083] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the AI ​​translation-based operation and maintenance knowledge base construction and alarm linkage method described above.

[0084] Alternatively, a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the following operations: The system obtains device alarm information that conforms to the original language, calls an AI translation engine preloaded with an operation and maintenance terminology database, performs semantic understanding and language conversion on the device alarm information, and generates target semantics that conforms to the target language. The operation and maintenance terminology database is pre-established and stores device operation and maintenance terms. Based on the target semantics, matching operation and maintenance solutions are retrieved from the equipment operation and maintenance knowledge base and displayed in multiple languages. The equipment operation and maintenance knowledge base is pre-established and stores operation and maintenance information for the entire life cycle of the equipment. Obtain operation feedback information based on the aforementioned operation and maintenance processing solution; Based on the operation feedback information, update the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology database.

[0085] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0086] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

[0087] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation, characterized in that, include: The system obtains device alarm information that conforms to the original language, calls an AI translation engine preloaded with an operation and maintenance terminology database, performs semantic understanding and language conversion on the device alarm information, and generates target semantics that conforms to the target language. The operation and maintenance terminology database is pre-established and stores device operation and maintenance terms. Based on the target semantics, matching operation and maintenance solutions are retrieved from the equipment operation and maintenance knowledge base and displayed in multiple languages. The equipment operation and maintenance knowledge base is pre-established and stores operation and maintenance information for the entire life cycle of the equipment. Obtain operation feedback information based on the aforementioned operation and maintenance processing solution; Based on the operation feedback information, update the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology database.

2. The method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation as described in claim 1, characterized in that, The process of obtaining device alarm information that conforms to the original language, calling an AI translation engine preloaded with an operation and maintenance terminology database, performing semantic understanding and language conversion on the device alarm information, and generating target semantics that conforms to the target language includes: The AI ​​translation engine, which is preloaded with the operation and maintenance terminology database, is invoked. Based on the device alarm information, the corresponding professional terms are matched from the operation and maintenance terminology database to obtain the term matching result. Using the term matching results as constraints, the device alarm information is subjected to contextual semantic association analysis to generate a translation result to be evaluated that conforms to the target language and to conduct a reliability assessment to obtain a first confidence level. Based on the first confidence level, the corresponding processing procedure is performed on the translation result to be evaluated to obtain the target semantics.

3. The method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation as described in claim 2, characterized in that, The AI ​​translation engine employs a deep neural network model. The step of performing a corresponding processing procedure on the translation result to be evaluated based on the first confidence level to obtain the target semantics includes: Compare the first confidence level with a preset threshold; If the first confidence level is greater than or equal to the preset threshold, the translation result to be evaluated is directly determined as the target semantics; If the first confidence level is less than the preset threshold, a manual correction process is executed, the translation result to be evaluated is pushed to the manual review end for correction, the corrected translation result is obtained, and the corrected translation result is used as the target semantics.

4. The method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation as described in claim 3, characterized in that, The deep neural network model adopts the Transformer architecture and is used for: The device alarm information is encoded using the self-attention mechanism of the Transformer architecture, and the global semantic features of the device alarm information are extracted to form an encoded feature representation. Based on the encoded feature representation, and through the multi-head attention mechanism of the Transformer architecture, combined with the term matching results obtained from the operation and maintenance terminology database, the contextual semantic association analysis of the device alarm information is performed to identify the professional terminology correspondence of the device alarm information in the operation and maintenance scenario. Based on the identified correspondences of the technical terms and the encoded feature representations, the decoder of the Transformer architecture generates the translation result to be evaluated that conforms to the target language.

5. The method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation according to claim 3 or 4, characterized in that, After obtaining the corrected translation result, the process also includes: The equipment operation and maintenance knowledge base is updated based on the corrected translation results; Add the first terminology comparison relationship and contextual information contained in the corrected translation results to the operation and maintenance professional terminology database; The parameters of the deep neural network model are adjusted based on at least one of the first terminology comparison relationship and the obtained user translation correction suggestions.

6. The method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation according to any one of claims 1-4, characterized in that, The step of retrieving matching operation and maintenance solutions from the equipment operation and maintenance knowledge base based on the target semantics and displaying them in multiple languages ​​includes: The keywords in the target semantics that conform to the target language are parsed, and the query is performed in the equipment operation and maintenance knowledge base based on the keywords to obtain the corresponding operation and maintenance processing solution that conforms to the target language and the original language; Based on the user display interface, a synchronized display view including at least a first display area and a second display area is created, and a language switching button is displayed for the user to select; wherein, the first display area is used to display the device alarm information and the corresponding operation and maintenance handling scheme in accordance with the original language, and the second display area is used to display the operation and maintenance handling scheme in accordance with the target semantics.

7. The method for constructing and linking an operation and maintenance knowledge base based on AI translation according to any one of claims 1-4, characterized in that, The AI ​​translation engine employs a deep neural network model, and updating the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology database based on the operation feedback information includes: The operation feedback information is parsed to identify its information category; If the operation feedback information is identified to contain new equipment fault handling experience, the processing experience is converted into the operation and maintenance handling scheme that at least conforms to the original language and the target language through the AI ​​translation engine, and the operation and maintenance handling scheme is added as a new entry to the equipment operation and maintenance knowledge base. If the operation feedback information includes suggestions for correcting existing terminology translations or adding new terms, then a corresponding standard terminology comparison relationship is established based on the suggestions for correcting or adding new terms, and the operation and maintenance professional terminology database is updated based on the standard terminology comparison relationship. Based on the newly added entries and the standard terminology comparison relationship, the parameters of the deep neural network of the AI ​​translation engine are adjusted.

8. The method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation according to any one of claims 1-4, characterized in that, Pre-establishing a knowledge base for equipment operation and maintenance includes: Collect equipment operation and maintenance data, which includes at least equipment technical parameters, historical operation records, and alarm handling solutions; The equipment operation and maintenance data is structured and classified according to equipment type and information category to generate standardized data entries; wherein each standardized data entry contains at least a content field that conforms to the original language, and a reserved field for filling in the target language translation content; The AI ​​translation engine is invoked to translate the content fields in the standardized data entries that conform to the original language, generating corresponding target language translation content, and generating corresponding translation confidence based on the translation process; The content fields conforming to the original language, the translated content in the target language, and the translation confidence level are associated and stored to construct the equipment operation and maintenance knowledge base.

9. The method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation according to any one of claims 1-4, characterized in that, The AI ​​translation engine employs a deep neural network model, and the method for constructing an operation and maintenance knowledge base and linking alarms based on AI translation also includes: Regularly extract new terminology entries from equipment manufacturers' technical manuals and industry standard documents; The newly added terminology entries are pushed to the expert review end, where the expert review end determines their standard terminology comparison relationship and interpretation, and generates standardized terminology data. Add the standardized terminology data to the operation and maintenance professional terminology library; Based on the standardized terminology data, the parameters of the deep neural network of the AI ​​translation engine are adjusted.

10. A system for constructing an operation and maintenance knowledge base and linking alarms based on AI translation, characterized in that, The method for constructing and linking an operation and maintenance knowledge base based on AI translation as described in any one of claims 1-9 includes: The alarm information acquisition and translation module is used to: acquire device alarm information that conforms to the original language, call an AI translation engine with a pre-loaded operation and maintenance terminology library, perform semantic understanding and language conversion on the device alarm information, and generate target semantics that conforms to the target language. The operation and maintenance terminology library is a pre-established library that stores device operation and maintenance terms. The solution matching and display module is used to: retrieve matching operation and maintenance solutions from the equipment operation and maintenance knowledge base based on the target semantics and display them in multiple languages, wherein the equipment operation and maintenance knowledge base is pre-established and stores equipment operation and maintenance information throughout its entire life cycle; The feedback acquisition module is used to: acquire operation feedback information based on the operation and maintenance processing plan; The knowledge update module is used to update the equipment operation and maintenance knowledge base and / or the operation and maintenance terminology base based on the operation feedback information.

Images