A hierarchical cognitive method and system for equipment fault diagnosis

Abstract

The application relates to the technical field of fault diagnosis, and discloses a layered cognition method and system for equipment fault diagnosis, which comprises the following steps: acquiring running data in real time; matching the running data with a rule library; if the matching is successful, obtaining a fault type and a maintenance measure, and performing fault treatment; if the matching fails, pre-processing the running data, extracting features, and obtaining feature data; inputting the running data and the feature data into a fault analysis model for identification; if the identification is successful, obtaining the fault type and the maintenance measure as a preliminary diagnosis scheme; based on the preliminary diagnosis scheme, the feature data and the running data, comprehensively analyzing and generating a fault type and a maintenance measure, and performing fault treatment; if the identification fails, inputting the running data and the feature data into a large model engine for reasoning; if the reasoning result comprises a fault type and a maintenance measure, performing fault treatment; and if the reasoning result does not comprise a fault type and a maintenance measure, not performing fault treatment. The application can cope with complex fault scenarios, and improve fault diagnosis accuracy and efficiency.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and more specifically, to a hierarchical cognitive method and system for equipment fault diagnosis. Background Technology

[0002] With the rapid development of industrial automation and artificial intelligence technologies, equipment fault diagnosis is gradually becoming more intelligent. Fault diagnosis methods based on neural networks are receiving increasing attention and have become a new research hotspot in the field of fault diagnosis.

[0003] In existing technologies, fault diagnosis is often analyzed from a single dimension. When faced with complex faults such as multi-source concurrency, it is difficult to accurately identify the type, which can easily lead to misjudgment or missed detection, resulting in delayed fault detection and thus increasing downtime and maintenance costs. Summary of the Invention

[0004] The technical objective of this invention is to address the shortcomings of the prior art by providing a hierarchical cognitive method and system for equipment fault diagnosis that can effectively improve the accuracy and efficiency of fault diagnosis in complex fault scenarios.

[0005] The first technical solution adopted in this invention is as follows: A hierarchical cognitive method for equipment fault diagnosis, the hierarchical cognitive method comprising the following steps: S1. Acquire real-time operating data of the equipment; S2. Based on the running data, perform fault matching with the preset rule base. If the matching is successful, obtain the fault type and corresponding maintenance measures, and perform fault handling based on the maintenance measures. If the matching fails, preprocess the running data and extract features to obtain feature data. S3. Input the operating data and feature data into a preset fault analysis model for fault identification. If the identification is successful, obtain the fault type and corresponding maintenance measures as a preliminary diagnosis plan. Based on the preliminary diagnosis plan, feature data, and operating data, perform comprehensive analysis to generate the fault type and corresponding maintenance measures, and perform fault handling based on the maintenance measures. If the identification fails, input the operating data and feature data into a preset large model engine for reasoning to generate a reasoning result. If the reasoning result includes the fault type and corresponding maintenance measures, perform fault handling based on the maintenance measures. If the reasoning result does not include the fault type and corresponding maintenance measures, do not perform fault handling.

[0006] The above-mentioned technical measures can quickly identify simple faults by using a rule base, identify faults that the rule base cannot handle by using a fault analysis model, and identify faults that neither the rule base nor the fault analysis model can handle by using a large model engine. By setting up a three-layer progressive diagnosis of "rule base", "fault analysis model" and "large model engine", the fault identification and judgment coverage can be improved, and effective response to diverse fault scenarios can be achieved, thereby improving the accuracy and efficiency of fault diagnosis.

[0007] Furthermore, the preprocessing includes standardizing the runtime data.

[0008] The above-mentioned technical measures effectively improve the quality of operational data by standardizing the processing of operational data, thereby enhancing the stability and accuracy of subsequent fault diagnosis.

[0009] Furthermore, the feature extraction includes multi-dimensional feature extraction from the preprocessed running data.

[0010] The above-mentioned technical measures, through multi-dimensional feature extraction, can perform analysis from multiple dimensions, effectively avoiding the problem of delayed fault occurrence caused by single-dimensional analysis, and can reduce downtime and maintenance costs.

[0011] Furthermore, the fault analysis model includes multiple analysis models trained based on various common fault types.

[0012] The above-mentioned technical measures can improve the coverage of fault identification and judgment, effectively respond to diverse fault scenarios, and improve the accuracy and efficiency of fault diagnosis.

[0013] Furthermore, the large model engine is built on a large language model.

[0014] The above-mentioned technical measures, by utilizing large language models, can reason about faults that rule bases and fault analysis models cannot handle, generate diagnostic results, and improve generalization ability.

[0015] Furthermore, the hierarchical cognitive method also includes: S4. Generate feedback samples based on the operation data, feature data, fault types and corresponding maintenance measures, and update the rule base, fault analysis model and large model engine based on the feedback samples.

[0016] The above-mentioned technical measures update the rule base, fault analysis model, and large model engine based on feedback samples, thereby achieving iterative optimization of the rule base, fault analysis model, and large model engine, effectively improving the accuracy and efficiency of fault identification and judgment.

[0017] The second technical solution adopted in this invention is as follows: A hierarchical cognitive system for equipment fault diagnosis, the hierarchical cognitive system comprising: The system includes a data acquisition module, a rule base module, a data processing module, an analysis model module, a diagnostic decision-making module, a large model module, and a response execution module. The data acquisition module is used to collect the device's operating data in real time and send the operating data to the rule base module; The rule base module is used to receive the operation data sent by the data acquisition module, perform fault matching based on the operation data and the preset rule base, if the matching is successful, obtain the fault type and the corresponding maintenance measures, and perform fault handling based on the maintenance measures; if the matching fails, send the operation data to the data processing module. The data processing module is used to receive the running data sent by the rule base module, preprocess the running data and extract features to obtain feature data, and send the running data and feature data to the analysis model module. The analysis model module is used to receive the operating data and feature data sent by the data processing module, input the operating data and feature data into a preset fault analysis model for fault identification, if the identification is successful, the fault type and corresponding maintenance measures are obtained and used as a preliminary diagnosis plan, and the preliminary diagnosis plan, feature data and operating data are sent to the diagnosis decision module; if the identification fails, the operating data and feature data are sent to the large model module. The diagnostic decision module is used to receive the preliminary diagnostic plan, feature data and operation data sent by the analysis model module, perform comprehensive analysis based on the preliminary diagnostic plan, feature data and operation data to generate fault types and corresponding maintenance measures and send them to the response execution module; The large model module is used to receive the running data and feature data sent by the analysis model module, input the running data and feature data into the preset large model engine for inference to generate inference results. If the inference results include the fault type and the corresponding maintenance measures, the fault type and the corresponding maintenance measures are sent to the response execution module. If the inference results do not include the fault type and the corresponding maintenance measures, no fault processing is performed. The response execution module is used to receive the fault type and corresponding maintenance measures sent by the rule base module, diagnostic decision module or large model module, and to handle the fault according to the maintenance measures.

[0018] The aforementioned technical measures, through the collaborative cooperation between the data acquisition module, rule base module, data processing module, analysis model module, diagnostic decision module, large model module, and response execution module, can improve the coverage of fault identification and judgment, effectively respond to diverse fault scenarios, and improve the accuracy and efficiency of fault diagnosis.

[0019] Furthermore, the hierarchical cognitive system also includes a feedback update module, which is connected to the rule base module, the analysis model module, the diagnostic decision module, and the large model module, respectively. The feedback update module is used to acquire the operating data, feature data, fault types, and corresponding maintenance measures and generate feedback samples. Based on the feedback samples, the rule base, the fault analysis model, and the large model engine are updated respectively.

[0020] The above-mentioned technical measures, by setting up a feedback update module, transform the data in the diagnostic process into feedback samples to update the rule base, fault analysis model, and large model engine, thereby achieving continuous improvement and upgrading of the hierarchical cognitive system.

[0021] Furthermore, the large model module has an optimization learning module, which is connected to the feedback update module. The optimization learning module is used to periodically optimize the large model engine based on the feedback samples.

[0022] The above-mentioned technical measures can improve diagnostic accuracy by setting up an optimized learning module.

[0023] Furthermore, the large model module has a knowledge update module, which is used to update the large model engine periodically through online learning.

[0024] The aforementioned technical measures, by setting up a knowledge update module and utilizing online learning, can improve diagnostic accuracy.

[0025] One or more technical solutions provided by this invention have at least the following technical effects or advantages: This invention enables rapid and timely identification of simple faults by utilizing a rule base, identifies faults that the rule base cannot handle by utilizing a fault analysis model, and identifies faults that neither the rule base nor the fault analysis model can handle by utilizing a large model engine. By setting up a three-layer progressive diagnosis of "rule base", "fault analysis model" and "large model engine", this three-layer diagnostic approach can improve the coverage of fault identification and judgment, effectively cope with diverse fault scenarios, and improve the accuracy and efficiency of fault diagnosis. Attached Figure Description

[0026] The accompanying drawings, which are provided to further illustrate embodiments of the invention and constitute a part of this invention, are not intended to limit the scope of the invention. Figure 1 This is a flowchart of the hierarchical cognition method for equipment fault diagnosis in this invention. Detailed Implementation

[0027] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, where there is no conflict, the embodiments of the present invention and the features thereof can be combined with each other.

[0028] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0029] Reference Figure 1 This embodiment proposes a hierarchical cognitive method for equipment fault diagnosis, which includes the following steps: S1. Acquire real-time operating data of the equipment; The system collects operational data in real time through multiple devices, sensors, and actuators. The operational data includes vibration data, temperature data, current data, etc., and the sampling frequency is 10kHz.

[0030] S2. Based on the running data and the preset rule base, perform fault matching. If the matching is successful, obtain the fault type and the corresponding maintenance measures, and handle the fault based on the maintenance measures. If the matching fails, preprocess the running data and extract features to obtain feature data. The rule base includes approximately 500 common fault rules, which can be set according to the actual situation.

[0031] Preprocessing includes standardizing the runtime data.

[0032] Specifically, standardization processes include data alignment, data cleaning, data imputation, and outlier removal. Data alignment methods can employ time synchronization algorithms, while data cleaning methods can employ outlier detection algorithms.

[0033] Feature extraction includes multi-dimensional feature extraction from the preprocessed running data.

[0034] Specifically, the PCA algorithm can be used for feature extraction, with a feature dimension not exceeding 100. Dimensional features include time-domain features, frequency-domain features, and statistical features.

[0035] S3. Input the operating data and feature data into the preset fault analysis model for fault identification. If the identification is successful, obtain the fault type and corresponding maintenance measures as a preliminary diagnosis plan. Based on the preliminary diagnosis plan, feature data, and operating data, perform comprehensive analysis to generate the fault type and corresponding maintenance measures, and perform fault handling based on the maintenance measures. If the identification fails, input the operating data and feature data into the preset large model engine for reasoning to generate reasoning results. If the reasoning results include the fault type and corresponding maintenance measures, perform fault handling based on the maintenance measures. If the reasoning results do not include the fault type and corresponding maintenance measures, do not perform fault handling.

[0036] The specific steps for generating fault types and corresponding maintenance measures based on a comprehensive analysis of preliminary diagnostic plans, characteristic data, and operational data are as follows: Based on the preliminary diagnostic plan, feature data, and operational data analysis, multiple candidate fault types and corresponding candidate maintenance measures, along with their matching degrees, are generated. A preset matching degree threshold is used to obtain the fault type and corresponding maintenance measures. Other types of thresholds can also be set.

[0037] The fault analysis model includes multiple analysis models trained based on various common fault types. It contains approximately 20 common fault analysis models.

[0038] The large model engine is built on a large language model.

[0039] The large model engine uses the Qwen3-VL-8B-Thinking model, which has 8B parameters and an inference latency of no more than about 1 second.

[0040] The hierarchical cognitive approach also includes: S4. Generate feedback samples based on operational data, feature data, fault types, and corresponding maintenance measures. Update the rule base, fault analysis model, and large model engine based on the feedback samples.

[0041] The fault analysis model is updated every week, and the rule base is updated every day.

[0042] Based on the above-described hierarchical cognition method, this embodiment also proposes a hierarchical cognition system for equipment fault diagnosis, which includes: The system includes a data acquisition module, a rule base module, a data processing module, an analysis model module, a diagnostic decision-making module, a large model module, and a response execution module. The data acquisition module is used to collect the device's operating data in real time and send the operating data to the rule base module; The rule base module is used to receive the operation data sent by the data acquisition module, and perform fault matching based on the operation data and the preset rule base. If the match is successful, the fault type and corresponding maintenance measures are obtained, and the fault is handled based on the maintenance measures. If the match fails, the operation data is sent to the data processing module. The data processing module receives the runtime data sent by the rule base module, preprocesses the runtime data and extracts features to obtain feature data, and then sends the runtime data and feature data to the analysis model module. The analysis model module receives the operational data and feature data sent by the data processing module, inputs the operational data and feature data into the preset fault analysis model for fault identification. If the identification is successful, the fault type and corresponding maintenance measures are obtained and used as a preliminary diagnosis plan. The preliminary diagnosis plan, feature data, and operational data are then sent to the diagnosis decision module. If the identification fails, the operational data and feature data are sent to the large model module. The diagnostic decision module receives the preliminary diagnostic plan, feature data, and operational data sent by the analysis model module. Based on the preliminary diagnostic plan, feature data, and operational data, it performs comprehensive analysis to generate fault types and corresponding maintenance measures, and sends them to the response execution module. Specifically, based on the preliminary diagnostic plan, feature data, and operational data, multiple candidate fault types and corresponding candidate maintenance measures, as well as matching degrees, are generated through analysis. A matching degree threshold is preset, and the fault type and corresponding maintenance measures are obtained based on the matching degree threshold. The matching degree threshold can be dynamically adjusted according to the load rate, with an adjustment cycle of approximately 30 minutes.

[0043] The large model module is used to receive the running data and feature data sent by the analysis model module, input the running data and feature data into the preset large model engine for inference to generate inference results. If the inference results include the fault type and the corresponding maintenance measures, the fault type and the corresponding maintenance measures are sent to the response execution module. If the inference results do not include the fault type and the corresponding maintenance measures, no fault processing is performed. The response execution module is used to receive the fault type and corresponding maintenance measures sent by the rule base module, diagnostic decision module or large model module, and to handle the fault according to the maintenance measures.

[0044] The fault handling latency of the response execution module is no more than approximately 100ms.

[0045] The hierarchical cognitive system also includes a feedback update module, which is connected to the rule base module, analysis model module, diagnostic decision module and large model module respectively. It is used to acquire operational data, feature data, fault types and corresponding maintenance measures and generate feedback samples. Based on the feedback samples, the rule base, fault analysis model and large model engine are updated respectively.

[0046] The large model module has an optimization learning module, which is connected to the feedback update module. The optimization learning module is used to optimize the large model engine periodically based on feedback samples.

[0047] The learning rate of the optimization learning module is approximately 0.01, and the learning period is 1 day.

[0048] The large model module has a knowledge update module, which is used to update the large model engine periodically through online learning.

[0049] The online learning frequency for the knowledge update module is once per hour.

[0050] The hierarchical cognitive system also includes a management module, which is connected to the data acquisition module, rule base module, data processing module, analysis model module, diagnostic decision module, large model module, response execution module, and feedback update module. The management module is used to allocate diagnostic tasks and manage data transmission between modules. The management module adopts a trust-based distributed decision algorithm to dynamically adjust task allocation according to the diagnostic accuracy and reliability of each module to achieve collaborative diagnosis. The task allocation strategy is a dynamic polling strategy.

[0051] The hierarchical cognition method and system in this embodiment can be applied to machine tools or other electromechanical equipment.

[0052] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0053] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A hierarchical cognitive method for equipment fault diagnosis, characterized in that: The hierarchical cognition method includes the following steps: S1. Acquire real-time operating data of the equipment; S2. Based on the running data, perform fault matching with the preset rule base. If the matching is successful, obtain the fault type and corresponding maintenance measures, and perform fault handling based on the maintenance measures. If the matching fails, preprocess the running data and extract features to obtain feature data. S3. Input the operating data and feature data into a preset fault analysis model for fault identification. If the identification is successful, obtain the fault type and corresponding maintenance measures as a preliminary diagnosis plan. Based on the preliminary diagnosis plan, feature data, and operating data, perform comprehensive analysis to generate the fault type and corresponding maintenance measures, and perform fault handling based on the maintenance measures. If the identification fails, input the operating data and feature data into a preset large model engine for reasoning to generate a reasoning result. If the reasoning result includes the fault type and corresponding maintenance measures, perform fault handling based on the maintenance measures. If the reasoning result does not include the fault type and corresponding maintenance measures, do not perform fault handling.

2. The hierarchical cognitive method for equipment fault diagnosis according to claim 1, characterized in that: The preprocessing includes standardizing the runtime data.

3. The hierarchical cognitive method for equipment fault diagnosis according to claim 1, characterized in that: The feature extraction includes multi-dimensional feature extraction from the preprocessed running data.

4. The hierarchical cognitive method for equipment fault diagnosis according to claim 1, characterized in that: The fault analysis model includes multiple analysis models trained based on various common fault types.

5. The hierarchical cognitive method for equipment fault diagnosis according to claim 1, characterized in that: The large model engine is built on a large language model.

6. The hierarchical cognitive method for equipment fault diagnosis according to claim 1, characterized in that: The hierarchical cognitive method also includes: S4. Generate feedback samples based on the operation data, feature data, fault types and corresponding maintenance measures, and update the rule base, fault analysis model and large model engine based on the feedback samples.

7. A hierarchical cognitive system for equipment fault diagnosis, characterized in that: The hierarchical cognitive system includes: The system includes a data acquisition module, a rule base module, a data processing module, an analysis model module, a diagnostic decision-making module, a large model module, and a response execution module. The data acquisition module is used to collect the device's operating data in real time and send the operating data to the rule base module; The rule base module is used to receive the operation data sent by the data acquisition module, perform fault matching based on the operation data and the preset rule base, if the matching is successful, obtain the fault type and the corresponding maintenance measures, and perform fault handling based on the maintenance measures; if the matching fails, send the operation data to the data processing module. The data processing module is used to receive the running data sent by the rule base module, preprocess the running data and extract features to obtain feature data, and send the running data and feature data to the analysis model module. The analysis model module is used to receive the operating data and feature data sent by the data processing module, input the operating data and feature data into a preset fault analysis model for fault identification, if the identification is successful, the fault type and corresponding maintenance measures are obtained and used as a preliminary diagnosis plan, and the preliminary diagnosis plan, feature data and operating data are sent to the diagnosis decision module; if the identification fails, the operating data and feature data are sent to the large model module. The diagnostic decision module is used to receive the preliminary diagnostic plan, feature data and operation data sent by the analysis model module, perform comprehensive analysis based on the preliminary diagnostic plan, feature data and operation data to generate fault types and corresponding maintenance measures and send them to the response execution module; The large model module is used to receive the running data and feature data sent by the analysis model module, input the running data and feature data into the preset large model engine for inference to generate inference results. If the inference results include the fault type and the corresponding maintenance measures, the fault type and the corresponding maintenance measures are sent to the response execution module. If the inference results do not include the fault type and the corresponding maintenance measures, no fault processing is performed. The response execution module is used to receive the fault type and corresponding maintenance measures sent by the rule base module, diagnostic decision module or large model module, and to handle the fault according to the maintenance measures.

8. The hierarchical cognitive system for equipment fault diagnosis according to claim 7, characterized in that: The hierarchical cognitive system also includes a feedback update module, which is connected to the rule base module, analysis model module, diagnostic decision module and large model module respectively. The feedback update module is used to acquire the operation data, feature data, fault types and corresponding maintenance measures and generate feedback samples, and update the rule base, fault analysis model and large model engine based on the feedback samples respectively.

9. The hierarchical cognitive system for equipment fault diagnosis according to claim 8, characterized in that: The large model module has an optimization learning module, which is connected to the feedback update module. The optimization learning module is used to optimize the large model engine periodically based on the feedback samples.

10. The hierarchical cognitive system for equipment fault diagnosis according to claim 7, characterized in that: The large model module has a knowledge update module, which is used to update the large model engine periodically through online learning.

Images