Industrial equipment fault diagnosis method, device, equipment, storage medium and program product

Abstract

The application discloses an industrial equipment fault diagnosis method, device, equipment, storage medium and program product, and relates to the technical field of intelligent operation and maintenance of industrial equipment. The method comprises the following steps: extracting a target alarm code from a fault triggering message of an industrial equipment; calling a Markov chain state model to process the target alarm code, so as to obtain a diagnosis context output by the Markov chain state model; the Markov chain state model comprises a state set and a state transition probability matrix between states, the state set comprises an alarm code state, a failure mode state and a failure cause state; the diagnosis context comprises candidate failure mode information and candidate failure cause information; and generating a fault diagnosis scheme according to the candidate failure mode information and the candidate failure cause information. The technical scheme of the embodiment of the application realizes the automation and intelligence of fault diagnosis, and improves the diagnosis efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for industrial equipment, and in particular to a method, apparatus, equipment, storage medium, and program product for diagnosing industrial equipment faults. Background Technology

[0002] In the context of intelligent manufacturing and industrial digital transformation, the operational status of industrial equipment directly affects production efficiency and product quality. With the continuous increase in equipment integration and complexity, failure modes are exhibiting diverse and cascading characteristics. Traditional fault diagnosis methods mainly rely on engineers' experience and preset threshold alarms, which suffer from problems such as time-consuming troubleshooting, poor diagnostic consistency, and a lack of a systematic reasoning mechanism from alarm triggering to the cause of failure. How to quickly and accurately locate the root cause of the fault and generate an effective diagnostic solution after an alarm is triggered has become a critical problem that urgently needs to be solved in the field of industrial equipment fault diagnosis. Summary of the Invention

[0003] This invention provides a method, apparatus, equipment, storage medium, and program product for diagnosing industrial equipment faults, thereby improving the efficiency and accuracy of fault diagnosis.

[0004] According to one aspect of the present invention, a method for diagnosing faults in industrial equipment is provided, comprising:

[0005] Extract the target alarm code from the fault trigger message of industrial equipment;

[0006] The target alarm code is processed by calling a Markov chain state model to obtain the diagnostic context output by the Markov chain state model; the Markov chain state model includes a set of states and a state transition probability matrix between states, the set of states includes alarm code states, failure mode states, and failure cause states; the diagnostic context includes candidate failure mode information and candidate failure cause information.

[0007] A fault diagnosis scheme is generated based on the candidate failure mode information and the candidate failure cause information.

[0008] According to another aspect of the present invention, an industrial equipment fault diagnosis device is provided, comprising:

[0009] The alarm code extraction module is used to extract target alarm codes from fault trigger messages of industrial equipment;

[0010] The diagnostic context acquisition module is used to call the Markov chain state model to process the target alarm code in order to obtain the diagnostic context output by the Markov chain state model; the Markov chain state model includes a set of states and a state transition probability matrix between states, the set of states includes alarm code states, failure mode states, and failure cause states; the diagnostic context includes candidate failure mode information and candidate failure cause information.

[0011] The diagnostic scheme generation module is used to generate a fault diagnosis scheme based on the candidate failure mode information and the candidate failure cause information.

[0012] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0013] At least one processor; and

[0014] A memory communicatively connected to the at least one processor; wherein,

[0015] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the industrial equipment fault diagnosis method according to any embodiment of the present invention.

[0016] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the industrial equipment fault diagnosis method according to any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the industrial equipment fault diagnosis method according to any embodiment of the present invention.

[0018] The technical solution of this invention extracts the target alarm code from the fault trigger message of industrial equipment, uses the Markov chain state model to propagate the fault to obtain a diagnostic context including candidate failure modes and candidate failure causes, and generates a fault diagnosis scheme based on the diagnostic context, thereby realizing the automation and intelligence of fault diagnosis and improving diagnostic efficiency and accuracy.

[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0021] Figure 1 This is a flowchart of an industrial equipment fault diagnosis method provided according to an embodiment of the present invention;

[0022] Figure 2 This is a flowchart of another industrial equipment fault diagnosis method provided according to an embodiment of the present invention;

[0023] Figure 3 This is a flowchart of another industrial equipment fault diagnosis method provided according to an embodiment of the present invention;

[0024] Figure 4 This is a schematic diagram of the structure of an industrial equipment fault diagnosis device according to an embodiment of the present invention;

[0025] Figure 5 This is a schematic diagram of the structure of an electronic device that implements the industrial equipment fault diagnosis method of this invention. Detailed Implementation

[0026] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0028] Figure 1This is a flowchart illustrating an industrial equipment fault diagnosis method provided by an embodiment of the present invention. This embodiment is applicable to scenarios where industrial equipment triggers an alarm during operation, requiring rapid location of the root cause of the fault and generation of an executable diagnostic solution. It is particularly suitable for industrial automation equipment such as transmitters, sensors, and actuators. This method can be executed by an industrial equipment fault diagnosis device, which can be implemented in hardware and / or software. This device can be configured in an electronic device that carries industrial equipment fault diagnosis functions, such as a client or server.

[0029] See Figure 1 The industrial equipment fault diagnosis method shown includes:

[0030] Step S101: Extract the target alarm code from the fault trigger message of the industrial equipment;

[0031] During the operation of industrial equipment, when the equipment monitoring system (such as a programmable logic controller, distributed control system, or embedded status monitoring unit) detects that the operating parameters exceed preset thresholds or an abnormal state occurs, it generates a fault trigger message. This message can be reported to the diagnostic server of the operation and maintenance platform via industrial Ethernet, fieldbus, or wireless transmission protocol to trigger subsequent diagnostic processes. The fault trigger message can be received and processed by a fault diagnosis device. Optionally, the fault diagnosis device can be a fault diagnosis intelligent agent, i.e., an LLM (Large Language Model) with an MCP (Model Context Protocol) client deployed; the fault diagnosis device can also be independent of the large language model; regardless of the method, it can be deployed in computing devices such as edge computing gateways, industrial computers, or servers.

[0032] Fault trigger messages can be encapsulated in a structured data format. The message body can include device context (such as device ID, product type, product model, site identifier, subsystem identifier) ​​and alarm context (such as alarm code, standardized pre-alarm text, alarm timestamp, and alarm description). Upon receiving a fault trigger message from a user / device, the fault diagnosis device parses the message content and extracts the target alarm code (such as ALM-201) as the entry parameter for subsequent fault diagnosis processes. If the alarm context includes both the original alarm text and the standardized alarm code, the standardized alarm code is used first to ensure consistency with the encoding system of the state set in the Markov chain state model used later, avoiding state matching failures due to differences in textual representation. If the alarm context only includes the original alarm text, the original alarm text can be standardized to generate a standardized alarm code.

[0033] Step S102: Call the Markov chain state model to process the target alarm code to obtain the diagnostic context output by the Markov chain state model; the Markov chain state model includes a set of states and a state transition probability matrix between states, the set of states includes alarm code states, failure mode states and failure cause states; the diagnostic context includes candidate failure mode information and candidate failure cause information.

[0034] After extracting the target alarm code, the fault diagnosis device calls a pre-built Markov chain state model (i.e., a fault propagation model), inputting the target alarm code as a diagnostic input parameter to the Markov chain state model, enabling the model to calculate the fault propagation path. The Markov chain state model characterizes the propagation relationship from the alarm code to the failure mode, and from the failure mode to the failure cause. The Markov chain state model includes a set of states and state transition probability matrices between states. The state set includes alarm code states, failure mode states, and failure cause states; the diagnostic context includes candidate failure mode information and candidate failure cause information.

[0035] In one optional implementation, each row of the state transition probability matrix corresponds to a source state, each column corresponds to a target state, each element in a row represents the probability of transitioning from the source state to the corresponding target state, and the sum of the probabilities in each row is 1.

[0036] In the state transition probability matrix, the source state is either the alarm code state or the failure mode state, and the target state is either the failure mode state or the failure cause state, corresponding to two levels of diagnostic transitions: from alarm to failure mode and from failure mode to failure cause, respectively. Each element in the state transition probability matrix represents the conditional probability of transitioning from the source state to the target state, denoted as P(target state|source state). When an element is 1, it indicates that the source state is a termination state (i.e., an absorption state), and there is no subsequent transition to other states.

[0037] Alarm code states correspond to various device alarm events, failure mode states correspond to the manifestations of device failure, and failure cause states correspond to the root causes of failure. By calling a Markov chain state model to calculate the fault propagation path, candidate failure modes and failure causes can be ranked based on the quantified probability values ​​in the state transition probability matrix among multiple possible fault propagation branches. This transforms the fault root cause localization process, which relies on expert experience, into repeatable and computable quantitative reasoning, thereby improving diagnostic efficiency and the consistency of results.

[0038] Step S103: Generate a fault diagnosis scheme based on the candidate failure mode information and candidate failure cause information.

[0039] After obtaining the diagnostic context returned by the Markov chain state model, the fault diagnosis device generates a fault diagnosis plan for field users based on candidate failure mode information and candidate failure cause information. Optionally, the fault diagnosis device inputs the candidate failure mode information and candidate failure cause information output by the Markov chain state model into a large language model. The large language model then generates a fault diagnosis plan, including a fault summary, candidate failure cause ranking, suggested troubleshooting steps, and suggested remedial measures, based on the causal relationship between candidate failure modes and failure causes and the probability ranking of each candidate path, and returns it to the user / equipment. By combining the quantitative diagnostic results output by the Markov chain state model with the semantic understanding and generation capabilities of the large language model, candidate failure causes and path probabilities can be transformed into clearly structured and actionable troubleshooting steps and remedial suggestions, making the diagnostic results easier for field engineers to understand and implement, thereby shortening fault repair time.

[0040] It should be noted that this step is not limited to the implementation method that relies on a large language model. The fault diagnosis device can also directly generate a fault diagnosis plan based on the candidate failure mode information and candidate failure cause information, according to preset rules (such as path probability sorting or table lookup matching).

[0041] The technical solution of this invention extracts the target alarm code from the fault trigger message of industrial equipment, uses the Markov chain state model to propagate the fault to obtain a diagnostic context including candidate failure modes and candidate failure causes, and generates a fault diagnosis scheme based on the diagnostic context, thereby realizing the automation and intelligence of fault diagnosis and improving diagnostic efficiency and accuracy.

[0042] In one optional embodiment, the method further includes: receiving an execution feedback message of the fault diagnosis scheme; the execution feedback message includes an execution result, a set of valid actions, a set of invalid actions, and a confirmed root cause; based on the execution result, positively associating the valid actions with their corresponding failure causes, negatively associating the invalid actions with their corresponding failure causes, and marking the confirmed root cause as confirmed, for use in correcting the fault diagnosis scheme.

[0043] Specifically, after executing the fault diagnosis plan, the field engineer or automated execution system encapsulates the execution result into a plan execution feedback message and returns it to the fault diagnosis device. This feedback message includes a plan identifier (the ID of the fed-back plan), execution result (valid, partially valid, invalid), a set of valid actions (verified valid troubleshooting or repair actions), a set of invalid actions (verified invalid troubleshooting or repair actions), and a confirmed root cause (the final confirmed cause of failure on-site). After parsing the feedback message, the fault diagnosis device records positive associations between valid actions and their corresponding failure causes, negative associations between invalid actions and their corresponding failure causes, and marks the confirmed root cause as confirmed. This correction result is used for the dynamic optimization of subsequent fault diagnosis plans, thereby achieving a closed-loop improvement in fault diagnosis accuracy.

[0044] Figure 2 This is a flowchart of another industrial equipment fault diagnosis method provided by an embodiment of the present invention. Based on the above embodiments, this embodiment of the present invention specifies "calling a Markov chain state model to process the target alarm code to obtain the diagnostic context output by the Markov chain state model" as "taking the alarm code state corresponding to the target alarm code as the diagnostic entry state of the Markov chain state model; obtaining candidate failure mode information based on the first transition probability from the diagnostic entry state to each failure mode state; obtaining candidate failure cause information based on the second transition probability from the failure mode state corresponding to the candidate failure mode information to each failure cause state; generating a diagnostic context including the candidate failure mode information and the candidate failure cause information," further improving the accuracy and interpretability of the diagnostic context. It should be noted that parts not detailed in this embodiment of the present invention can be found in the descriptions of other embodiments. The industrial equipment fault diagnosis method includes:

[0045] Step S201: Extract the target alarm code from the fault trigger message of the industrial equipment;

[0046] Step S202: Take the alarm code state corresponding to the target alarm code as the diagnostic entry state of the Markov chain state model; the Markov chain state model includes a set of states and a state transition probability matrix between states, the set of states includes alarm code state, failure mode state and failure cause state;

[0047] In the Markov chain state model, the state propagation direction is from alarm code state to failure mode state, and from failure mode state to failure cause state. That is, alarm code state can be transferred to failure mode state, and failure mode state can be transferred to failure cause state. There is no reverse transfer from failure cause state to failure mode state or from failure mode state to alarm code state, thus ensuring that the fault propagation direction is consistent with the actual fault causal chain direction.

[0048] Upon receiving a target alarm code, the fault diagnosis device performs precise coding matching within the state set to locate the alarm code state corresponding to that alarm code, and uses this state as the diagnostic entry point for the current fault propagation. By accurately mapping the target alarm code to the corresponding alarm code state in the state set, the original alarm code can be transformed into a computable state node in a Markov chain state model. This provides an accurate starting point for subsequent quantitative fault propagation calculations based on the state transition probability matrix, avoiding diagnostic errors caused by state location deviations. If the target alarm code does not exist in the state set, an error message is returned and propagation is terminated.

[0049] Step S203: Obtain candidate failure mode information based on the first transition probability from the diagnostic entry state to each failure mode state;

[0050] After determining the diagnostic entry state, the Markov chain state model reads the transition probability distribution of the row containing the alarm code state from the state transition probability matrix, obtaining the first transition probability from the diagnostic entry state to each failure mode state. The first transition probability reflects the relative probability of each failure mode occurring given the current alarm code.

[0051] In one possible implementation, this probability value can be determined based on the normalized quantification results of the probability of occurrence and detectability of failure modes in the fault causation dataset. The fault causation dataset refers to a structured data set extracted from Design Failure Mode and Effects Analysis (D-FMEA) documents, which includes alarm codes, failure modes, and diagnostic path relationships and their associated attributes among failure causes.

[0052] The Markov chain state model traverses all reachable failure mode states, filters out failure modes with probabilities greater than a preset minimum path probability threshold, and sorts them from highest to lowest according to their first transition probability to form candidate failure mode information. By filtering and sorting failure modes based on their first transition probability, high-probability failure modes can be prioritized from multiple possible failure modes, effectively narrowing down the scope of fault investigation and thus improving diagnostic efficiency.

[0053] Step S204: Based on the second transition probability from the failure mode state to each failure cause state corresponding to the candidate failure mode information, obtain candidate failure cause information; generate a diagnostic context including the candidate failure mode information and the candidate failure cause information; the diagnostic context includes the candidate failure mode information and the candidate failure cause information.

[0054] For each failure mode state in the candidate failure mode information, the Markov chain state model reads the transition probability distribution of the row containing that failure mode state from the state transition probability matrix, obtaining the second transition probability from that failure mode state to each failure cause state. The second transition probability reflects the relative likelihood of each failure cause occurring given the current failure mode; this probability value can be determined based on the normalized quantification result of the central severity of the fault causal dataset. The Markov chain state model continues to traverse all reachable failure cause states (including absorbing states, i.e., terminating states with a probability of 1 when there are no successor states), filtering and sorting them from high to low according to the second transition probability to form candidate failure cause information. Furthermore, the Markov chain state model generates a diagnostic context including candidate failure mode information and candidate failure cause information, returning it to the fault diagnosis device. Through the hierarchical propagation calculation of two levels of probability, the root cause localization of the fault is transformed from subjective judgment relying on expert experience to objective ranking based on quantified probabilities, making the diagnostic results repeatable and interpretable. This allows for orderly troubleshooting according to probability priority, avoiding maintenance delays caused by improper troubleshooting order.

[0055] Step S205: Generate a fault diagnosis scheme based on the candidate failure mode information and candidate failure cause information.

[0056] In one alternative implementation, the first transition probability and the second transition probability are determined in the following manner:

[0057] A. Extract diagnostic paths, including alarm codes, failure modes, and failure causes, from the fault cause dataset of the industrial equipment;

[0058] The system architecture and functional specifications documents for the input industrial equipment products are parsed. Following the standard workflow of industrial equipment product development, key information is extracted, including some or all of the following: product functions, alarm codes, failure modes, failure consequences, severity (S), failure causes, preventive measures, detection measures, probability of failure mode occurrence (O), and detectability of failure modes (D). Based on this, a diagnostic path of "alarm code → failure mode → failure cause" is extracted from each record in the fault causality dataset, and a unique identifier (diagnostic path ID) is established for each diagnostic path. This diagnostic path completely describes a causal chain from alarm triggering to failure mode occurrence and then to tracing the failure cause.

[0059] B. Construct a failure propagation chain based on the extracted diagnostic paths;

[0060] After extracting the diagnostic paths, each diagnostic path and its associated additional attributes are transformed into a structured failure propagation chain including nodes, edges, and attributes. Nodes include alarm code nodes, failure mode nodes, and failure cause nodes; edges represent causal relationships between nodes; attributes include original D-FMEA parameters such as product function, failure consequences, severity, preventative measures, detection measures, failure mode occurrence probability, and failure mode detectability. During construction, one-to-many relationships (one alarm code corresponding to multiple failure modes, or one failure mode corresponding to multiple failure causes) and many-to-one relationships (multiple failure modes corresponding to the same failure cause) are preserved without pre-pruning to retain all possible branches of fault propagation. Simultaneously, entries from different records in the diagnostic path set that have the same diagnostic path are merged to avoid data redundancy.

[0061] C. Based on the failure mode occurrence probability and failure mode detectability of the same alarm code to each failure mode in the failure propagation chain, calculate the first transition probability of the diagnostic path from the alarm code to each failure mode.

[0062] In the constructed structured failure propagation chain, the same alarm code node may be connected to different failure mode nodes through multiple edges. Each edge is associated with two attribute values: the probability of occurrence of the failure mode and the detectability of the failure mode. The first transition probability of the alarm code to each failure mode is calculated based on the product of the probability of occurrence of the failure mode and the detectability of the failure mode, which comprehensively reflects the probability of occurrence of the failure mode and the difficulty of detection.

[0063] In one optional implementation, the step of calculating the first transition probability of the diagnostic path from the alarm code to each failure mode based on the failure mode occurrence probability and failure mode detectability of the same alarm code to each failure mode in the failure propagation chain includes: normalizing the product of the failure mode occurrence probability and failure mode detectability of the same alarm code to each failure mode in the failure propagation chain, and using the normalization result as the first transition probability of the diagnostic path from the alarm code to each failure mode.

[0064] Based on this, the product values ​​of the same alarm code to each failure mode are normalized, and the normalized result is used as the first transition probability of the alarm code to each failure mode. Taking the O×D value of alarm code ALM-201 to failure mode load cell drift as 48 and the O×D value to failure mode excitation voltage abnormality as 112 as an example, after normalization, the first transition probability of alarm code ALM-201 to load cell drift is 48 / (48+112)=0.3, and the first transition probability to failure mode excitation voltage abnormality is 112 / (48+112)=0.7. This normalization operation ensures that the sum of all first transition probabilities originating from the same alarm code node is 1, which meets the row normalization requirement of the Markov chain state transition probability matrix.

[0065] D. Based on the severity of the same failure mode to each failure cause in the failure propagation chain, calculate the second transition probability of the diagnostic path from the failure mode to each failure cause.

[0066] In a structured failure propagation chain, the same failure mode node may be connected to different failure cause nodes through multiple edges, each edge being associated with a severity attribute value. Severity reflects the severity of the consequences caused by the failure cause once it occurs, and is an important basis for prioritizing fault diagnosis. Specifically, the severity attribute values ​​of the same failure mode to each failure cause in the failure propagation chain are normalized, and the normalization result is used as the second transition probability of the diagnostic path from the failure mode to each failure cause. This normalization operation ensures that the sum of all second transition probabilities originating from the same failure mode node is 1, which, together with the first transition probability, constitutes the complete Markov chain state transition probability matrix.

[0067] The technical solution of this invention extracts the target alarm code from the fault trigger message of industrial equipment, uses the alarm code state corresponding to the target alarm code as the diagnostic entry state of the Markov chain state model, obtains candidate failure mode information based on the first transition probability, obtains candidate failure cause information based on the second transition probability, and then generates a diagnostic context including both information and a fault diagnosis scheme accordingly. This solution transforms the professional analysis results in the D-FMEA field into a computable diagnostic reasoning framework through the conversion of the structured failure propagation chain and the Markov chain state model, achieving accurate quantification and ranking from alarm code to failure cause.

[0068] Figure 3This is a flowchart of another industrial equipment fault diagnosis method provided by an embodiment of the present invention. Based on the above embodiments, this embodiment of the present invention specifies "generating a fault diagnosis scheme based on the candidate failure mode information and the candidate failure cause information" as "initiating a knowledge retrieval request to a product knowledge database based on the candidate failure mode information and the candidate failure cause information; receiving a knowledge retrieval response message returned by the product knowledge database; the knowledge retrieval response message includes knowledge evidence matching the candidate failure mode information and the candidate failure cause information; fusing the candidate failure mode information, the candidate failure cause information, and the knowledge evidence to generate the fault diagnosis scheme," thereby improving the accuracy and interpretability of the diagnosis scheme. It should be noted that parts not detailed in this embodiment of the present invention can be referred to in other embodiments. The industrial equipment fault diagnosis method includes:

[0069] Step S301: Extract the target alarm code from the fault trigger message of the industrial equipment;

[0070] Step S302: Invoke the Markov chain state model to process the target alarm code to obtain the diagnostic context output by the Markov chain state model; the Markov chain state model includes a set of states and a state transition probability matrix between states, the set of states includes alarm code states, failure mode states and failure cause states; the diagnostic context includes candidate failure mode information and candidate failure cause information.

[0071] Step S303: Based on the candidate failure mode information and the candidate failure cause information, initiate a knowledge retrieval request to the product knowledge database;

[0072] After obtaining the diagnostic context returned by the Markov chain state model, the fault diagnosis device constructs retrieval query conditions based on candidate failure mode information and candidate failure cause information, and initiates a retrieval request to the product knowledge database. This retrieval request is used to recall knowledge evidence related to the current candidate failure path, candidate failure mode, and candidate failure cause from the product knowledge database. By initiating a retrieval request to the product knowledge database, the structured diagnostic results output by the Markov state model can be correlated with the knowledge in the product knowledge base, providing factual basis for the generation of subsequent fault diagnosis solutions.

[0073] In one optional implementation, the product knowledge database is constructed based on multi-source knowledge, which includes static knowledge, dynamic knowledge, and operation and maintenance knowledge; the static knowledge includes product design documents and functional specifications, the dynamic knowledge includes historical operating data, operating status monitoring data, and alarm logs, and the operation and maintenance knowledge includes a fault handling experience base and a glossary.

[0074] Multi-source knowledge is constructed into a product knowledge database that can be used for vector retrieval through a standard process under the Retrieval-Augmented Generation (RAG) technology framework, which includes knowledge extraction, text segmentation, vectorization, and data storage.

[0075] Step S304: Receive a knowledge retrieval response message returned by the product knowledge database; the knowledge retrieval response message includes knowledge evidence that matches the candidate failure mode information and the candidate failure cause information;

[0076] Upon receiving a search request, the product knowledge database performs a similarity search, retrieving knowledge evidence related to the current fault from product functional documents, fault manuals, detection measures, and preventative measures. This knowledge evidence is then returned to the fault diagnosis device via a knowledge retrieval response message. The response message includes knowledge evidence matching candidate failure mode information and candidate failure cause information. This knowledge evidence may include fields such as evidence ID, source type, title, content fragment, relevance, association path ID, and associated failure cause, providing a traceable basis for subsequent fusion. By receiving knowledge data returned from the product knowledge base, the diagnostic results not only rely on probability ranking but also have the support of domain knowledge, enhancing the credibility of the diagnostic results.

[0077] Step S305: The candidate failure mode information, the candidate failure cause information and the knowledge evidence are fused to generate the fault diagnosis scheme.

[0078] The fault diagnosis device integrates the candidate failure mode information and candidate failure cause information output by the Markov chain state model with the knowledge evidence returned by the product knowledge database to generate the final fault diagnosis solution.

[0079] In one optional implementation, fusing the candidate failure mode information, the candidate failure cause information, and the knowledge evidence to generate the fault diagnosis scheme includes: matching each candidate failure cause with the knowledge evidence to determine the relevance of each candidate failure cause; determining the priority of each candidate failure cause based on the path probability of the candidate failure path in which each candidate failure cause is located and the relevance; generating fault diagnosis prompt words based on the priority of the candidate failure causes, the candidate failure mode information, the candidate failure cause information, and the knowledge evidence; and inputting the fault diagnosis prompt words into a large language model to generate the fault diagnosis scheme.

[0080] Here, a candidate fault path refers to the diagnostic link that starts from the target alarm code state, transitions through the failure mode state to the failure cause state, and the path probability is used to characterize the likelihood of the candidate fault path occurring. The path probability can be the result of multiplying the probabilities of each level of transition along the candidate fault path in the Markov chain state model, and is used to quantify the relative probability of the candidate fault path occurring.

[0081] Specifically, for each candidate failure cause, its name, description, and content fragments in the knowledge evidence are semantically matched to calculate the relevance. A higher relevance indicates a better match between the failure cause and the description in the existing knowledge base. The path probability and relevance are then fused according to preset weights to obtain the priority of each candidate failure cause, which is then sorted from highest to lowest priority. The sorted candidate failure mode information, candidate failure cause information, and matched knowledge evidence are integrated into a structured fault diagnosis prompt. This prompt clearly indicates the solution structure that the large language model needs to output, including, for example, a fault summary, candidate failure cause ranking, suggested troubleshooting steps, suggested remedial measures, risk warnings, and references. The large language model generates the final fault diagnosis solution based on this prompt and returns it to the user / device in the format of a fault solution message. By fusing path probability and knowledge evidence relevance to obtain the priority of candidate failure causes, the generated diagnosis solution balances the objectivity of statistical probability with the professionalism of domain knowledge, effectively improving the accuracy of the fault diagnosis solution.

[0082] For example, regarding the fault with alarm code ALM-201, when the candidate path "ALM-201 → load cell drift → sensor aging" has a high probability and is supported by knowledge evidence such as "comparing the zero-point value with the reference value" and "replacing the aging sensor," the fault diagnosis device can determine sensor aging as the primary suspected cause of failure. Simultaneously, the path with a lower probability, "ALM-201 → load cell drift → analog-to-digital converter sampling noise," is retained as a secondary candidate path. The generated fault diagnosis solution can output the following suggestions: check whether the load cell drift is caused by sensor aging, and perform the operation of comparing the zero-point value with the reference value; if sensor aging is confirmed, replace the aging sensor; if no aging problem is found, further investigate the analog-to-digital converter sampling noise according to the secondary path.

[0083] The technical solution of this invention extracts the target alarm code from the fault trigger message of industrial equipment, calls the Markov chain state model to propagate the fault to obtain the diagnostic context, and initiates a knowledge retrieval request to the product knowledge database based on the candidate failure mode information and candidate failure cause information. After receiving the returned knowledge evidence, it fuses the evidence to generate a fault diagnosis solution, thereby realizing the structured generation and knowledge enhancement of the fault diagnosis solution and improving the accuracy and interpretability of the diagnosis solution.

[0084] Figure 4 This is a schematic diagram of an industrial equipment fault diagnosis device provided in an embodiment of the present invention. This embodiment of the present invention is applicable to scenarios where industrial equipment triggers an alarm during operation and requires rapid location of the root cause of the fault and generation of an executable diagnostic solution. It is particularly suitable for industrial automation equipment such as transmitters, sensors, and actuators. This device can execute industrial equipment fault diagnosis methods. The device can be implemented in hardware and / or software and can be configured in electronic devices that carry industrial equipment fault diagnosis functions, such as in a client or server.

[0085] See Figure 4 The industrial equipment fault diagnosis device shown includes:

[0086] Alarm code extraction module 410 is used to extract target alarm codes from fault trigger messages of industrial equipment;

[0087] The diagnostic context acquisition module 420 is used to call the Markov chain state model to process the target alarm code in order to obtain the diagnostic context output by the Markov chain state model; the Markov chain state model includes a set of states and a state transition probability matrix between states, the set of states includes alarm code states, failure mode states and failure cause states; the diagnostic context includes candidate failure mode information and candidate failure cause information.

[0088] The diagnostic scheme generation module 430 is used to generate a fault diagnosis scheme based on the candidate failure mode information and the candidate failure cause information.

[0089] In an optional embodiment of the present invention, the diagnostic context acquisition module 420 includes:

[0090] The diagnostic entry unit is used to take the alarm code state corresponding to the target alarm code as the diagnostic entry state of the Markov chain state model.

[0091] The candidate failure mode unit is used to obtain candidate failure mode information based on the first transition probability from the diagnostic entry state to each failure mode state.

[0092] The candidate failure cause unit is used to obtain candidate failure cause information based on the second transition probability from the failure mode state corresponding to each failure cause state.

[0093] A diagnostic context unit is used to generate a diagnostic context that includes the candidate failure mode information and the candidate failure cause information.

[0094] In an optional embodiment of the present invention, the diagnostic context acquisition module 420 includes a transition probability generation unit, the transition probability generation unit comprising:

[0095] The diagnostic path subunit is used to extract a diagnostic path, including alarm codes, failure modes, and failure causes, from the fault cause dataset of the industrial equipment.

[0096] The propagation chain subunit is used to construct a failure propagation chain based on the extracted diagnostic path;

[0097] The first transition probability subunit is used to calculate the first transition probability of the diagnostic path from the alarm code to each failure mode based on the failure mode occurrence probability and failure mode detectability of the same alarm code to each failure mode in the failure propagation chain.

[0098] The second transition probability subunit is used to calculate the second transition probability of the diagnostic path from the same failure mode to each failure cause in the failure propagation chain based on the severity of the failure mode to each failure cause.

[0099] In an optional embodiment of the present invention, the first transition probability subunit is specifically used for:

[0100] The product of the failure mode occurrence probability and the failure mode detectability of the same alarm code to each failure mode in the failure propagation chain is normalized, and the normalization result is used as the first transition probability of the diagnostic path from the alarm code to each failure mode.

[0101] In an optional embodiment of the present invention, each row of the state transition probability matrix corresponds to a source state, each column corresponds to a target state, each element in a row represents the probability of transitioning from the source state to the corresponding target state, and the sum of the probabilities in each row is 1.

[0102] In an optional embodiment of the present invention, the diagnostic scheme generation module 430 includes:

[0103] The knowledge retrieval request unit is used to initiate a knowledge retrieval request to the product knowledge database based on the candidate failure mode information and the candidate failure cause information.

[0104] A knowledge retrieval response unit is configured to receive a knowledge retrieval response message returned by the product knowledge database; the knowledge retrieval response message includes knowledge evidence that matches the candidate failure mode information and the candidate failure cause information.

[0105] The diagnostic scheme unit is used to fuse the candidate failure mode information, the candidate failure cause information, and the knowledge evidence to generate the fault diagnosis scheme.

[0106] In an optional embodiment of the present invention, the diagnostic scheme unit is specifically used for:

[0107] Each candidate failure cause is matched with the knowledge evidence to determine the relevance of each candidate failure cause;

[0108] The priority of each candidate failure cause is determined based on the path probability of the candidate failure path in which each candidate failure cause is located, and the relevance.

[0109] Based on the priority of the candidate failure causes, the candidate failure mode information, the candidate failure cause information, and the knowledge evidence, fault diagnosis prompts are generated;

[0110] The fault diagnosis prompts are input into a large language model to generate a fault diagnosis solution.

[0111] In an optional embodiment of the present invention, the product knowledge database is constructed based on multi-source knowledge, which includes static knowledge, dynamic knowledge, and operation and maintenance knowledge.

[0112] The static knowledge includes product design documents and functional specifications; the dynamic knowledge includes historical operating data, operating status monitoring data and alarm logs; and the operation and maintenance knowledge includes a fault handling experience base and a glossary.

[0113] In an optional embodiment of the present invention, the industrial equipment fault diagnosis device further includes a diagnosis scheme feedback module, which is specifically used for:

[0114] Receive the execution feedback message of the fault diagnosis scheme; the execution feedback message includes the execution result, the set of valid actions, the set of invalid actions, and the confirmed root cause;

[0115] Based on the execution results, the effective actions are positively correlated with the corresponding failure causes, the invalid actions are negatively correlated with the corresponding failure causes, and the confirmed root causes are marked as confirmed, in order to correct the fault diagnosis scheme.

[0116] The industrial equipment fault diagnosis device provided in this embodiment of the invention can execute the industrial equipment fault diagnosis method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0117] According to embodiments of the present invention, the present invention also provides an electronic device, a computer-readable storage medium, and a computer program product.

[0118] like Figure 5As shown, the electronic device 500 may include a memory 51, a processor 52 and a bus, and may also include a computer program stored in the memory 51 and executable on the processor 52, such as an industrial equipment fault diagnosis program.

[0119] The memory 51 includes at least one type of computer-readable storage medium, including flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 51 can be an internal storage unit of the electronic device 500, such as the portable hard drive of the electronic device 500. In other embodiments, the memory 51 can also be an external storage device of the electronic device 500, such as a plug-in portable hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 500. Furthermore, the memory 51 can include both internal and external storage units of the electronic device 500. The memory 51 can be used not only to store application software and various types of data installed on the electronic device 500, such as fault diagnosis codes for industrial equipment, but also to temporarily store data that has been output or will be output.

[0120] In some embodiments, the processor 52 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 52 is the control unit of the electronic device 500, connecting various components of the electronic device 500 via various interfaces and lines. It executes programs or modules (such as industrial equipment fault diagnosis programs) stored in the memory 51, and calls data stored in the memory 51 to perform various functions and process data of the electronic device 500.

[0121] The processor 52 executes the operating system of the electronic device 500 and various installed application programs. The processor 52 executes the application programs to implement the steps in the above-described industrial equipment fault diagnosis method.

[0122] For example, a computer program may be divided into one or more modules, one or more of which are stored in memory 51 and executed by processor 52 to complete this application. One or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in electronic device 500.

[0123] For example, a computer program can be divided into an alarm code extraction module 410, a diagnostic context acquisition module 420, and a diagnostic scheme generation module 430.

[0124] The integrated unit implemented as a software functional module described above can be stored in a computer-readable storage medium, which can be non-volatile or volatile. The software functional module stored in the storage medium includes several instructions to cause a computer device (which may be a personal computer, computer equipment, or network device, etc.) or processor to execute some functions of the industrial equipment fault diagnosis method of the various embodiments of this application.

[0125] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0126] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for diagnosing faults in industrial equipment, characterized in that, include: Extract the target alarm code from the fault trigger message of industrial equipment; The target alarm code is processed by calling a Markov chain state model to obtain the diagnostic context output by the Markov chain state model; the Markov chain state model includes a set of states and a state transition probability matrix between states, the set of states includes alarm code states, failure mode states, and failure cause states; the diagnostic context includes candidate failure mode information and candidate failure cause information. A fault diagnosis scheme is generated based on the candidate failure mode information and the candidate failure cause information.

2. The method according to claim 1, characterized in that, The process of calling the Markov chain state model to process the target alarm code to obtain the diagnostic context output by the Markov chain state model includes: The alarm code state corresponding to the target alarm code is used as the diagnostic entry state of the Markov chain state model; Based on the first transition probability from the diagnostic entry state to each failure mode state, candidate failure mode information is obtained; Based on the second transition probability from the failure mode state to each failure cause state corresponding to the candidate failure mode information, the candidate failure cause information is obtained; Generate a diagnostic context that includes the candidate failure mode information and the candidate failure cause information.

3. The method according to claim 2, characterized in that, The first transition probability and the second transition probability are determined in the following manner: Extract diagnostic paths, including alarm codes, failure modes, and failure causes, from the fault cause dataset of the industrial equipment; Construct a failure propagation chain based on the extracted diagnostic pathways; Based on the failure mode occurrence probability and failure mode detectability of the same alarm code to each failure mode in the failure propagation chain, calculate the first transition probability of the diagnostic path from the alarm code to each failure mode. Based on the severity of the same failure mode to each failure cause in the failure propagation chain, calculate the second transition probability of the diagnostic path from the failure mode to each failure cause.

4. The method according to claim 3, characterized in that, The step of calculating the first transition probability of the diagnostic path from the alarm code to each failure mode based on the failure mode occurrence probability and failure mode detectability in the failure propagation chain includes: The product of the failure mode occurrence probability and the failure mode detectability of the same alarm code to each failure mode in the failure propagation chain is normalized, and the normalization result is used as the first transition probability of the diagnostic path from the alarm code to each failure mode.

5. The method according to claim 1, characterized in that, In the state transition probability matrix, each row corresponds to a source state, each column corresponds to a target state, and each element in a row represents the probability of transitioning from the source state to the corresponding target state. The sum of the probabilities in each row is 1.

6. The method according to claim 1, characterized in that, The step of generating a fault diagnosis scheme based on the candidate failure mode information and candidate failure cause information includes: Based on the candidate failure mode information and the candidate failure cause information, a knowledge retrieval request is initiated to the product knowledge database; Receive a knowledge retrieval response message returned by the product knowledge database; the knowledge retrieval response message includes knowledge evidence that matches the candidate failure mode information and the candidate failure cause information; The candidate failure mode information, the candidate failure cause information, and the knowledge evidence are fused to generate the fault diagnosis scheme.

7. The method according to claim 6, characterized in that, The process of fusing the candidate failure mode information, the candidate failure cause information, and the knowledge evidence to generate the fault diagnosis solution includes: Each candidate failure cause is matched with the knowledge evidence to determine the relevance of each candidate failure cause; The priority of each candidate failure cause is determined based on the path probability of the candidate failure path in which each candidate failure cause is located, and the relevance; the candidate failure path refers to the diagnostic link that starts from the target alarm code state, transitions through the failure mode state to the failure cause state; the path probability is used to characterize the probability of the candidate failure path occurring. Based on the priority of the candidate failure causes, the candidate failure mode information, the candidate failure cause information, and the knowledge evidence, fault diagnosis prompts are generated; The fault diagnosis prompts are input into a large language model to generate a fault diagnosis solution.

8. The method according to claim 6, characterized in that, The product knowledge database is built based on multi-source knowledge, which includes static knowledge, dynamic knowledge, and operation and maintenance knowledge. The static knowledge includes product design documents and functional specifications; the dynamic knowledge includes historical operating data, operating status monitoring data and alarm logs; and the operation and maintenance knowledge includes a fault handling experience base and a glossary.

9. The method according to claim 1, characterized in that, The method further includes: Receive the execution feedback message of the fault diagnosis scheme; the execution feedback message includes the execution result, the set of valid actions, the set of invalid actions, and the confirmed root cause; Based on the execution results, the effective actions are positively correlated with the corresponding failure causes, the invalid actions are negatively correlated with the corresponding failure causes, and the confirmed root causes are marked as confirmed, in order to correct the fault diagnosis scheme.

10. An industrial equipment fault diagnosis device, characterized in that, include: The alarm code extraction module is used to extract target alarm codes from fault trigger messages of industrial equipment; The diagnostic context acquisition module is used to call the Markov chain state model to process the target alarm code in order to obtain the diagnostic context output by the Markov chain state model; the Markov chain state model includes a set of states and a state transition probability matrix between states, the set of states includes alarm code states, failure mode states, and failure cause states; the diagnostic context includes candidate failure mode information and candidate failure cause information. The diagnostic scheme generation module is used to generate a fault diagnosis scheme based on the candidate failure mode information and the candidate failure cause information.

11. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the method according to any one of claims 1-9.

13. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-9.

Images