Industrial knowledge query method and device, equipment and medium

Abstract

The invention discloses an industrial knowledge query method and device, equipment and a medium, and the industrial knowledge query method comprises the steps: obtaining a voice instruction through a user terminal, and converting the voice instruction into a query text under the support of a voice recognition model; then, the equipment data acquisition end synchronously collects real-time operation data of the industrial equipment, and the real-time operation data and the query text are subjected to space-time alignment; afterwards, the knowledge query end extracts industrial knowledge meeting requirements from the local knowledge base according to the real-time operation data and the query text and based on a preset retrieval strategy, and finally presents the industrial knowledge to the user through the user terminal. The ability of workers and managers to obtain knowledge and solve problems in the industrial production process is greatly improved, the urgent demand of modern industrial production for timely and accurate information query is met, and improvement of production efficiency and optimization of decision are promoted.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an industrial knowledge query method, apparatus, equipment, and medium. Background Technology

[0002] Currently, knowledge retrieval in industrial production environments primarily relies on traditional document retrieval, keyword matching, and simple database search methods. These methods, lacking contextual understanding capabilities, struggle to accurately identify and quickly respond to complex or technical terms. While some systems have incorporated cloud-based AI technology to enhance intelligence, network dependence and data security concerns lead to issues such as response delays, data privacy risks, and a lack of real-time dynamic adjustment capabilities in field applications. As the manufacturing industry's demands for production efficiency, equipment reliability, and information security continue to rise, existing knowledge retrieval technologies can no longer fully meet enterprises' needs for efficient, accurate, and secure knowledge acquisition. Therefore, current technologies are insufficient to meet the demands of efficient, intelligent, safe, and integrated knowledge retrieval and fault handling in modern industrial production. There is an urgent need to develop a localized industrial knowledge retrieval method that integrates multimodal information processing and possesses self-learning capabilities to improve the intelligence level of the manufacturing industry. Summary of the Invention

[0003] The embodiments of the present invention provide an industrial knowledge query method, apparatus, equipment and medium, which aims to solve the technical problem that existing knowledge query methods in industrial production environments are insufficient to meet the current industrial production needs.

[0004] In a first aspect, embodiments of the present invention provide an industrial knowledge query method applied to a knowledge retrieval system. The system includes a user terminal, an equipment data acquisition terminal, and a knowledge query terminal. The method includes: receiving a voice command input by a user to query industrial knowledge through the user terminal, and converting the voice command into query text through a speech recognition model; synchronously collecting real-time operating data of industrial equipment through the equipment data acquisition terminal based on the query text and its query time; querying a local knowledge base based on the real-time operating data and the query text using a preset retrieval strategy through the knowledge query terminal to obtain matching industrial knowledge query results; and displaying the industrial knowledge query results to the user through the user terminal.

[0005] Secondly, embodiments of the present invention also provide an industrial knowledge query device for executing the industrial knowledge query method described above.

[0006] Thirdly, embodiments of the present invention also provide a computer device, the computer device including a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to perform the steps of the above-described industrial knowledge query method.

[0007] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, can implement the steps of the above-described industrial knowledge query method.

[0008] Compared with the prior art, the beneficial effects of the present invention are:

[0009] In the technical solution of this invention, the industrial knowledge query method acquires voice commands through a user terminal and converts them into query text with the support of a speech recognition model. Then, the equipment data acquisition terminal synchronously collects real-time operating data of the industrial equipment and aligns it spatiotemporally with the query text. Subsequently, the knowledge query terminal, based on the real-time operating data and the query text, and using a preset retrieval strategy, extracts the required industrial knowledge from a local knowledge base and finally presents it to the user through the user terminal. This significantly enhances the ability of workers and managers to acquire knowledge and solve problems during industrial production, meets the urgent need for timely and accurate information retrieval in modern industrial production, and promotes improved production efficiency and optimized decision-making. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A flowchart of the industrial knowledge query method provided by the present invention;

[0012] Figure 2 This is a first sub-flowchart of the industrial knowledge query method provided by the present invention;

[0013] Figure 3 This is a sub-flowchart of the second sub-flowchart of the industrial knowledge query method provided by the present invention;

[0014] Figure 4 The third sub-flowchart of the industrial knowledge query method provided by the present invention;

[0015] Figure 5 The fourth sub-flowchart of the industrial knowledge query method provided by the present invention;

[0016] Figure 6 The fifth sub-flowchart of the industrial knowledge query method provided by the present invention;

[0017] Figure 7 The sixth sub-flowchart of the industrial knowledge query method provided by the present invention;

[0018] Figure 8 A schematic block diagram of the unit of the industrial knowledge query device provided by the present invention;

[0019] Figure 9 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0022] It should also be understood that the terminology used in this specification is for the purpose of describing embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0024] To address the technical problem that existing knowledge retrieval methods in industrial production environments are insufficient to meet current industrial production needs, this invention discloses an industrial knowledge retrieval method. This method is applied to a knowledge retrieval system in a manufacturing workshop, which consists of a user terminal, an equipment data acquisition terminal, and a knowledge retrieval terminal.

[0025] Reference Figures 1 to 7 The method includes the following steps:

[0026] S110. Receive the user's voice command to query industrial knowledge through the user terminal, and convert the voice command into query text through a speech recognition model;

[0027] S120. Real-time operating data of industrial equipment is synchronously collected through the device data acquisition terminal according to the query text and its query time;

[0028] S130. The knowledge query terminal performs a query from the local knowledge base based on the real-time operation data and the query text using a preset retrieval strategy to obtain matching industrial knowledge query results.

[0029] S140. The industrial knowledge query results are displayed to the user through the user terminal.

[0030] The user first inputs voice commands through a user terminal, which includes a device for inputting commands—specifically, a portable device integrated with voice recognition capabilities and equipped with a noise-reducing microphone array. The user terminal uses a built-in voice recognition model to convert and recognize the user's voice commands, obtaining the corresponding query text. This voice recognition model is typically based on an end-to-end deep learning network and has undergone specialized training for industrial scenarios, enabling it to accurately recognize professional voice commands from on-site operators, providing foundational data for subsequent retrieval. Specifically, the user terminal collects audio through the microphone array and uses an industrial-grade voice recognition model for noise reduction and voice conversion. This industrial-grade voice recognition model uses the Kaldi framework and a custom acoustic model, specifically optimized for industrial scenarios, and can recognize PLC fault codes, equipment models, and other technical terms, allowing users to directly input professional information via voice.

[0031] After the instruction conversion is completed, the equipment data acquisition terminal begins to synchronously collect real-time operating data of the industrial equipment. The data acquisition terminal monitors the industrial equipment through various hardware components such as temperature and pressure sensors, PLC interfaces, and vibration monitoring devices. A synchronization mechanism ensures accurate temporal and spatial alignment between real-time data and query text; that is, it uses timestamps or synchronization protocols to correlate the latest parameter status of the equipment with the user's query intent. After data collection, the collected data undergoes preprocessing, such as filtering out outliers or extracting features, to ensure data quality and integrity. This synchronization process provides a crucial foundation of on-site information for subsequent retrieval, supporting the contextual relevance and timeliness of search results.

[0032] Subsequently, based on synchronized real-time data and query text, the system uses a combination of keyword retrieval and semantic vector retrieval strategies to search the local knowledge base for industrial knowledge that matches the real-time operational data and query text as query results. The knowledge query terminal is a central server located locally, which can connect to and utilize the computing power of local edge computing device nodes. Keyword retrieval first matches the user's explicit keywords, such as fault codes, process parameter names, and core terms in operating instructions. Semantic vector retrieval, on the other hand, converts the query text and entries in the knowledge base into low-dimensional vector representations and calculates their semantic similarity, thereby better understanding the user's intent and the context of the scenario. Specifically, the local knowledge base includes a basic process library obtained by batch converting paper SOP documents using an OCR scanner, storing structured documents such as enterprise standard operating procedures and equipment manuals, as well as a case experience library accumulated from the maintenance work order system, containing historical fault handling solutions. The combination of these two retrieval methods significantly improves the accuracy and adaptability of the retrieval, especially in complex or ambiguous query scenarios, ensuring that the matched industrial knowledge query results meet actual needs.

[0033] Finally, the system outputs the matched industrial knowledge query results to the user terminal, which also includes an information interaction device. Specifically, the information interaction device is an operating interface with graphical display and voice broadcast functions. The output process has been formatted and optimized to ensure that users can easily understand and implement the solution. The entire process realizes a closed loop from user voice input, data synchronization, mixed retrieval, to result output, providing industrial field workers with a high-speed and accurate knowledge support means.

[0034] Furthermore, this method offers flexibility and scalability. For instance, the weighting of the retrieval strategy can be adjusted based on different industrial scenarios, edge computing technology can be introduced to optimize local semantic understanding capabilities, and even autonomous learning mechanisms can be incorporated into key stages to continuously improve retrieval accuracy and user experience. Ultimately, this approach achieves an efficient, intelligent, and reliable industrial knowledge query solution, significantly enhancing the intelligence level and operational efficiency of production sites.

[0035] In one embodiment, reference is made to Figure 2 The steps in S130 include:

[0036] S131. Extract keywords from the query text using a search engine;

[0037] S132. Generate a semantic vector from the domain semantic model of the query text input;

[0038] S133. Perform a joint query of keyword retrieval and semantic vector retrieval based on the keywords and the semantic vector to obtain an initial result candidate set from the local knowledge base;

[0039] S134. Combining the real-time operation data, dynamically filter conflict schemes from the initial result candidate set to obtain industrial knowledge query results that match the real-time operation data and the query text.

[0040] In the process of extracting industrial terminology keywords from query text, the system uses word segmentation algorithms combined with a predefined industry terminology dictionary to identify keywords in the user's query. Specifically, the system uses the Elasticsearch search engine for keyword extraction. In practical use, when a user asks "How to handle excessive vibration of a dryer E12", the system parses out the core terms: "dryer" (equipment type), "E12" (equipment model), and "excessive vibration" (fault phenomenon). This dictionary is built based on the company's equipment manual and historical work orders, employing a dual filtering mechanism: first, it matches standard models using regular expressions, and then filters noun-based technical terms based on part-of-speech tagging (POS) to ensure that invalid words are excluded.

[0041] The query text is input into a pre-trained domain semantic model, generating corresponding semantic vectors. This model is based on a deep learning architecture optimized for industrial scenarios, such as a low-dimensional semantic space embedding model, like the industrial version of BERT. Trained on a large amount of industrial scenario text, this model can efficiently and accurately map professional terms to a fixed semantic space. The generated semantic vectors characterize the deeper meaning of the query, rather than relying solely on surface keywords, thus better capturing the underlying intent and contextual information of the query, laying the foundation for subsequent deep semantic matching.

[0042] The system simultaneously performs keyword retrieval and semantic vector retrieval from the local knowledge base, forming a joint query. Keyword retrieval utilizes fast matching mechanisms, such as inverted indexes or hash algorithms, to filter out all candidate solutions containing the extracted keywords. Semantic vector retrieval, on the other hand, calculates the similarity between the query vector and the semantic vectors of answers or solutions stored in the knowledge base, such as cosine similarity, to select the most relevant candidates. The results of both are combined to cross-validate the relevance of the candidate set, forming a relatively comprehensive and accurate preliminary set of solution candidates.

[0043] To ensure that the selected solutions meet the actual needs of the on-site operating conditions, the candidate results are dynamically filtered based on the operating constraints in real-time operational data. At this point, real-time collected operating parameters, such as sensor data on temperature, pressure, and vibration, are used to determine whether each solution conflicts with the current equipment status or process conditions. For example, if a solution requires the equipment to operate at a high temperature, but real-time temperature information shows that the equipment is currently operating at a low temperature, this solution will be filtered out. Through this dynamic filtering mechanism, inapplicable or potentially conflicting solutions can be eliminated, thereby selecting the solution that best meets the actual needs of the on-site scenario.

[0044] Through the close coordination of the above steps, from keyword extraction and deep semantic retrieval to working condition constraint filtering, the accuracy and practicality of solution matching are greatly improved, ensuring that the optimal solution that conforms to the process and equipment status can be obtained quickly and reliably in the industrial field, effectively supporting on-site fault handling and process optimization.

[0045] Furthermore, referring to Figure 3 The steps in S133 include:

[0046] S1331. When the keywords include equipment fault feature keywords, a preset decision path is matched through a fault association tree model.

[0047] S1332. Based on the decision path, reorder the results of the joint query to generate the initial result candidate set.

[0048] In practical applications, to further improve the accuracy of Hall effect results, a joint query combining keyword retrieval and semantic vector retrieval is performed to obtain an initial candidate set of results that matches the scenario. This process goes beyond simple matching. In the initial keyword retrieval stage, the system first extracts industry-specific terms or fault-related keywords from the user's query, such as "fault code," "temperature anomaly," and "pressure fluctuation." When the system detects that the query text contains characteristic keywords related to equipment faults, it activates a matching mechanism based on a fault association tree model.

[0049] The fault association tree model is a decision tree structure based on the causal chain of equipment faults and fault diagnosis paths, used to match pre-defined diagnostic paths. When the system identifies a keyword representing a certain fault in the equipment, it matches it along a predefined path—the diagnostic decision path—within the fault association tree. These paths typically consist of fault type, diagnostic conditions, corresponding solutions, and handling suggestions. By matching with the fault association tree model, the system can quickly pinpoint the diagnostic path of potential faults, thereby filtering out the solutions most relevant to the equipment fault. In this step, once a match is successful, a preliminary set of solutions closely related to the currently queried fault characteristics is formed.

[0050] After obtaining the initial set of solutions, the system employs a decision path-based re-ranking strategy to further optimize the candidate set. Specifically, the matching results of the diagnostic path are used as the priority ranking criterion, and the degree of consistency between the path and the query is used as the weight, prioritizing solutions with higher consistency and closer relevance to fault characteristics. This effectively improves the relevance and practicality of the solutions. The re-ranking result considers not only keyword matching but also the priority of path matching, ensuring that the final output of the initial candidate set not only matches the expressed keywords but is also closely related to equipment fault inference and better meets actual on-site needs.

[0051] Furthermore, referring to Figure 4 The steps in S134 include:

[0052] S1341. Filter the equipment fault-related schemes in the initial result candidate set using the fault association tree model;

[0053] S1342. When the number of initial results after filtering is not unique, prioritize the initial results after filtering based on the historical adoption rate.

[0054] S1343. Extract key features from the real-time running data, perform conflict filtering on the sorted initial results based on the key features, and select the first one among the initial results obtained after conflict filtering as the industrial knowledge query result.

[0055] The system utilizes a fault association tree model to filter solutions related to equipment faults. Specifically, the fault association tree model filters out solutions from candidate results that are associated with the current equipment fault type based on the symptoms, diagnostic paths, and solutions of the equipment fault. This process first analyzes the equipment's real-time operating data, such as sensor data on temperature, pressure, and vibration, to identify states that match specific fault characteristics. Then, it combines this with predefined paths in the fault association tree to quickly filter out the set of solutions most closely related to the current equipment fault.

[0056] If, after filtering using the fault association tree, several candidate results still exist that are not unique, the system will prioritize these solutions based on historical adoption rates. Specifically, the system will query and statistically analyze the actual adoption frequency of each solution in historical data—that is, the proportion of times a particular solution was used during equipment maintenance or fault resolution out of the total number of solutions. Based on these statistics, solutions with high adoption rates will be ranked higher, thus prioritizing recommendations for solutions that are effective in similar scenarios. This process significantly improves the practicality and efficiency of the solutions, making field operators more inclined to adopt efficient solutions that have been validated in the field.

[0057] By combining real-time operating constraints, the system performs conflict filtering on the ranked candidate results. This step is particularly critical because on-site operating conditions often change rapidly, and some solutions may conflict with the current actual conditions. For example, a solution might involve equipment operating outside the permissible temperature range under high load, or a solution might require shutdown for maintenance when the equipment is continuously running. Therefore, the system eliminates potentially conflicting solutions based on real-time collected operating parameters, ensuring the feasibility of the selected solution under the current operating conditions. After conflict filtering, the solution ranked first is presented as the final solution and pushed to the operator. In this embodiment, the closed-loop optimization process formed by fault correlation screening, historical verification ranking, and operating condition conflict filtering effectively improves the adaptability, scientific nature, and reliability of the solutions.

[0058] In one embodiment, reference is made to Figure 5 The steps following S140 also include:

[0059] S150. Perform adaptive optimization on the local knowledge base based on the feedback information of the industrial knowledge query results input by the user from the user terminal.

[0060] In the industrial knowledge query method of this invention, after the solution is output to the user terminal, the system also collects user feedback data through the user terminal to continuously optimize the local knowledge base. Specifically, while displaying or broadcasting the solution, the user terminal has a feedback interface that allows operators to evaluate the effectiveness, applicability, or accuracy of the solution, usually through simple operation responses or detailed text descriptions. This user feedback data is collected and stored in real time, serving as the core input for subsequent optimization.

[0061] After collecting feedback data, the system invokes its integrated learning mechanism to analyze user evaluations of solutions and determine the actual effectiveness of a solution in the current scenario. If multiple feedback responses indicate that a solution is rated as "ineffective" or "difficult to execute," the system will reduce its weight according to a predefined strategy, decreasing its priority in similar future scenarios. Simultaneously, the system will increase the confidence score of solutions rated as "effective," raising their priority in relevant scenarios. Based on this feedback, the system can automatically adjust the priority ranking and parameter configuration of relevant solutions in the knowledge base, and even revise or supplement solution content to continuously improve the accuracy and usability of the knowledge base.

[0062] Furthermore, by incorporating machine learning algorithms, such as reinforcement learning or incremental learning, the system continuously adjusts retrieval model parameters based on feedback data, including keyword matching weights and semantic vector model parameters, thereby improving overall retrieval performance. All these optimization operations are performed locally, ensuring data security while guaranteeing real-time and targeted knowledge updates. This adaptive optimization mechanism enables the local knowledge base to continuously learn from real-world usage experience, effectively adapting to changing production environments and equipment conditions, and gradually improving the accuracy and efficiency of fault diagnosis and solution recommendation.

[0063] Furthermore, referring to Figure 6 The steps of S150 include:

[0064] S151. Obtain the user's validity labeling information and actual execution result information for the industrial knowledge query results through the user terminal as the feedback information;

[0065] S152. The feedback information is context-associated with the corresponding real-time running data and the query text and stored as incremental data.

[0066] S153. When the incremental data reaches the preset training threshold, comparative learning training of the domain semantic model is started based on the incremental data.

[0067] S154. Embed the trained domain semantic model into the knowledge retrieval system, and synchronously adjust the scheme adoption weights in the local knowledge base according to the incremental data.

[0068] After obtaining user feedback on the solution's validity and actual execution results at the user terminal, the system stores this data as feedback. Specifically, after the solution is presented, on-site operators can select "valid" or "invalid" via the interface, or fill in a brief description to record their evaluation of the solution. Simultaneously, the system associates this feedback information with the corresponding real-time equipment operating data and the user's query text, forming context-sensitive feedback entries. This associative storage method helps the system understand the specific scenario behind the user feedback, contributing to the accuracy of subsequent model training.

[0069] The system then continuously collects this associated feedback data until it accumulates to a preset training threshold, such as a certain number of feedback samples or the number of feedback samples within a continuous feedback period. At this point, it triggers contrastive learning training of the domain semantic model based on incremental data. Contrastive learning is a training strategy that strengthens the model's ability to distinguish between different categories or similarities. It helps the model better differentiate specialized semantics in different scenarios, improving its understanding capabilities. For example, the system compares newly collected feedback data with existing semantic vectors, optimizing model parameters to make the query semantic mapping more accurate in similar scenarios, thereby improving the solution matching effect.

[0070] After training, the new domain semantic model is embedded into the knowledge retrieval system. Simultaneously, the system adjusts the adoption weights of solutions in the knowledge base based on feedback data. Specifically, the adoption weight of a solution is an indicator that measures its applicability and priority. During model training, the model identifies which solutions have been repeatedly verified as effective in practice and assigns them higher weights; conversely, solutions repeatedly marked as "ineffective" have their weights reduced. This adjustment is synchronously updated in the knowledge base in the background, enabling the system to prioritize recommending solutions that have been verified as efficient and meet the specific needs of the situation during the next knowledge retrieval.

[0071] This embodiment enables the knowledge retrieval system to form a closed-loop, continuously optimized system under the operation of industrial knowledge query methods. This system comprises user feedback, context-dependent storage, incremental model training, and knowledge base weight adjustment, all working together to drive the system to continuously learn and adapt to changing field environments. This achieves dynamic evolution of knowledge and continuous optimization of solution recommendations. This not only significantly improves the accuracy of fault diagnosis and solution matching but also enhances the adaptability and intelligence of the knowledge base to actual field changes.

[0072] In one embodiment, reference is made to Figure 7 The steps following S154 also include:

[0073] S155. When the information in the incremental data conflicts with the information in the local knowledge base, the false conflict information is eliminated through semantic similarity analysis, and the real conflict information is pushed and submitted to the user for manual review through the user terminal.

[0074] The system utilizes an updated, pre-defined domain semantic model to perform semantic similarity analysis on the information in incremental data and the information in the local knowledge base. Specifically, the system converts the description, parameter settings, and process elements of each specific solution into semantic vectors, and then calculates the similarity between these vectors, for example, using cosine similarity. When the semantic similarity between a new solution and an old solution exceeds a preset threshold, a pseudo-conflict may exist, meaning that the two solutions are actually highly similar in content, differing only slightly in expression or details, but having little impact on the on-site process.

[0075] After detecting high similarity, the system further analyzes the detailed differences between the two schemes, eliminating pseudo-conflicts caused by essentially identical content and only different descriptions. This process mainly relies on the structured text comparison of the two schemes in terms of parameter configuration, process description, and operation steps, combined with the results of semantic similarity analysis, to exclude schemes that do not substantially conflict, thereby reducing misjudgments. In this way, pseudo-conflicting schemes that mislead on-site adoption and have no practical impact can be effectively eliminated.

[0076] For genuine conflicts in content or process that remain after eliminating false conflicts, the system will label these genuine conflict situations with information such as conflict type and involved solutions. Combining this with the requirements for manual review, the system will push the conflict information to the manual review interface via the user terminal. Manual reviewers can make a final judgment based on the detailed analysis information provided by the system and the actual situation on site, deciding whether to adjust the solution or take other countermeasures. This process ensures the practicality of the solution while avoiding errors introduced by automatic judgment, guaranteeing the scientific nature of the process solution and the safety of on-site operation. By eliminating false conflicts through semantic similarity analysis, accurately identifying genuine conflicts, and finally confirming them through human review, the reliability and usability of the knowledge base solutions are greatly improved. This provides strong support for the fault repair and process improvement of industrial equipment, promoting continuous system optimization and intelligent development.

[0077] Figure 8 This is a schematic block diagram of an industrial knowledge query device 600 provided in an embodiment of the present invention. Figure 8 As shown, corresponding to the above-described industrial knowledge query method, the present invention also provides an industrial knowledge query device 600. This industrial knowledge query device 600 includes a unit for executing the above-described industrial knowledge query method, and the device can be configured in a desktop computer, tablet computer, smartphone, or other terminal.

[0078] Specifically, please refer to Figure 8 The industrial knowledge query device 600 includes:

[0079] The instruction acquisition unit 610 is used to receive a voice instruction input by a user to query industrial knowledge through the user terminal, and convert the voice instruction into query text through a speech recognition model;

[0080] The equipment information acquisition unit 620 is used to synchronously acquire real-time operating data of industrial equipment through the equipment data acquisition terminal based on the query text and its query time.

[0081] The solution generation unit 630 is used to retrieve matching industrial knowledge query results from the local knowledge base based on the real-time operation data and the query text using the knowledge query terminal and a preset retrieval strategy.

[0082] The solution output unit 640 is used to display the industrial knowledge query results to the user through the user terminal.

[0083] In one embodiment, the scheme generation unit 630 includes:

[0084] The keyword extraction unit is used to extract keywords from the query text through a search engine.

[0085] A semantic vector generation unit is used to generate semantic vectors from the domain semantic model of the query text input;

[0086] The initial result candidate set acquisition unit is used to perform a joint query of keyword retrieval and semantic vector retrieval based on the keywords and the semantic vector, and obtain an initial result candidate set from the local knowledge base;

[0087] The conflict scheme filtering unit is used to dynamically filter conflict schemes from the initial result candidate set by combining the real-time operation data, so as to obtain industrial knowledge query results that match the real-time operation data and the query text.

[0088] In one embodiment, the initial result candidate set acquisition unit includes:

[0089] The fault association decision unit is used to match a preset decision path through a fault association tree model when the keywords include equipment fault feature keywords.

[0090] The result reordering unit is used to reorder the results of the joint query based on the decision path to generate the initial result candidate set.

[0091] Furthermore, the conflict scheme filtering unit includes:

[0092] The equipment failure related solution screening unit is used to screen equipment failure related solutions in the initial result candidate set through a fault association tree model;

[0093] The priority sorting unit is used to sort the candidate results after screening based on the historical adoption rate when the number of candidate results after screening is not unique.

[0094] The final selection unit is used to extract key features from the real-time running data, perform conflict filtering on the sorted initial results based on the key features, and select the first one among the initial results obtained after conflict filtering as the industrial knowledge query result.

[0095] Furthermore, the solution output unit 640 is further provided with:

[0096] An adaptive optimization unit is used to perform adaptive optimization on the local knowledge base based on feedback information from the user inputting the industrial knowledge query results from the user terminal.

[0097] Furthermore, the adaptive optimization unit includes:

[0098] The feedback data generation unit is used to obtain the user's validity labeling information and actual execution result information of the industrial knowledge query results through the user terminal as the feedback information;

[0099] The feedback data generation unit is used to contextually associate the feedback information with the corresponding real-time running data and the query text, and store it as incremental data.

[0100] A contrastive learning training unit is used to initiate contrastive learning training on the domain semantic model based on the incremental data when the incremental data reaches a preset training threshold.

[0101] The adoption weight adjustment unit is used to embed the trained domain semantic model into the knowledge retrieval system and synchronously adjust the adoption weights of the solutions in the local knowledge base according to the incremental data.

[0102] In one embodiment, after adopting the weight adjustment unit, the following is further provided:

[0103] The manual final review unit is used to push the real conflict information after eliminating false conflicts in the conflict information through semantic similarity analysis when the information in the incremental data conflicts with the information in the local knowledge base, and then submit it to the user for manual review through the user terminal.

[0104] The aforementioned industrial knowledge query device 600 can be implemented as a computer program, which can, for example... Figure 9 It runs on the computer device shown.

[0105] Please see Figure 9 , Figure 9 This is a schematic block diagram of a computer device 500 provided in an embodiment of this application. The computer device 500 can be a terminal or a server. The terminal can be an electronic device with communication functions, such as a desktop computer, tablet computer, or smartphone. The server can be a standalone server or a server cluster composed of multiple servers.

[0106] See Figure 9 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0107] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform an industrial knowledge query method.

[0108] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0109] The internal memory 504 provides an environment for the execution of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute an industrial knowledge query method.

[0110] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0111] The processor 502 is used to run a computer program 5032 stored in a memory to implement the steps of the above method.

[0112] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0113] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0114] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. When executed by a processor, the program instructions cause the processor to perform the steps of the above-described method.

[0115] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0116] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0117] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0118] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0119] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0120] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A method for querying industrial knowledge, characterized in that, Applied to a knowledge retrieval system, the system including a user terminal, a device data acquisition terminal, and a knowledge query terminal, the method includes: The user terminal receives voice commands from users to query industrial knowledge, and converts the voice commands into query text using a speech recognition model. The device data acquisition terminal synchronously collects real-time operating data of industrial equipment based on the query text and its query time. The knowledge query terminal retrieves matching industrial knowledge query results from the local knowledge base based on the real-time operation data and the query text using a preset retrieval strategy. The industrial knowledge query results are displayed to the user through the user terminal.

2. The industrial knowledge query method according to claim 1, characterized in that, The step of obtaining matching industrial knowledge query results by querying the local knowledge base based on the real-time operation data and the query text using the knowledge query terminal and a preset retrieval strategy includes: Keywords are extracted from the query text using a search engine; Generate a semantic vector from the domain semantic model of the query text input; Based on the keywords and semantic vectors, a joint query of keyword retrieval and semantic vector retrieval is performed to obtain an initial result candidate set from the local knowledge base; By combining the real-time operational data, conflicting schemes are dynamically filtered from the initial result candidate set to obtain industrial knowledge query results that match the real-time operational data and the query text.

3. The industrial knowledge query method according to claim 2, characterized in that, The step of performing a joint query of keyword retrieval and semantic vector retrieval based on the keywords and semantic vectors to obtain an initial result candidate set from the local knowledge base includes: When the keywords include equipment fault feature keywords, a preset decision path is matched using a fault association tree model; The results of the joint query are reordered based on the decision path to generate the initial result candidate set.

4. The industrial knowledge query method according to claim 3, characterized in that, The step of dynamically filtering conflicting schemes from the initial result candidate set based on the real-time operational data to obtain industrial knowledge query results that match the real-time operational data and the query text further includes: The device fault-related schemes in the initial result candidate set are filtered using a fault association tree model; When the number of initial results after filtering is not unique, the initial results after filtering are prioritized based on the historical adoption rate. Key features are extracted from the real-time running data, and conflict filtering is performed on the sorted initial results based on the key features. The first result among the initial results obtained after conflict filtering is selected as the industrial knowledge query result.

5. The industrial knowledge query method according to claim 2, characterized in that, Following the step of displaying the industrial knowledge query results to the user through the user terminal, the method further includes: The local knowledge base is adaptively optimized based on the feedback information from the user's input of the industrial knowledge query results from the user terminal.

6. The industrial knowledge query method according to claim 5, characterized in that, The step of performing adaptive optimization of the local knowledge base based on the feedback information of the industrial knowledge query results input by the user from the user terminal includes: The user terminal obtains the user's validity labeling information and actual execution result information of the industrial knowledge query results as the feedback information; The feedback information is stored in context with the corresponding real-time running data and the query text, and is used as incremental data. When the incremental data reaches a preset training threshold, comparative learning training of the domain semantic model is initiated based on the incremental data. The trained domain semantic model is embedded into the knowledge retrieval system, and the solution adoption weights in the local knowledge base are adjusted synchronously based on the incremental data.

7. The industrial knowledge query method according to claim 6, characterized in that, The step of embedding the trained domain semantic model into the knowledge retrieval system and synchronously adjusting the solution adoption weights in the local knowledge base based on the incremental data further includes: When the information in the incremental data conflicts with the information in the local knowledge base, the real conflict information is pushed after eliminating false conflicts through semantic similarity analysis and then submitted to the user terminal for manual review.

8. An industrial knowledge query device, characterized in that, Used to perform the industrial knowledge query method as described in any one of claims 1 to 7.

9. A computer device, characterized in that, The computer device includes a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program stored in the memory to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, can implement the steps of the method as described in any one of claims 1 to 7.

Images