Fault diagnosis system and method, computing device cluster and computer readable storage medium

Abstract

The embodiment of the invention discloses a fault diagnosis system and method, a computing device cluster and a computer readable storage medium, and relates to the technical field of fault processing. The system comprises a first interaction platform used for uploading a question text; the large model is used for querying the experience database according to input equipment data and question text to obtain target experience and initial fault diagnosis information, returning the target experience and the initial fault diagnosis information to the first interaction platform, and dynamically analyzing the target experience, the equipment data and the question text to obtain a fault diagnosis result; a fault diagnosis result is returned to the first interaction platform; the second interaction platform is used for inputting a query text of a fault diagnosis result into the large model; and the large model is also used for intelligently analyzing the questioning text to obtain a final diagnosis result, and returning the final diagnosis result to the first interaction platform. Therefore, according to the scheme, an interaction mechanism can be provided for the user, so that the user can obtain detailed information of the fault.

Description

Technical Field

[0001] This application relates to the field of fault handling technology, specifically to a fault diagnosis system, method, computing device cluster, and computer-readable storage medium. Background Technology

[0002] In the field of equipment fault diagnosis, there has long been a reliance on two main approaches: manual experience-based inspection and simple threshold judgment. Manual experience-based inspection is susceptible to errors due to the operator's skill level, and its long inspection cycle and limitations lead to delayed fault detection. Simple threshold judgment, relying solely on a single fixed value to trigger fault alerts, cannot adapt to the complex operating conditions and dynamic changes of equipment, easily overlooking potential hazards and resulting in a high rate of missed detections.

[0003] Related technologies utilize artificial intelligence (AI) for equipment fault diagnosis. However, AI typically outputs fault diagnosis results directly based on the input information, making it difficult for users to know the diagnostic basis, fault logic, and other key details, which brings inconvenience to subsequent fault investigation and equipment maintenance decisions. Summary of the Invention

[0004] This application provides a fault diagnosis system, method, computing device cluster, and computer-readable storage medium, which can provide an interactive mechanism to users so that users can obtain detailed information about the fault.

[0005] In a first aspect, embodiments of this application provide a fault diagnosis system, including: The first interactive platform is used to upload question text; The large model is used to query the experience database based on the input device data and the question text to obtain the target experience and the initial fault diagnosis information, and return the target experience and the initial fault diagnosis information to the first interactive platform. It also performs dynamic analysis on the target experience, the device data and the question text to obtain the fault diagnosis result, and returns the fault diagnosis result to the first interactive platform. The second interactive platform is used to input the follow-up text of the fault diagnosis results into the large model; The large model is also used to perform intelligent analysis on the follow-up text to obtain a final diagnostic result, and return the final diagnostic result to the first interactive platform.

[0006] Secondly, embodiments of this application provide a fault diagnosis method applied to a fault diagnosis system, the fault diagnosis system comprising: a first interactive platform, a second interactive platform, and a large model; the method comprising: Obtain device data and the question text uploaded by the first interactive platform; The device data and the question text are input into the large model to obtain the target experience and initial fault diagnosis information obtained by the large model from the experience database based on the device data and the question text, and the target experience and initial fault diagnosis information are returned to the first interactive platform. Based on the large model, the target experience, the equipment data and the question text are dynamically analyzed to obtain the fault diagnosis result, and the fault diagnosis result is returned to the first interactive platform; Obtain follow-up text of the fault diagnosis results input from the second interactive platform; The large model is used to perform intelligent analysis on the follow-up text to obtain the final diagnosis result, and the final diagnosis result is returned to the first interactive platform.

[0007] Thirdly, embodiments of this application also provide a computing device cluster, including at least one computing device, each computing device including a processor and a memory; the processor is used to execute instructions stored in the memory so that the computing device cluster performs the methods provided in the various optional implementations described in the embodiments of this application.

[0008] Fourthly, embodiments of this application also provide a computer-readable storage medium storing instructions that, when executed on at least one computing device, cause the at least one computing device to perform the methods provided in the various optional implementations described in the embodiments of this application.

[0009] Fifthly, embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described in embodiments of this application.

[0010] The embodiments of this application have the following beneficial effects: The large-scale model can query the experience database based on equipment data and query text uploaded by the first interactive platform to obtain target experience. Based on this target experience, it performs experience-based diagnosis to determine if the equipment has malfunctioned, obtaining initial fault diagnosis information. This target experience and initial fault diagnosis information are then returned to the first interactive platform. The large-scale model can further perform more intelligent dynamic analysis based on the target experience, equipment data, and query text to obtain fault diagnosis results, which are then returned to the first interactive platform. Thus, the first interactive platform can obtain detailed and accurate fault information based on the target experience, initial fault diagnosis information, and fault diagnosis results. The second interactive platform can follow up on the fault diagnosis results by inputting the follow-up query text into the large-scale model. The large-scale model can intelligently analyze the follow-up query text to obtain the final diagnosis result, which is then returned to the first interactive platform. This provides an interactive mechanism for users, allowing them to ask about fault details and repair methods. The large-scale model responds to users' personalized follow-up queries, providing various information to help users accurately locate and resolve equipment faults. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the structure of a fault diagnosis system provided in an embodiment of this application; Figure 2 This is a schematic diagram of the architecture of a fault diagnosis system provided in an embodiment of this application; Figure 3 This is an interactive schematic diagram of a fault diagnosis system provided in an embodiment of this application; Figure 4 This is a schematic diagram of the steps of a fault diagnosis method provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a fault diagnosis device provided in one embodiment of this application; Figure 6 This is a schematic diagram of the structure of a computing device cluster provided in one embodiment of this application. Detailed Implementation

[0013] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0014] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar entities or operations, and do not necessarily require or imply any such actual relationship or order between these entities or operations. It should be understood that such terms can be used interchangeably where appropriate; this is merely a way of distinguishing entities or operations with the same attributes in the description of embodiments of this application.

[0015] Reference Figure 1 This is a schematic diagram of the structure of a fault diagnosis system. For example... Figure 1 As shown, the fault diagnosis system may include a first interactive platform, a second interactive platform, and a large model, and the first interactive platform, the second interactive platform, and the large model may establish direct or indirect communication connections through wired or wireless means. Figure 1 The example provided is a fault diagnosis system consisting of a first interactive platform and a second interactive platform. In actual applications, a fault diagnosis system may include multiple first interactive platforms and multiple second interactive platforms, and there is no limitation on this.

[0016] In the after-sales service scenario for equipment, the first interaction platform can be a platform used by after-sales personnel (such as web pages, client applications, software applications, etc.), and the second interaction platform can be a platform used by engineers (such as web pages, client applications, software applications, etc.). Engineers can refer to personnel with expert experience in the field of the equipment. Expert experience refers to the sum of professional knowledge, practical skills, and problem-solving wisdom accumulated by relevant personnel in the field through long-term practice and learning; expert experience integrates a deep understanding of industry trends and the ability to adapt to complex scenarios, providing efficient and practical guidance for troubleshooting and related decision-making. After-sales personnel can receive equipment fault feedback and provide on-site or remote fault diagnosis and repair services. After-sales personnel possess both equipment expertise and communication and coordination skills, enabling them to efficiently resolve faults and restore normal equipment operation, as well as respond to user inquiries and ensure service experience and user satisfaction.

[0017] Figure 2 This is a schematic diagram of the architecture of a fault diagnosis system provided in an embodiment of this application; see reference. Figure 2 This fault diagnosis system may include a web front-end, an Internet of Things (IoT) cloud, and a database. The web front-end is the user-facing web application presentation and interaction layer. It transforms the data and functions provided by the back-end into a visual interface, handles page interaction logic, and serves as a bridge connecting the user and the system's core functions. The web front-end may include a first interaction platform and a second interaction platform. The web front-end can communicate and connect with the IoT cloud.

[0018] IoT Cloud is a cloud service platform supporting the connection, management, and data processing of IoT devices. It can receive real-time operational data and historical fault records uploaded by multiple devices, and leverage cloud computing power to extract fault features, issue early warnings, and trace root causes, making fault analysis more efficient and accurate. IoT Cloud may include, but is not limited to, a Web backend, business service modules, and an AI-agent module. The Web backend is the technical layer responsible for handling web application business logic, data storage, and interface services. It can read and write data, execute business rules, and interact with the Web frontend. The Web backend can interact with the first and second interaction platforms of the Web frontend to obtain question texts sent by the first and second interaction platforms. The business service module refers to the set of modules provided by IoT Cloud for device fault diagnosis-related business functions. Business service modules may include, but are not limited to: device information, product information, shadow and / or real-time diagnostics. Device information is used to manage and maintain basic device information, such as device model, parameters, and type, providing device identity and attribute basis for fault analysis. Product information refers to the product series information to which the equipment belongs, including product specifications, design standards, production batches, etc., which can help analyze common faults from a product perspective. The "shadow" is a digital mapping of the equipment in the cloud, which can synchronize equipment status and cache equipment commands, facilitating the acquisition of real-time and historical status data during fault analysis. Real-time diagnostics is used to monitor equipment operating status in real time, analyze abnormal indicators, and quickly locate faults. The intelligent agent module is a module for intelligent analysis in the IoT cloud. It can call the large model of the AI ​​cloud via the network, building a bridge between business services and the AI ​​cloud, providing intelligent decision-making, anomaly identification, and root cause reasoning support for equipment fault diagnosis. In another embodiment, the large model of the AI ​​cloud can be directly deployed in the IoT cloud, and the fault diagnosis system directly performs fault analysis based on the large model deployed in the IoT cloud.

[0019] A database can include multiple different databases, each used to store different types of data. For example, a database can include a log service database, a distributed database, and a document database. The log service database can be used for the collection, storage, and rapid retrieval of unstructured / semi-structured data such as device operation logs and anomaly alarm logs, providing raw log support for fault tracing. The distributed database, using a distributed architecture, can achieve high-concurrency storage and efficient querying of massive amounts of structured data such as device operation indicators and fault records, ensuring data throughput and access performance during fault analysis. The document database can be used to store document data such as device specifications, fault handling manuals, and product parameter documents, supporting flexible querying and providing knowledge references for fault root cause analysis. In one embodiment, the database can include an experience database, which stores historical fault information for multiple devices. Based on the historical fault information of multiple devices of the same type / model, fault patterns and / or fault judgment rules for that type / model of device can be summarized. The IoT cloud can acquire and query data from the database, and can also store data in the database.

[0020] In one embodiment, the fault diagnosis system may further include: a device for determining the load type of the device, determining a data acquisition frequency based on the load type of the device, and acquiring multimodal device data according to the data acquisition frequency; the device is further configured to perform data preprocessing on the multimodal device data to obtain the device data, and upload the device data to the large model.

[0021] Equipment load types can be categorized into high-load and low-load types. High-load equipment refers to devices that bear heavy workloads during operation, whose core components are prone to wear and heat generation, and / or whose operating parameters fluctuate significantly, such as motors and gearboxes. Low-load equipment refers to devices that experience moderate workloads during operation, whose core components experience minimal wear and / or whose operating parameters are relatively stable, such as fans. The equipment load type can be predetermined. Data acquisition frequencies corresponding to different equipment load types can be preset, with high-load equipment having a higher data acquisition frequency than low-load equipment. For example, the data acquisition frequency for high-load equipment might be 1 time per 1 second, while the data acquisition frequency for low-load equipment might be 1 time per 5 seconds.

[0022] Equipment data can include equipment operational data and basic equipment data. Equipment operational data refers to various data generated during equipment operation that reflect the equipment's status, performance, operating conditions, and environmental impacts. This data may include time-related information. Basic equipment data may include equipment load type, equipment model, serial number, installation location, and / or maintenance records. Basic equipment data can be pre-acquired and uploaded to the IoT cloud or database for use in fault diagnosis using large-scale models. Equipment operational data can be collected in real-time according to the data acquisition frequency corresponding to the equipment's load type and uploaded to the cloud via 5G, Ethernet, or Message Queuing Telemetry Transport (MQTT) protocol, with a transmission latency typically less than or equal to 10 seconds. When the network is down, the device can temporarily store the operational data locally and re-upload it to the cloud upon reconnection.

[0023] Equipment operation data includes multimodal data, which can include structured and unstructured data. Structured data may include, but is not limited to, current (accuracy ±0.1A), voltage (±0.1V), temperature (±0.5℃), pressure (±0.01MPa), and / or rotational speed (±1r / min), which can be acquired through the equipment's built-in sensors. Unstructured data may include, but is not limited to, vibration waveforms (sampling rate 1000Hz, recording vibration frequency / amplitude) and / or equipment operation sound data (sampling rate 44.1kHz, WAV format), which can be acquired through external vibration sensors and microphones.

[0024] After collecting device data, the device can preprocess the data to obtain preprocessed device data, and then upload the preprocessed device data to the cloud for use by large models. In another embodiment, the device can upload raw device data, and the cloud can receive the raw device data, preprocess it, and store the preprocessed device data for use by large models.

[0025] Data preprocessing for equipment data can include, but is not limited to, denoising, completion, and normalization. Denoising can use wavelet transform to remove environmental noise (such as factory background noise) from vibration / sound data; completion can use linear interpolation to fill in missing data caused by temporary communication interruptions (effective when missing data is ≤5 seconds); normalization can map data to the [0,1] interval (such as temperature 0-100℃→0-1) to avoid the influence of dimensional differences on the model.

[0026] In this way, the device can upload device data to the cloud, providing a basis for device fault diagnosis. Moreover, the device data has been preprocessed, which greatly reduces the computing cost of fault diagnosis in the cloud and significantly improves the efficiency and accuracy of device fault diagnosis.

[0027] Figure 3 This is an interactive schematic diagram of a fault diagnosis system provided in an embodiment of this application; see reference. Figure 3 The system consists of a first interactive platform for uploading question text; a large model for querying the experience database based on the input device data and question text to obtain target experience and initial fault diagnosis information, returning the target experience and initial fault diagnosis information to the first interactive platform, and dynamically analyzing the target experience, device data, and question text to obtain fault diagnosis results, which are then returned to the first interactive platform; a second interactive platform for inputting follow-up questions based on the fault diagnosis results into the large model; and the large model for intelligently analyzing the follow-up questions to obtain the final diagnosis result, which is then returned to the first interactive platform.

[0028] The first interactive platform can be a platform for uploading question texts according to a preset cycle, or it can be a platform uploaded in response to fault diagnosis requests triggered by relevant users. For example, if the device is an air conditioner, the first interactive platform is a platform used by after-sales personnel. When the user of the air conditioner discovers an abnormality, they can proactively consult after-sales personnel to inquire whether the air conditioner is faulty. The first interactive platform can obtain the question texts entered by after-sales personnel via voice or text.

[0029] The question text may include, but is not limited to: descriptive text of equipment data, analysis requirements for equipment data, and / or description requirements of fault conditions. For example, the question text may include: You are a heating equipment data analysis expert, capable of analyzing the operating data of heating machines. The attachment contains equipment data for a heating unit, specifically time-series data generated during the operation of a single heating unit from 20:30 to 21:30 on the first, second, third, and fourth days. The first line of the file contains the attribute names of the heating unit during operation, including (time, set temperature, inlet water temperature, outlet water temperature, outdoor ambient temperature, defrosting status, exhaust superheat, compressor return (suction) temperature, outdoor coil temperature, four-way valve status, exhaust temperature). The second line onwards contains attribute values. It is known that the data from the fourth day's operation shows an anomaly, but the location of the anomaly is unknown. Please use an anomaly detection algorithm based on the 3σ criterion to compare the differences between the data from these four days, calculating only the columns (set temperature, inlet water temperature, outlet water temperature, outdoor ambient temperature, defrosting status, exhaust superheat, compressor return (suction) temperature, outdoor coil temperature, four-way valve status, exhaust temperature). Identify the abnormal data from the fourth day and the corresponding time period, and describe in words what equipment malfunction occurred during that time period. The attached content contains equipment data for a heating machine, specifically time-series data generated during the operation of a heating machine from 20:30 to 21:30 on the first, second, third, and fourth days. The first line of the file contains the attribute names of the heating machine during operation, including (time, set temperature, inlet water temperature, outlet water temperature, outdoor ambient temperature, defrosting status, exhaust superheat, compressor return (suction) temperature, outdoor coil temperature, four-way valve status, and exhaust temperature). The second line onwards contains attribute values. This can be considered a descriptive text for the equipment data. The following section, "It is known that the data from the fourth day's operation is abnormal, but the location of the abnormality is unknown. Please use an anomaly detection algorithm based on the 3σ criterion to compare the differences between the four days' operating data, calculating only the columns (set temperature, inlet water temperature, outlet water temperature, outdoor ambient temperature, defrosting status, exhaust superheat, compressor return (suction) temperature, outdoor coil temperature, four-way valve status, and exhaust temperature)," can be considered an analysis requirement for the equipment data. The following section, "Identify the abnormal data and corresponding time periods from the fourth day, and describe in words what equipment malfunction occurred during that time period," can be considered a description requirement for the malfunction.

[0030] The first interactive platform can also select or actively upload device data associated with the query text from the database. The fault diagnosis system can input the device data and the query text sent by the first interactive platform into the large model. The large model can perform a preliminary analysis of the problem based on the query text and device data, and then query the experience database for target experience associated with the device data and query text. Target experience may include device data of the same type / model as the device data, or fault judgment rules summarized from historical fault data of multiple devices of the same type / model. The large model can analyze the device data and query text based on the queried target experience to determine whether there is an anomaly in the device, thereby obtaining initial fault diagnosis information. The analysis of device data and query text based on target experience by the large model has less computational load than the dynamic analysis of target experience, device data, and query text by the large model, thus obtaining target experience and initial fault diagnosis information more efficiently. The large model returns the target experience and initial fault diagnosis information to the first interactive platform, which can display the target experience and initial fault diagnosis information so that users of the first interactive platform can gain a preliminary understanding and location of the fault after seeing the target experience and initial fault diagnosis information.

[0031] After obtaining target experience and initial fault diagnosis information, the large-scale model can continue to perform more intelligent dynamic analysis based on the acquired data to obtain fault diagnosis results. Specifically, the large-scale model can dynamically analyze target experience, equipment data, and query text together to obtain fault diagnosis results. During this dynamic analysis, the large-scale model performs more intelligent calculations, with a computational workload exceeding that required for obtaining the initial fault diagnosis information, resulting in more detailed and accurate fault diagnosis results. The large-scale model can then return the fault diagnosis results to the primary interactive platform, which can display the results so that users on the platform can understand and locate the fault in detail.

[0032] In one embodiment, the large model analyzes equipment data and query text based on target experience. This is achieved by comparing whether the equipment data matches the fault data associated with the query text as represented by the target experience, thereby obtaining initial fault diagnosis information. This initial fault diagnosis information is obtained through a simple matching calculation between the equipment data and the target experience. The fault diagnosis results obtained by the large model through dynamic analysis of target experience, equipment data, and query text represent intelligent analysis results derived from deep reasoning and exploration by the large model. Therefore, the fault diagnosis results are more detailed and accurate.

[0033] In one embodiment, when the large model returns target experience, initial fault diagnosis information, and fault diagnosis results to the first interactive platform, it can simultaneously return the entire session record and / or the analysis process performed internally by the large model to the first interactive platform for display. The session record is a collection of various interactive information recorded chronologically during the interaction between the first and / or second interactive platforms and the large model.

[0034] The second interactive platform can read the entire conversation log. It can also obtain follow-up questions from users (e.g., engineers) regarding the fault diagnosis results. The fault diagnosis system can input these follow-up questions into the large model. The large model can combine the follow-up questions and the conversation log to perform intelligent analysis (deep reasoning) to obtain the final diagnosis result, which is then returned to both the first and second interactive platforms. Follow-up questions can include, but are not limited to, expert experience text and / or fault-related follow-up questions. Expert experience text allows the large model's responses to follow-up questions to better reflect the actual fault scenario, avoiding generalized conclusions. Expert experience text can supplement industry-specific fault characteristics, diagnostic logic, and other implicit knowledge, helping the large model quickly locate core issues and improve the targeting of follow-up questions and the professional accuracy of responses.

[0035] For example, the follow-up question text could be: "When the compressor's return air (suction) temperature is consistently 15 degrees higher than the inlet water temperature, it indicates an excessively high suction temperature; in the three abnormal time periods identified above, did the suction temperature exhibit excessively high levels?" Here, "When the compressor's return air (suction) temperature is consistently 15 degrees higher than the inlet water temperature, it indicates an excessively high suction temperature" can be considered expert experience text, while "in the three abnormal time periods identified above, did the suction temperature exhibit excessively high levels?" can be considered a troubleshooting follow-up question text.

[0036] By employing the technical solution of this application embodiment, the large model can query the experience database based on device data and the question text uploaded by the first interactive platform to obtain target experience. Based on this target experience, it performs experience-based diagnosis to determine whether a device malfunction has occurred, obtaining initial fault diagnosis information. The target experience and initial fault diagnosis information are then returned to the first interactive platform. The large model can further perform more intelligent dynamic analysis based on the target experience, device data, and question text to obtain fault diagnosis results, which are then returned to the first interactive platform. Thus, the first interactive platform can obtain detailed and accurate fault information based on the target experience, initial fault diagnosis information, and fault diagnosis results. The second interactive platform can follow up on the fault diagnosis results by inputting the follow-up question text into the large model. The large model can intelligently analyze the follow-up question text to obtain the final diagnosis result, which is then returned to the first interactive platform. This provides an interactive mechanism for users, allowing them to ask about fault details and repair methods. The large model responds to the user's personalized follow-up questions, providing various information to help users accurately locate and resolve device faults.

[0037] Follow-up text can be entered by the user of the second interaction platform based on prompts from a large model (or other AI), or it can be entered proactively after reviewing the conversation history.

[0038] Based on the above technical solution, as an example, the large model can also be used to determine the uncertainty of the fault diagnosis result, and automatically output a device information request to the second interactive platform when the uncertainty is greater than the uncertainty threshold; the second interactive platform can also be used to input follow-up text of the fault diagnosis result to the large model based on the device information request.

[0039] When determining fault diagnosis results, large-scale models can simultaneously determine the uncertainty of those results. Higher uncertainty generally indicates lower reliability of the diagnosis. Large-scale models can determine the uncertainty based on a quantitative evaluation process considering input data quality, fault scenario adaptability, and their own knowledge boundaries. The uncertainty can be determined by comprehensively analyzing the completeness and reliability of equipment data, the coverage of target experience, the rarity of fault types, and the complexity of multi-fault coupling. Large-scale models can also determine the uncertainty of fault diagnosis results by quantifying the impact of factors such as missing data, feature ambiguity, and lack of seen-fitting scenarios.

[0040] When the uncertainty of the fault diagnosis result exceeds a preset uncertainty threshold, it indicates that the reliability of the fault diagnosis result is low, and the large model needs more information to arrive at a more accurate final diagnosis result. Therefore, the large model can output a device information request to the second interactive platform. Based on the device information request, the second interactive platform obtains the data required by the large model and then inputs follow-up text of the fault diagnosis result into the large model. The device information request is used to request the data needed to improve the reliability of the final diagnosis result. This follow-up text can include the data needed to improve the reliability of the final diagnosis result. For example, if the fault diagnosis result is a four-way valve leak, the device information request could be "Please provide the compressor return gas temperature." Because a four-way valve leak may affect the pressure balance of the equipment system, thus affecting the compressor return gas temperature. Based on this device information request, the follow-up text can include the compressor return gas temperature. The large model combines the compressor return gas temperature to verify whether the four-way valve is leaking, thereby obtaining an accurate final diagnosis result.

[0041] By employing the technical solution of this application embodiment, the uncertainty of the fault diagnosis result can provide engineers with a reliability reference, avoiding maintenance errors due to misjudgment. It also provides direction for subsequent follow-up inquiries, optimization, data supplementation, or adjustment of diagnostic logic, making fault diagnosis both efficient and rigorous, further improving the decision-making accuracy and practicality of the fault diagnosis system. When the uncertainty of the fault diagnosis result is large, a device information request can be issued to obtain follow-up text, thereby improving the accuracy of the final diagnostic result based on the follow-up text.

[0042] Based on the above technical solution, as an example, the database of the fault diagnosis system can be used to store session records; the second interactive platform can view the session records and actively input follow-up text of the fault diagnosis results.

[0043] After reviewing the session logs, users on the second interactive platform can choose to actively input follow-up questions if they have any questions, require further details, or inquire about troubleshooting methods. For example, follow-up questions could include "Will this fault affect adjacent gearboxes?" or "What tools are needed to replace the bearing?" The large model can combine the follow-up questions and data from the session logs to output a more accurate final diagnostic result that meets the user's needs.

[0044] By adopting the technical solution of this application embodiment, users can actively input follow-up questions to meet personalized inquiries. Users can specifically explore uncovered information such as fault correlation impacts and maintenance details, avoiding the limitations of general conclusions. By leveraging historical data in the conversation records and the reasoning capabilities of large models, follow-up questions and answers become more coherent and targeted, reducing the cost of repeated communication. In addition, the final diagnosis can help users quickly obtain key information such as maintenance operation guidance and risk prediction, shortening the fault handling cycle and improving problem-solving efficiency.

[0045] Based on the above technical solution, as an embodiment, the equipment data may include a data sequence. The large model queries an experience database based on the input equipment data and query text to obtain target experience and initial fault diagnosis information. This may include: the large model performing trend analysis on the data sequence to obtain trend data; querying the experience database based on the query text to obtain target experience; the target experience includes fault determination rules; and the large model analyzing the trend data and equipment data according to the fault determination rules to determine initial fault diagnosis information.

[0046] Equipment data carries a time attribute. Based on multiple pieces of equipment data with time attributes, a data sequence can be obtained, which reflects the evolution of the equipment's operating status. Large-scale models can use the LSTM (Long Short-Term Memory) algorithm to perform trend analysis on the data sequence, obtaining trend data. Trend data, generated by the large-scale model after feature extraction and pattern analysis of the equipment data sequence, reflects the trend characteristics of the equipment's operating status evolution and can represent predictions of future equipment data.

[0047] Large-scale models can combine information such as query text and equipment type to accurately search experience databases and filter out highly relevant target experiences. Target experiences can include fault judgment rules, which are standardized judgment logic extracted from massive amounts of historical fault data. For example, fault judgment rules can include "vibration peak exceeding X and lasting for Y seconds is judged as bearing abnormality" and "temperature trend slope greater than Z and accompanied by current fluctuations is judged as circuit fault," etc.

[0048] The large model can compare and quantify the abnormal features in the trend data, the original equipment data, and the judgment thresholds and correlation conditions in the fault judgment rules one by one according to the fault judgment rules. By verifying whether the trend data and equipment data meet the fault feature patterns represented by the rules, it can clarify whether the equipment is abnormal, the type of abnormality, and the initial scope of impact, and finally generate initial fault diagnosis information, laying the foundation for subsequent in-depth analysis and accurate diagnosis.

[0049] The technical solution adopted in this application embodiment can determine the trajectory of equipment state evolution through trend analysis, thereby discovering faults in advance and identifying gradual faults; relying on the fault judgment rules of the experience database, it provides a standardized basis for diagnosis, reduces subjective misjudgment, and improves the accuracy and consistency of initial fault diagnosis; the large model automatically completes data processing, experience matching and rule verification, which greatly shortens the initial diagnosis cycle, reduces the cost of manual analysis, and improves fault response efficiency.

[0050] Based on the above technical solution, as an embodiment, the first interactive platform uploading the question text may include: the first interactive platform uploading the initial text to the large model, and uploading the complete question text according to the information supplementation prompt when the large model returns the information supplementation prompt; the large model is also used to perform information recognition on the initial text when it receives the initial text uploaded by the first interactive platform, and output information supplementation prompt to the first interactive platform when it recognizes that the information of the initial text is incomplete.

[0051] The first interactive platform can first upload the initial text input by the user to the large model, and then enter the information verification and feedback process. After receiving the initial text, the large model can activate the intelligent information recognition mechanism to determine the completeness of the information by analyzing whether the key elements in the text (equipment type, description of the fault phenomenon, occurrence scenario, core request, etc.) are complete. If the large model identifies that the initial text has missing key information (such as not specifying the equipment model, not specifying the timing of the fault, or vague description of the fault phenomenon), it can automatically generate targeted information supplementary prompts and feed them back to the first interactive platform, with the prompts accurately pointing to the missing items (such as "Please supplement the equipment model and the operating stage in which the abnormal noise occurred" or "Please specify the specific value and duration of the high temperature"). After receiving the information supplementary prompts, the first interactive platform can supplement the corresponding information or improve the content by combining the basic equipment information in the system database, and finally form a complete question text, which is then re-uploaded to the large model.

[0052] If the large model identifies that the initial text does not lack key information, it can directly identify the initial text as the question text and perform equipment fault diagnosis based on the question text.

[0053] By adopting the technical solution of this application embodiment, the closed-loop interaction of uploading, recognizing, completing, and re-uploading can effectively avoid diagnostic biases caused by incomplete initial information, ensure that the input obtained by the large model has complete elements such as device identification, fault details, and scene background, significantly reduce invalid diagnostic attempts, reduce the redundant costs of subsequent data matching and analysis, lay a solid information foundation for subsequent accurate diagnosis and efficient retrieval of experience databases by combining device data, and improve the coherence and reliability of the entire fault diagnosis process.

[0054] Based on the above technical solution, as an embodiment, the second interactive platform can also be used to output feedback information on the final diagnostic results; the large model can also be used to perform experience summary processing on the equipment data and the final diagnostic results based on the large model when the feedback information represents the agreement, to obtain fault experience, and to add the fault experience to the experience database.

[0055] After obtaining the final diagnostic results, the large model can return these results to the second interactive platform. This second platform allows users (such as engineers) to provide targeted feedback. Feedback can take the form of clear approval indicators, confirmation instructions based on practical verification, etc.

[0056] The large model receives feedback information uploaded from the second interactive platform and can determine whether the feedback represents agreement through semantic recognition (such as user annotations like "accurate diagnosis" or "problem solved"). When the large model recognizes user agreement, it can automatically trigger an experience summarization process.

[0057] The large model can trace back the entire diagnostic process, including the original equipment data, data sequence trend characteristics, question text, diagnostic logic derivation process, follow-up question text, and final diagnostic results. By extracting key elements such as core fault characteristics, matching conditions, judgment rules, and solutions, it can be structured and standardized to form reusable fault experience, which can be added to the experience database.

[0058] By adopting the technical solution of this application embodiment, fault experience is added to the experience database through experience summary, which can enrich the fault cases and judgment rules of similar equipment in the experience database, and provide more accurate and comprehensive experience support for the fault diagnosis of similar equipment in the future, thereby continuously improving the accuracy and efficiency of large model diagnosis.

[0059] Based on the above technical solution, as an example, the large model can be iteratively trained periodically. Historical data from different devices (motors, gearboxes, hydraulic systems, etc.) can be collected, including "normal operation data" and "fault data," labeled with fault type, location, and cause, such as "motor bearing wear - spindle - insufficient lubrication." Based on this data, supervised training or deep learning can be performed on the large model to improve its accuracy.

[0060] Based on the above technical solution, as an embodiment, the question text may include: a description of the equipment data, analysis requirements for the equipment data, and a description of the fault situation; the initial fault diagnosis information may include: abnormal data in the equipment data; the target experience may include: normal data corresponding to the abnormal data in the equipment data; the fault diagnosis result may include: fault analysis process text, fault cause, and / or fault handling suggestions; the follow-up question text may include: expert experience text and fault follow-up question text.

[0061] Based on the above technical solutions, as an example, the first and / or second interactive platforms may include functions such as result display, maintenance reminders, and feedback recording, forming a full-process support system from diagnostic output to maintenance execution and experience accumulation. The result display stage can present core diagnostic information in a multi-dimensional, visualized format, clearly displaying accurate diagnostic results (initial fault diagnosis information, fault diagnosis results, and final diagnosis results), covering the probability of fault occurrence, specific fault location, root cause analysis, and appropriate maintenance steps and spare parts recommendations, providing users with decision-making support. It can also simultaneously generate data trend charts, such as the change curves of key parameters like equipment temperature and vibration over the past hour, intuitively presenting the fault evolution trajectory. Simultaneously, it retains complete dialogue records, facilitating the review of key interactive information during the diagnostic process. The maintenance reminder function enables efficient maintenance outreach through a mobile app. The system can automatically push reminder messages containing fault type (e.g., "motor bearing wear"), urgency level, and maintenance time limit (e.g., "repair recommended within 12 hours"). Users can set maintenance countdowns based on their needs, monitor maintenance progress in real time, and avoid fault escalation due to delays. The feedback recording function allows users to mark the "fault resolved / unresolved" status on the platform after completing the repair, and fill in detailed repair details, including the model of the replacement spare parts (such as "SKF6205 bearing"), repair process and fault handling effect, etc. This data can be synchronously fed back to the cloud, providing real-world operational data for large model iterative training, and continuously improving the accuracy and adaptability of subsequent diagnoses.

[0062] Based on the above technical solution, as an example, the equipment data includes four days of operating data for the heating unit: data for day 1 (November 14, 2024), data for day 2 (November 15, 2024), data for day 3 (November 16, 2024), and data for day 4 (November 17, 2024). The question text includes: You are a heating unit equipment data analysis expert, capable of analyzing the operating data of the heating unit. The attachment contains equipment data for a heating unit, specifically time-series data generated during the operation of a single heating unit from 20:30 to 21:30 on the first, second, third, and fourth days. The first line of the file contains the attribute names of the heating unit during operation, including (time, set temperature, inlet water temperature, outlet water temperature, outdoor ambient temperature, defrosting status, exhaust superheat, compressor return (suction) temperature, outdoor coil temperature, four-way valve status, exhaust temperature). The second line onwards contains attribute values. It is known that the data from the fourth day's operation shows an anomaly, but the location of the anomaly is unknown. Please use an anomaly detection algorithm based on the 3σ criterion to compare the differences between the data from these four days, calculating only the columns (set temperature, inlet water temperature, outlet water temperature, outdoor ambient temperature, defrosting status, exhaust superheat, compressor return (suction) temperature, outdoor coil temperature, four-way valve status, exhaust temperature). Identify the abnormal data from the fourth day and the corresponding time period, and describe in words what equipment malfunction occurred during that time period.

[0063] The large model answers as follows: I. From 21:09:04 on November 17, 2024 to 21:10:17 on November 17, 2024; Abnormal attribute name: Outdoor ambient temperature / °C; The period of abnormality was from 21:09:04 on November 17, 2024 to 21:10:17 on November 17, 2024. Abnormal situation description: The outdoor ambient temperature rose abnormally from 6.8℃ to 8.6℃, which may indicate a sensor malfunction or an abnormality in the ambient temperature control system.

[0064] Anomaly Analysis: a. The average outdoor temperature during the abnormal period was 8.1°C, while the average temperature during the normal period was -1.82°C; b. During the abnormal period, the compressor return (intake) temperature rose significantly to 35.33°C, while the normal temperature is 6.87°C; c. The external coil temperature was -0.92°C during the abnormal period and -5.23°C under normal conditions; d. The exhaust superheat was 63.14°C during the abnormal period, and 65.32°C under normal conditions; e. The status of the four-way valve and the exhaust temperature change slightly during abnormal periods compared to normal periods, but the changes are not significant.

[0065] II. From 21:06:08 on November 17, 2024 to 21:07:21 on November 17, 2024; Abnormal attribute name: Compressor return (intake) temperature / ℃; The period of abnormality was from 21:06:08 on November 17, 2024 to 21:07:21 on November 17, 2024. Abnormal situation description: The compressor return (suction) temperature fluctuates abnormally, rising from 48.0℃ to 53.7℃ and then dropping to 48.8℃, indicating that the compressor is not operating stably, which may be due to compressor failure or refrigerant circulation problems.

[0066] Anomaly Analysis: a. The compressor return (intake) temperature was 51.59°C during the abnormal period, which is much higher than the normal 6.87°C; b. The outdoor ambient temperature during the abnormal period was -1.53°C, which is close to the normal range of -1.82°C; c. The external coil temperature was 4.4°C during the abnormal period, and -5.23°C under normal conditions; d. If both the defrosting status and the four-way valve status are 1.0 during the abnormal time period, it indicates that the equipment may have entered defrosting mode or the four-way valve status has changed.

[0067] III. From 21:07:36 on November 17, 2024 to 21:09:20 on November 17, 2024; Abnormal attribute name: External coil temperature / °C; The period of abnormality was from 21:07:36 on November 17, 2024 to 21:09:20 on November 17, 2024. Description of the abnormal situation: The temperature of the external coil changes abnormally, rising from 8.0℃ to 12.9℃ and then suddenly dropping to 7.9℃, indicating a problem with the heat exchange system. It may be due to frost on the coil or uneven flow leading to a decrease in heat exchange efficiency.

[0068] Anomaly Analysis: a. The external coil temperature was 10.43°C during the abnormal period, which is much higher than the normal temperature of -5.23°C; b. The outdoor ambient temperature was 4.06°C during the abnormal period, which is higher than the normal range of -1.82°C; c. The compressor return (intake) temperature was 40.14°C during the abnormal period, which is higher than the normal 6.87°C; d. A defrost status value of 0.95 during an abnormal time period indicates that the device may frequently enter defrost mode.

[0069] Based on the above equipment data, the manual analysis concluded that the equipment experienced "four-way valve leakage" between 21:08:28 and 21:09:29 on November 17, 2024. The abnormality analysis was based on "slow defrosting and high return gas temperature".

[0070] It can be seen that the time range of the large model's response and the human analysis conclusion are close (within 1 minute difference), and the analytical basis is similar (both mentioned abnormal return gas temperature). However, the large model's response is inaccurate because it does not understand the true business meaning of "four-way valve leakage".

[0071] Engineers can input the following follow-up question on the second interactive platform: When the compressor return air (suction air) temperature is consistently 15 degrees higher than the inlet water temperature, is this a phenomenon of excessively high suction air temperature? In the three abnormal time periods identified above, is there any phenomenon of excessively high suction air temperature?

[0072] In response to this follow-up question, the large model can provide the following answer: In the data from the fourth day, we observed a phenomenon of "excessively high suction temperature," specifically occurring between 21:06 and 21:07. During this period, the "compressor return (suction) temperature / ℃" consistently exceeded the "inlet water temperature / ℃" by 15 degrees Celsius. This indicates an increase in the compressor's workload, possibly due to changes in the heating system's operating mode or external environmental factors.

[0073] The second interactive platform allows for multiple follow-up questions, with engineers able to input multiple question texts. For example, an engineer could input the following question: "If the 'compressor return (suction) temperature' is significantly higher than the 'inlet water temperature' when defrosting begins, what could be the anomaly? Could it be related to the four-way valve?" The large model can respond to this follow-up question as follows: The four-way valve is a key component in a heating system, responsible for controlling the flow of refrigerant in the system to switch between heating and cooling modes. If the four-way valve malfunctions, it may affect the normal operation of the heating system, especially during defrosting. Possible four-way valve malfunctions include: 1. Four-way valve jamming or damage: If the four-way valve cannot switch normally, the defrosting mode may not start or end correctly, affecting defrosting effectiveness and system thermal efficiency. 2. Control signal problems: The operation of the four-way valve may be affected by the control signal. If the control signal is unstable or incorrect, the four-way valve may not respond correctly, leading to abnormal system operation. 3. Leakage: If the four-way valve leaks, it may affect the system's pressure balance, leading to abnormal compressor return gas temperature. Since the four-way valve's state is activated during the abnormal period (average value 1.0), this may indicate that the four-way valve is working normally during defrosting. However, if the four-way valve's operation is uncoordinated or there are problems with its interaction with other system components, it may still lead to the observed anomaly.

[0074] By analyzing the large model's response to the follow-up questions, it can be determined that the large model identified a possible fault in the four-way valve, which matches the "four-way valve leakage" identified by manual investigation.

[0075] One embodiment of this application provides a fault diagnosis method, such as... Figure 4 The fault diagnosis method shown can be applied to a fault diagnosis system, which may include: a first interactive platform, a second interactive platform, and a large model. The method includes at least steps S410 to S450, detailed below: In step S410, device data and the question text uploaded by the first interactive platform are obtained; In step S420, the device data and the question text are input into the large model to obtain the target experience and initial fault diagnosis information obtained by the large model from the experience database based on the device data and the question text, and the target experience and initial fault diagnosis information are returned to the first interactive platform. In step S430, the target experience, the equipment data, and the question text are dynamically analyzed based on the large model to obtain the fault diagnosis result, and the fault diagnosis result is returned to the first interactive platform. In step S440, the follow-up text of the fault diagnosis result input by the second interactive platform is obtained; In step S450, the follow-up text is intelligently analyzed based on the large model to obtain the final diagnosis result, and the final diagnosis result is returned to the first interactive platform.

[0076] The primary interaction platform can be used to upload query texts according to a preset cycle, or it can be used in response to fault diagnosis requests triggered by relevant users. For example, if the device is an air conditioner, and the primary interaction platform is used by after-sales personnel, when an air conditioner user discovers an abnormality, they can proactively consult after-sales personnel about whether the air conditioner is faulty. The primary interaction platform can obtain the query texts entered by after-sales personnel via voice or text. The query texts may include, but are not limited to: descriptive text of device data, analysis requirements for device data, and / or description requirements of fault conditions.

[0077] The first interactive platform can also select or actively upload device data associated with the query text from the database. The fault diagnosis system can input the device data and the query text sent by the first interactive platform into the large model. The large model can perform a preliminary analysis of the problem based on the query text and device data, and then query the experience database for target experience associated with the device data and query text. Target experience may include device data of the same type / model as the device data, or fault judgment rules summarized from historical fault data of multiple devices of the same type / model. The large model can analyze the device data and query text based on the queried target experience to determine whether there is an anomaly in the device, thereby obtaining initial fault diagnosis information. The analysis of device data and query text based on target experience by the large model has less computational load than the dynamic analysis of target experience, device data, and query text by the large model, thus obtaining target experience and initial fault diagnosis information more efficiently. The large model returns the target experience and initial fault diagnosis information to the first interactive platform, which can display the target experience and initial fault diagnosis information so that users of the first interactive platform can gain a preliminary understanding and location of the fault after seeing the target experience and initial fault diagnosis information.

[0078] After obtaining target experience and initial fault diagnosis information, the large-scale model can continue to perform more intelligent dynamic analysis based on the acquired data to obtain fault diagnosis results. Specifically, the large-scale model can dynamically analyze target experience, equipment data, and query text together to obtain fault diagnosis results. During this dynamic analysis, the large-scale model performs more intelligent calculations, with a computational workload exceeding that required for obtaining the initial fault diagnosis information, resulting in more detailed and accurate fault diagnosis results. The large-scale model can then return the fault diagnosis results to the primary interactive platform, which can display the results so that users on the platform can understand and locate the fault in detail.

[0079] In one embodiment, the large model analyzes equipment data and query text based on target experience. This is achieved by comparing whether the equipment data matches the fault data associated with the query text as represented by the target experience, thereby obtaining initial fault diagnosis information. This initial fault diagnosis information is obtained through a simple matching calculation between the equipment data and the target experience. The fault diagnosis results obtained by the large model through dynamic analysis of target experience, equipment data, and query text represent intelligent analysis results derived from deep reasoning and exploration by the large model. Therefore, the fault diagnosis results are more detailed and accurate.

[0080] In one embodiment, when the large model returns target experience, initial fault diagnosis information, and fault diagnosis results to the first interactive platform, it can simultaneously return the entire session record and / or the analysis process performed internally by the large model to the first interactive platform for display. The session record is a collection of various interactive information recorded chronologically during the interaction between the first and / or second interactive platforms and the large model.

[0081] The second interactive platform can read the entire conversation log. It can also obtain follow-up questions from users (e.g., engineers) regarding the fault diagnosis results. The fault diagnosis system can input these follow-up questions into the large model. The large model can combine the follow-up questions and the conversation log to perform intelligent analysis (deep reasoning) to obtain the final diagnosis result, which is then returned to both the first and second interactive platforms. Follow-up questions can include, but are not limited to, expert experience text and / or fault-related follow-up questions. Expert experience text allows the large model's responses to follow-up questions to better reflect the actual fault scenario, avoiding generalized conclusions. Expert experience text can supplement industry-specific fault characteristics, diagnostic logic, and other implicit knowledge, helping the large model quickly locate core issues and improve the targeting of follow-up questions and the professional accuracy of responses.

[0082] By employing the technical solution of this application embodiment, the large model can query the experience database based on device data and the question text uploaded by the first interactive platform to obtain target experience. Based on this target experience, it performs experience-based diagnosis to determine whether a device malfunction has occurred, obtaining initial fault diagnosis information. The target experience and initial fault diagnosis information are then returned to the first interactive platform. The large model can further perform more intelligent dynamic analysis based on the target experience, device data, and question text to obtain fault diagnosis results, which are then returned to the first interactive platform. Thus, the first interactive platform can obtain detailed and accurate fault information based on the target experience, initial fault diagnosis information, and fault diagnosis results. The second interactive platform can follow up on the fault diagnosis results by inputting the follow-up question text into the large model. The large model can intelligently analyze the follow-up question text to obtain the final diagnosis result, which is then returned to the first interactive platform. This provides an interactive mechanism for users, allowing them to ask about fault details and repair methods. The large model responds to the user's personalized follow-up questions, providing various information to help users accurately locate and resolve device faults.

[0083] In one embodiment, obtaining the follow-up text on the fault diagnosis result input by the second interactive platform includes: Obtain the uncertainty threshold, and determine the uncertainty of the fault diagnosis result based on the large model; When the uncertainty is greater than the uncertainty threshold, the device information request text is automatically output to the second interactive platform; Obtain the follow-up text of the fault diagnosis result requested by the second interactive platform based on the device information request text input.

[0084] In one embodiment, after obtaining the fault diagnosis result, the method further includes: Store session records; The step of obtaining follow-up text on the fault diagnosis result input by the second interactive platform includes: Obtain the follow-up question text entered by the second interactive platform after reviewing the session record.

[0085] In one embodiment, the device data includes a data sequence; The process of obtaining the target experience and initial fault diagnosis information obtained by querying the experience database based on the question text and the device data of the large model includes: The large model performs trend analysis on the data sequence to obtain trend data; The large model queries the experience database based on the question text to obtain the target experience; the target experience includes fault analysis rules. The large model analyzes the trend data and the equipment data according to the fault analysis rules to determine the initial fault diagnosis information.

[0086] In one embodiment, acquiring device data includes: Determine the load type of the equipment; The data acquisition frequency is determined based on the load type of the device; Acquire the multimodal device data collected at the stated data acquisition frequency; The device data is obtained by preprocessing the multimodal device data.

[0087] In one embodiment, obtaining the question text uploaded by the first interactive platform includes: Obtain the initial text uploaded by the first interactive platform; Information recognition is performed on the initial text based on the large model; When the initial text is found to be incomplete, a supplementary information prompt is output to the first interactive platform, and the complete question text uploaded by the first interactive platform according to the supplementary information prompt is obtained.

[0088] In one embodiment, after obtaining the final diagnostic result, the method further includes: Obtain feedback information from the second interactive platform regarding the final diagnostic result; When the feedback information is an affirmative information, the device data and the final diagnostic results are processed based on the large model to obtain fault experience. The fault experience is added to the experience database.

[0089] To facilitate better implementation of the fault diagnosis method of this application, this application also provides a fault diagnosis device based on the above-described fault diagnosis method. The meanings of the terms used are the same as in the fault diagnosis method described above, and specific implementation details can be found in the descriptions of the method embodiments.

[0090] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of the fault diagnosis device provided in the embodiments of this application, wherein the fault diagnosis device includes: The question acquisition module 501 is used to acquire device data and the question text uploaded by the first interactive platform; The question input module 502 is used to input the device data and the question text into the large model, obtain the target experience and initial fault diagnosis information obtained by the large model from the experience database based on the device data and the question text, and return the target experience and initial fault diagnosis information to the first interactive platform; The question analysis module 503 is used to dynamically analyze the target experience, the equipment data and the question text based on the large model, obtain the fault diagnosis result, and return the fault diagnosis result to the first interactive platform; The follow-up question module 504 is used to obtain the follow-up question text of the fault diagnosis result input by the second interactive platform; The follow-up analysis module 505 is used to perform intelligent analysis on the follow-up text based on the large model to obtain the final diagnosis result, and return the final diagnosis result to the first interactive platform.

[0091] For specific limitations regarding the fault diagnosis equipment, please refer to the limitations on the fault diagnosis methods above, which will not be repeated here. Each module in the aforementioned fault diagnosis equipment can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0092] Figure 6 The diagram illustrates the structure of a computing device cluster. Figure 6 The computing device cluster shown includes multiple computing devices 600, and the aforementioned fault diagnosis equipment includes multiple functional modules that can be distributed and deployed across multiple computing devices within this cluster. For example... Figure 6 As shown, the computing device cluster includes multiple computing devices 600. Each computing device 600 includes a memory 610, a processor 620, a communication interface 630, and a bus 640. The memory 610, the processor 620, and the communication interface 630 communicate with each other through the bus 640.

[0093] The processor 620 can be a CPU, GPU, ASIC, or one or more integrated circuits. The processor 620 can also be an integrated circuit chip with signal processing capabilities. In implementation, some functions of the aforementioned fault diagnosis device can be accomplished through integrated logic circuits in the hardware of the processor 620 or through software instructions. The processor 620 can also be a DSP, FPGA, general-purpose processor, other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, capable of implementing or executing some of the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory 610. In each computing device 600, the processor 620 reads information from memory 610, and in conjunction with its hardware, can complete some functions of the fault diagnosis device.

[0094] The memory 610 may include ROM, RAM, static storage devices, dynamic storage devices, hard disks (e.g., SSDs, HDDs), etc. The memory 610 can store program code. The communication interface 603 in each computing device 600 is used for communication with external systems, such as interacting with other computing devices 600.

[0095] Bus 640 can be a standard bus for interconnecting peripheral components or an extended industry standard structure bus, etc. For ease of representation, Figure 6 The bus 640 within each computing device 600 is represented by a single thick line, but this does not imply that there is only one bus or one type of bus.

[0096] The aforementioned multiple computing devices 600 establish communication channels through a communication network to realize the functions of the fault diagnosis device. Any computing device can be a computing device in a cloud environment (e.g., a server), a computing device in an edge environment, or a terminal device.

[0097] Furthermore, this application also provides a computer-readable storage medium storing instructions that, when executed on one or more computing devices, cause the one or more computing devices to perform the methods executed by the various modules of the fault diagnosis device described above.

[0098] Furthermore, this application also provides a computer program product, which, when executed by one or more computing devices, allows the computing devices to perform any of the aforementioned fault handling methods. The computer program product can be a software installation package; when any of the aforementioned fault handling methods is required, the computer program product can be downloaded and executed on a computer.

[0099] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0100] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0101] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0102] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0103] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0104] The above provides a detailed description of a fault diagnosis system, method, computing device cluster, and computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A fault diagnosis system, characterized in that, include: The first interactive platform is used to upload question text; The large model is used to query the experience database based on the input device data and the question text to obtain the target experience and the initial fault diagnosis information, and return the target experience and the initial fault diagnosis information to the first interactive platform. It also performs dynamic analysis on the target experience, the device data and the question text to obtain the fault diagnosis result, and returns the fault diagnosis result to the first interactive platform. The second interactive platform is used to input the follow-up text of the fault diagnosis results into the large model; The large model is also used to perform intelligent analysis on the follow-up text to obtain a final diagnostic result, and return the final diagnostic result to the first interactive platform.

2. The system according to claim 1, characterized in that, The large model is used to determine the uncertainty of the fault diagnosis result, and when the uncertainty is greater than the uncertainty threshold, it automatically outputs a device information request to the second interactive platform. The second interactive platform is specifically used to request follow-up text of the fault diagnosis result to be input into the large model based on the device information.

3. The system according to claim 1, characterized in that, The system also includes: A database used to store session records; The second interactive platform inputs the follow-up text of the fault diagnosis results into the large model, including: The second interactive platform enters the follow-up question text after reviewing the session record.

4. The system according to claim 1, characterized in that, The device data includes a data sequence; The large model queries the experience database based on the input device data and the query text to obtain the target experience and the initial fault diagnosis information, including: The large model performs trend analysis on the data sequence to obtain trend data, and queries the experience database based on the question text to obtain the target experience; the target experience includes fault judgment rules. The large model analyzes the trend data and the equipment data according to the fault determination rules to determine the initial fault diagnosis information.

5. The system according to claim 1, characterized in that, The system also includes: The device is used to determine the load type of the device, determine the data acquisition frequency based on the load type of the device, and acquire multimodal device data according to the data acquisition frequency; The device is also used to preprocess the multimodal device data to obtain the device data, and upload the device data to the large model.

6. The system according to claim 1, characterized in that, The first interactive platform uploads the question text, including: The first interactive platform uploads the initial text to the large model, and when the large model returns supplementary information prompts, it uploads the complete question text according to the supplementary information prompts; The large model is also used to perform information recognition on the initial text when it receives the initial text uploaded by the first interactive platform, and to output the information supplement prompt to the first interactive platform when the information of the initial text is found to be incomplete.

7. The system according to claim 1, characterized in that, The second interactive platform is also used to output feedback information on the final diagnostic result; The large model is also used to perform experience summarization processing on the device data and the final diagnosis result based on the large model when the feedback information is recognized, to obtain fault experience, and to add the fault experience to the experience database.

8. A fault diagnosis method, characterized in that, The method is applied to a fault diagnosis system, which includes: a first interactive platform, a second interactive platform, and a large model; the method includes: Obtain device data and the question text uploaded by the first interactive platform; The device data and the question text are input into the large model to obtain the target experience and initial fault diagnosis information obtained by the large model from the experience database based on the device data and the question text, and the target experience and initial fault diagnosis information are returned to the first interactive platform. Based on the large model, the target experience, the equipment data and the question text are dynamically analyzed to obtain the fault diagnosis result, and the fault diagnosis result is returned to the first interactive platform; Obtain follow-up text of the fault diagnosis results input from the second interactive platform; The large model is used to perform intelligent analysis on the follow-up text to obtain the final diagnosis result, and the final diagnosis result is returned to the first interactive platform.

9. A computing device cluster, characterized in that, It includes at least one computing device, each computing device including a processor and memory; The processor is configured to execute instructions stored in the memory to cause the computing device cluster to perform the method of claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on at least one computing device, cause the at least one computing device to perform the method as described in claim 8.

Images