A wedge cross rolling model fault early warning system and method based on large model and knowledge graph

Abstract

The present invention discloses a wedge-cross rolling model fault warning system and method based on a large model and a knowledge graph, which belongs to the field of fault diagnosis. The system includes: a model fault knowledge graph module performs structured analysis on the faults of the industrial model based on the model structure, operation standard and semantic information, and constructs a wedge-cross rolling model fault knowledge graph; a model fault distributed diagnosis module is used to construct a distributed decision model and determine the operating state, warning part object and warning cause of the wedge-cross rolling model based on the distributed decision model; a large model diagnosis reasoning module is used to perform fault diagnosis on the fault object based on the operating state of the wedge-cross rolling model in combination with the wedge-cross rolling model fault knowledge graph to obtain a diagnosis result. The present invention constructs a knowledge graph and combines machine learning and a large language model to perform wedge-cross rolling model fault diagnosis, thereby effectively improving the accuracy of wedge-cross rolling model fault diagnosis, optimizing the fault diagnosis process and improving the diagnosis efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis, and in particular, relates to a wedge cross rolling model fault early warning system and method based on a large model and a knowledge graph. Background Art

[0002] Wedge cross rolling technology is a technology that heats the bar and feeds it into the middle of two co-rotating dies with wedge-shaped protrusions. Driven by the dies, the bar makes reverse rotational motion. At the same time, the material undergoes radial compression deformation and axial extension deformation, thereby forming stepped shaft parts. This technology is an advanced forming technology for shaft parts and is listed as one of the key technologies promoted by the country. It is widely used in the production of shaft parts for automobiles, tractors, construction machinery, drilling machinery, coal mining machinery, and hardware tools.

[0003] Traditional industrial model fault detection technology usually relies on empirical rules and simple statistical analysis, which is difficult to deal with complex failure modes, resulting in low detection accuracy and efficiency, and often unable to detect potential problems in a timely manner. These traditional methods lack the ability to deeply analyze massive data and are easily affected by equipment noise and environmental changes.

[0004] Therefore, the present invention proposes a wedge cross rolling model fault warning system and method based on a large model and a knowledge graph. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes a wedge cross rolling model fault warning system and method based on a large model and a knowledge graph to solve the problems existing in the above-mentioned prior art.

[0006] To achieve the above object, the present invention provides a wedge cross rolling model fault early warning system and method based on a large model and a knowledge graph, comprising:

[0007] Model fault knowledge graph module, model fault distributed diagnosis module and large model diagnosis reasoning module;

[0008] The model fault knowledge graph module performs structured analysis on the faults of the industrial model based on the model structure, operation standard and semantic information, and constructs a knowledge graph of the cross-wedge rolling model faults;

[0009] The model fault distributed diagnosis module is used to construct a distributed decision model and determine the operating state, warning part objects and warning reasons of the cross-wedge rolling model based on the distributed decision model;

[0010] The large model diagnosis and reasoning module is used to perform fault diagnosis on the fault object based on the operating status of the cross-wedge rolling model in combination with the cross-wedge rolling model fault knowledge graph to obtain a diagnosis result.

[0011] Optionally, the model fault knowledge graph module includes: a first fault specification unit and a second fault specification unit;

[0012] Wherein, the first fault specification unit parses the model structure specification from the model part entity objects and the transmission relationship between the model part entity objects to obtain the model structure knowledge layer;

[0013] The second fault specification unit is used to parse the model structure specification from the model part entity object and the usage attribute standard of the model part entity object to obtain the part operation standard knowledge layer.

[0014] Optionally, the model fault distributed diagnosis module includes: a data receiving unit, a model state prediction unit and a data integration processing unit;

[0015] Wherein, the data receiving unit is used to record the operating data of the components;

[0016] The model state prediction unit is used to construct a decision tree model and an LSTM prediction model, and predict the operating state of the component in the next time segment based on the operating data of the component;

[0017] The data integration processing unit is used to collect the operating data of each component and splice the operating data of each component into a model overall data set.

[0018] Optionally, the model state prediction unit includes: a mechanical transmission device prediction model, a universal joint device prediction model, an equipment support device prediction model, a hydraulic device prediction model and a roller device prediction model;

[0019] Wherein, when the prediction results of the prediction model in the model state prediction unit are all normal, no diagnostic reasoning operation is performed;

[0020] When the prediction result of the existing prediction model is a fault state, the LSTM model prediction is not performed, and the fault state distribution data is sent directly;

[0021] When the prediction results of the prediction model are only normal state and warning state, the LSTM model prediction is performed to obtain the warning state distribution data.

[0022] Optionally, the large model diagnosis reasoning module includes: a specification information query unit and a fault warning diagnosis unit;

[0023] Wherein, the specification information query unit is used to query the part operation standard specification information of the warning part object from the cross-wedge rolling model fault knowledge graph;

[0024] The fault warning diagnosis unit uses the prompt engineering in combination with the part operation standard specification information and component operation data to complete the model operation fault warning diagnosis.

[0025] Optionally, the prompting project includes: setting background information, prompting domain knowledge, and providing task instructions.

[0026] The present invention also provides an early warning method of a wedge cross rolling model fault early warning system based on a large model and a knowledge graph, comprising the following steps:

[0027] According to the model structure and operation standards, a knowledge graph of wedge cross rolling model faults is constructed;

[0028] Constructing a distributed decision model, acquiring real-time data of cross wedge rolling, pre-diagnosing the operation status of the real-time data of cross wedge rolling based on the distributed decision model, and determining the warning part object;

[0029] Constructing a query statement, and retrieving the operation standard specification of the warning part object from the wedge cross rolling model fault knowledge graph based on the query statement;

[0030] The warning part object and the operating standard specification of the warning part object are combined and introduced into the large language model for diagnosis to generate fault warning diagnosis results.

[0031] The present invention also provides a computer terminal device, comprising:

[0032] one or more processors;

[0033] A memory, coupled to the processor, for storing one or more programs;

[0034] When the one or more programs are executed by the one or more processors, the one or more processors implement a wedge cross rolling model fault warning system based on a large model and a knowledge graph.

[0035] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a wedge cross rolling model fault warning system based on a large model and a knowledge graph is implemented.

[0036] Compared with the prior art, the present invention has the following advantages and technical effects:

[0037] (1) Through the knowledge graph module, the system of the present invention can perform structured analysis on the faults of the wedge cross rolling model, covering the structure, operation standards and semantic information of the model, and comprehensively and systematically organize the failure modes of the equipment, which helps to accurately locate the cause of the failure and the related equipment, thereby improving the efficiency and accuracy of fault diagnosis.

[0038] (2) The present invention uses a distributed decision model and LSTM prediction model to monitor and predict the operating status (normal state, warning state and fault state) of each branch device in real time, and discover potential problems in advance by predicting data of future time segments. This modular and distributed diagnosis method enhances the system's fault prediction ability, improves the system's flexibility and ability to cope with complex equipment environments.

[0039] (3) By introducing the fault information and prompt engineering in the knowledge graph into the large language model, the system can perform intelligent reasoning based on multi-level operating standards and historical data, thereby generating more accurate fault warning and diagnosis results. This combination enables the system to use the powerful reasoning ability of the large language model to provide intelligent support for early warning diagnosis when dealing with complex fault scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0041] Figure 1 It is a structural diagram of a wedge cross rolling model fault warning system based on a large model and a knowledge graph according to an embodiment of the present invention;

[0042] Figure 2 It is a flow chart of a fault warning method of a wedge cross rolling model based on a large model and a knowledge graph according to an embodiment of the present invention;

[0043] Figure 3 This is a structural fragment diagram of the knowledge graph of the roll device fault of the wedge cross rolling model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0045] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0046] Embodiment 1

[0047] In order to solve the problems of low accuracy and efficiency of traditional industrial model fault detection technology, the present invention provides a wedge cross rolling model fault warning system and method based on a large model and knowledge graph. By constructing a multi-dimensional and multi-granular standard specification knowledge graph, and a model fault distributed diagnosis module based on decision trees and LSTM, efficient and accurate retrieval of model operation anomalies in the knowledge graph is achieved, and a large language model is combined to realize model fault prediction and diagnosis.

[0048] Compared with traditional fault detection technology, the use of fault prediction and knowledge graph technology in equipment health management and fault warning systems can effectively promote the in-depth development of intelligent manufacturing. Fault prediction collects key equipment data (such as temperature, vibration, and pressure) through sensors, and predicts possible equipment failures through data preprocessing, feature extraction, and model training. Traditional statistical models (such as time series analysis), machine learning methods (such as support vector machines and random forests), and deep learning models (such as LSTM and CNN) are often used for such predictions. In addition, generative adversarial networks (GANs) and graph neural networks (GNNs) in deep learning have also been introduced to improve the accuracy and adaptability of predictions.

[0049] The knowledge graph systematizes and structures industrial knowledge to support knowledge query and assist decision making. Through information extraction technology, unstructured equipment documents and maintenance records are converted into structured knowledge and stored using a graph database. Through the reasoning function of the knowledge graph, the system can generate health management recommendations and provide references for fault diagnosis. The knowledge graph not only helps to understand complex data, but also assists fault analysis through reasoning capabilities, becoming an important supplement to the prediction model.

[0050] The combination of fault prediction and knowledge graphs has greatly enhanced the fault warning and intelligent decision-making capabilities of industrial systems. Knowledge graphs systematize domain knowledge and provide background information and references for prediction models; while prediction models inject real-time data into knowledge graphs to help dynamically update and expand knowledge graphs. The dual support of data-driven and knowledge-driven brought about by this fusion is the key to the advancement of intelligent manufacturing technology. On this basis, Figure 1 As shown, this embodiment provides a wedge cross rolling model fault warning system based on a large model and a knowledge graph, and the system includes: a model fault knowledge graph module, a model fault distributed diagnosis module and a large model diagnosis reasoning module.

[0051] The model fault knowledge graph module is used to perform structured analysis on the faults of the industrial model from the perspectives of model structure, operation standards and semantic information, and to construct a knowledge graph of wedge cross rolling model faults.

[0052] The model fault distributed diagnosis module is used to construct a distributed decision-making model, which is used to determine the operating status of each part of the decision-making model, the warning (fault) object and the warning (fault) cause.

[0053] The large model diagnosis reasoning module is used to retrieve each warning part object from the wedge cross rolling model fault knowledge graph according to the warning part objects and warning reasons of each branch, and introduce it into the large language model for fault warning diagnosis through prompt engineering combined with the model fault knowledge graph to generate a diagnosis result.

[0054] Furthermore, the model fault knowledge graph module includes: a first fault specification unit and a second fault specification unit; wherein, the first fault specification unit parses the model structure specification from the model part entity object and the transmission relationship between the model part entity objects to obtain the model structure knowledge layer; the second fault specification unit is used to parse the model structure specification from the model part entity object and the usage attribute standard of the model part entity object to obtain the part operation standard knowledge layer.

[0055] As a specific implementation of this embodiment, the method of constructing the model fault knowledge graph is as follows: a two-layer model architecture is used for the model part operation standard, and the model operation faults are standardized, constrained and integrated from the model structure knowledge layer and the part operation standard knowledge layer. Figure 3 shown.

[0056] Furthermore, the model structure knowledge layer is constructed based on the model transmission structure connection tree. By parsing the model structure manual, the model part entity objects and the transmission relationship between them are extracted from the semantic dimension. First, they are classified according to the characteristic attributes of the corresponding description of each part, and the classification categories are set as mechanical transmission device class, universal joint device class, equipment support device class, hydraulic device class and rolling mill device class; then the part knowledge contained in each type of device is summarized and organized, and refined into part entity objects; finally, the transmission relationship is defined between the related part objects.

[0057] Furthermore, the part operation standard knowledge layer is constructed based on the model performance standard combing. By parsing the model instruction manual, the model part entity objects and usage attribute standards are constructed from the semantic dimension. First, the part entity objects in the layer are listed according to the model structure knowledge layer; then the performance standard knowledge contained in each part entity object is summarized and organized, and refined into standard entity objects; finally, the behavioral relationship is defined between the associated part entity objects and standard entity objects.

[0058] As a specific implementation of this embodiment, the model fault distributed diagnosis module includes a data receiving unit, a model state prediction unit and a data integration processing unit. The data receiving unit is used to record the operating data of the components; the model state prediction unit is used to build a decision tree model and an LSTM prediction model, and predict the operating state of the components in the next time segment based on the operating data of the components; the data integration processing unit is used to collect the operating data of each component, and splice the operating data of each component into a model overall data set.

[0059] The data receiving unit is used to receive the operating data information of each component of the current marking type recorded by the model sensor. The sensor is deployed at the performance collection location of each component of the wedge cross rolling model, and label information is added according to the model division. The collected data includes but is not limited to the model current and voltage, amplitude and speed data, and finished product vortex center detection, necking offset data, etc. The information is pre-processed by slicing according to the label classification and then passed to the LSTM state prediction module.

[0060] The model state prediction unit constructs a decision tree model and an LSTM prediction model in a distributed manner to predict the operating status and future operating data of each branch model. First, the operating data of each category of parts are received respectively through the constructed decision tree model, and the operating status of each branch is predicted through the trained model: divided into normal state, warning state and fault state; then the corresponding distribution data of the part to be predicted is sent to the LSTM prediction model, and the part operating data of the next time segment is predicted through the trained model; finally, the data of the warning state or fault state is sent to the data integration processing module.

[0061] The data integration processing unit organizes and processes the received data from each section. First, it detects whether there is fault data. If so, it immediately interrupts the model running process to reduce resource loss; then it reorganizes the data into the [{part object: {type: type value, attribute: attribute value, ...}, ...}] format; finally, it splices the data from each section into the overall model data set and sends it to the large model diagnosis and reasoning module.

[0062] Furthermore, when the prediction model is deployed, it is constructed according to the type of equipment parts and is divided into a mechanical transmission device prediction model, a universal coupling device prediction model, an equipment support device prediction model, a hydraulic device prediction model and a rolling mill device prediction model, and a Leader is set to coordinate the prediction process: the Leader receives and records the Worker status prediction information. If all Worker prediction results are normal, no subsequent operations are performed; if there is a Worker prediction result that is a fault state, the LSTM model prediction is not performed, and the fault state distribution data is sent directly; if the Worker prediction results are only normal and warning states, the LSTM model is used to predict the corresponding warning state distribution data; finally, the Leader receives the corresponding amount of distribution data information according to the result record, encapsulates the data, and sends it to the data integration processing module.

[0063] Furthermore, the large model diagnosis and reasoning module includes a specification information query unit and a fault warning diagnosis unit. The specification information query unit is used to query the part operation standard specification information of the warning part object from the wedge cross rolling model fault knowledge graph; the fault warning diagnosis unit completes the model operation fault warning diagnosis based on the part operation standard specification information and component operation data.

[0064] As a specific implementation method of this embodiment, the specific process of the audit method of the large model diagnostic reasoning module includes: first, based on the warning (fault) part object obtained by the model fault distributed diagnosis module, a Cypher statement is constructed to query the relevant part operation standard specification information from the knowledge graph; secondly, a prompt project is set up to guide the large model to diagnose the warning (fault) parts according to the part operation standard specifications; finally, after the prediction model warning part operation knowledge is automatically queried, the prompt project is used to combine the part data and the operation standard specification knowledge, and simultaneously introduce them into the large model to complete the model operation fault warning diagnosis process.

[0065] Furthermore, in the Cypher query process, since some nodes on the path will be frequently filtered according to their attributes when searching for all paths starting from the warning (fault) part object, an index can be created on the warning (fault) part object node to improve query efficiency. The Cypher query statement for the roller part constructed is:

[0066] MATCH path =(n:part{name:'upper roller'})-[:attribute*]->(m)

[0067] RETURN path

[0068] Furthermore, the prompting project includes: setting background information, setting appropriate background information according to the model operation environment, guiding and limiting the thinking scope and language style of the large model to a certain extent; prompting domain knowledge, giving reference to the model operation standard specification content, so that the large model can fully understand the diagnostic knowledge points of the standard specification; providing task instructions, providing the large model with clear and accurate task descriptions, so that the large model can give correct early warning diagnosis results according to the set requirements.

[0069] After collecting model operation data, judging the operation status and predicting future operation data, the prompt engineering is combined with the standard specification knowledge of the wedge cross rolling model operation parts and introduced into the large model, such as ChatGPT, etc., using the powerful semantic understanding and reasoning capabilities of LLM to complete the model fault prediction and diagnosis process.

[0070] Embodiment 2

[0071] like Figure 2 As shown, this embodiment also provides a wedge cross rolling model fault warning method based on a large model and a knowledge graph, and the method includes the following steps:

[0072] Step 1: Construct a knowledge graph of wedge cross rolling model faults based on the model structure and operation standards;

[0073] Step 2: Pre-diagnose the running status of the real-time data of the model, obtain the warning (fault) part object, and predict the future running data of the part;

[0074] Step 3: construct a query statement based on the part entity object obtained after processing by the prediction module, and retrieve the part entity object and its operating standard specification from the knowledge graph of the wedge cross rolling model operating standard specification;

[0075] Step 4: Combine the model warning part object and the model part operation standard specification object, and introduce them into the large language model for diagnosis to generate fault warning diagnosis results.

[0076] This embodiment constructs a knowledge graph and combines machine learning and a large language model to perform wedge cross rolling model fault diagnosis, thereby effectively improving the accuracy of wedge cross rolling model fault diagnosis, optimizing the fault diagnosis process, and improving diagnosis efficiency.

[0077] The present invention uses modern intelligent detection methods based on large models and knowledge graphs to collect and analyze equipment operating data, such as temperature, vibration, pressure, etc. in real time. The fault warning system can quickly identify abnormal conditions and warn of potential problems, thereby realizing intelligent monitoring and predictive diagnosis of equipment operating status and ensuring production continuity and safety.

[0078] Embodiment 3

[0079] This embodiment also provides a computer terminal device, including:

[0080] one or more processors;

[0081] A memory, coupled to the processor, for storing one or more programs;

[0082] When the one or more programs are executed by the one or more processors, the one or more processors implement a wedge cross rolling model fault warning system based on a large model and a knowledge graph.

[0083] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, a wedge cross rolling model fault warning system based on a large model and a knowledge graph is implemented.

[0084] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A wedge cross rolling model fault warning system based on a large model and knowledge graph, characterized in that: include: Model fault knowledge graph module, model fault distributed diagnosis module and large model diagnosis reasoning module; The model fault knowledge graph module performs structured analysis on the faults of the industrial model based on the model structure, operation standard and semantic information, and constructs a knowledge graph of the cross-wedge rolling model faults; The model fault knowledge graph module includes: a first fault specification unit and a second fault specification unit; wherein the first fault specification unit parses the model structure specification from the model part entity object and the transmission relationship between the model part entity objects to obtain the model structure knowledge layer; the second fault specification unit is used to parse the model structure specification from the model part entity object and the use attribute standard of the model part entity object to obtain the part operation standard knowledge layer; The model fault distributed diagnosis module is used to construct a distributed decision model and determine the operating state, warning part objects and warning reasons of the cross-wedge rolling model based on the distributed decision model; The model fault distributed diagnosis module includes: a data receiving unit, a model state prediction unit and a data integration processing unit; wherein the data receiving unit is used to record the operating data of the components; the model state prediction unit is used to construct a decision tree model and an LSTM prediction model, and predict the operating state of the components in the next time segment based on the operating data of the components; the data integration processing unit is used to collect the operating data of each component, and splice the operating data of each component into a model overall data set; The large model diagnosis and reasoning module is used to perform fault diagnosis on the fault object based on the operating state of the cross-wedge rolling model in combination with the cross-wedge rolling model fault knowledge graph to obtain a diagnosis result; The large model diagnostic reasoning module includes: a specification information query unit and a fault warning diagnosis unit; wherein, the specification information query unit is used to query the part operation standard specification information of the warning part object from the wedge cross rolling model fault knowledge graph; the fault warning diagnosis unit uses prompt engineering combined with the part operation standard specification information and component operation data to complete the model operation fault warning diagnosis.

2. The wedge cross rolling model fault early warning system based on a large model and knowledge graph according to claim 1 is characterized in that: The model state prediction unit includes: a mechanical transmission device prediction model, a universal joint device prediction model, an equipment support device prediction model, a hydraulic device prediction model and a roller device prediction model; Wherein, when the prediction results of the prediction model in the model state prediction unit are all normal, no diagnostic reasoning operation is performed; When the prediction result of the existing prediction model is a fault state, the LSTM model prediction is not performed, and the fault state distribution data is sent directly; When the prediction results of the prediction model are only normal state and warning state, the LSTM model prediction is performed to obtain the warning state distribution data.

3. The wedge cross rolling model fault early warning system based on a large model and knowledge graph according to claim 1 is characterized in that: The prompting project includes: setting background information, prompting domain knowledge, and providing task instructions.

4. The early warning method of the wedge cross rolling model fault early warning system based on a large model and a knowledge graph according to claim 1 is characterized in that: The following steps are involved: According to the model structure and operation standards, a knowledge graph of wedge cross rolling model faults is constructed; Constructing a distributed decision model, acquiring real-time data of cross wedge rolling, pre-diagnosing the operation status of the real-time data of cross wedge rolling based on the distributed decision model, and determining the warning part object; Constructing a query statement, and retrieving the operation standard specification of the warning part object from the wedge cross rolling model fault knowledge graph based on the query statement; The warning part object and the operating standard specification of the warning part object are combined and introduced into the large language model for diagnosis to generate fault warning diagnosis results.

5. A computer terminal device, characterized in that: include: one or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the early warning method as claimed in claim 4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the early warning method as claimed in claim 4 is implemented.

Images