Downhole equipment fault tracing method based on large language model

Abstract

The invention discloses an underground equipment fault tracing method based on a large language model. The method comprises the steps that fault signals and fault reasons and solutions corresponding to the fault signals are arranged according to a maintenance ledger; predicting possible fault signals and corresponding fault reasons and solutions thereof; establishing a fault traceability database through the fault signals, the fault reasons and the solutions; establishing a pre-training model based on the BERT large model, and performing pre-training through the fault traceability database; establishing a fault traceability network, and accessing the fault signal of the underground equipment to the fault traceability network; the fault traceability network is used as an input end to access the trained BERT large model; and when the underground equipment sends out a fault signal, the BERT large model correspondingly outputs a fault reason and a solution. According to the method, the fault signals are extracted and analyzed through the BERT large model, the corresponding fault reasons and solutions are output, the analysis speed of the fault reasons and the output speed of the solutions are increased, and the fault processing speed is increased.

Description

Technical Field

[0001] The invention belongs to the technical field of power fault detection and maintenance, and relates to a method for tracing the fault of underground equipment based on a large language model. Background Art

[0002] The mine power grid is the power network of the entire coal mine. From large equipment such as main fans, hoists, drainage, belts to small equipment such as lighting, communications, and various sensors, all require safe and reliable power supply support. When a power grid fails, it may affect multiple devices. If the fault location cannot be found and repaired in time, production progress may be delayed or even dangerous.

[0003] At present, when a fault occurs, the fault signal is mainly analyzed and checked manually, and a solution is formulated. The troubleshooting process is long, so the troubleshooting time is long, which is not conducive to solving the problem in time. Summary of the invention

[0004] In view of the above problems, the present invention proposes a method for tracing the fault of underground equipment based on a large language model, which well solves the problems in the prior art.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for tracing the source of underground equipment faults based on a large language model, comprising:

[0007] Step S1, sorting out fault signals and their corresponding fault causes and solutions according to the maintenance ledger;

[0008] Step S2: predicting possible fault signals and their corresponding fault causes and solutions;

[0009] Step S3: Establish a fault tracing database based on fault signals, fault causes and solutions;

[0010] Step S4: Establish a pre-training model based on the BERT large model and perform pre-training through the fault tracing database;

[0011] Step S5, establishing a fault tracing network, and connecting the fault signal of the downhole equipment to the fault tracing network;

[0012] Step S6: The fault tracing network is used as an input terminal to access the trained BERT large model;

[0013] Step S7: When the downhole equipment sends a fault signal, the BERT large model outputs the fault cause and solution accordingly.

[0014] Optionally, the fault signal includes an independent signal and a combined signal when multiple independent signals occur synchronously;

[0015] The step S2 also includes combining independent signals, screening possible combined signals and predicting corresponding fault causes and solutions.

[0016] Optionally, step S7 includes, when the fault signal includes multiple fault causes, the BERT large model reversely traces the fault cause to other fault signal sources that may cause the fault and detects them.

[0017] Optionally, step S7 includes, based on the detection of other fault signal sources, sorting the fault causes according to likelihood, and outputting a maintenance order according to likelihood.

[0018] Optionally, the method further includes step S8, establishing an emergency contact list of the emergency group and synchronizing it in the solution. When the solution is output, the fault tracing network emergency contact list notifies the corresponding emergency group.

[0019] Optionally, step S8 further includes the emergency response team uploading a fault log after troubleshooting, supplementing the fault causes and solutions not involved, and providing feedback on the fault causes and solutions of the analyzed errors.

[0020] Optionally, the fault signal includes a narrative signal formed by narrative voice or text;

[0021] The step S3 also includes extracting narration signal keywords and establishing a keyword database;

[0022] The step S7 includes extracting the keywords of the narrative signal by the BERT large model, comparing them with the keyword database, and outputting the fault cause and solution accordingly.

[0023] Optionally, the step S3 further includes sorting the fault signals according to the control levels of the relevant devices;

[0024] The step S7 also includes, when multiple fault signals alarm and the combined signal database does not include this situation, analyzing the signal with a high control level first.

[0025] Optionally, the fault cause includes at least the fault location and the cause of the fault.

[0026] Compared with the prior art, the present invention has the following beneficial effects: by connecting the fault signal to the fault tracing network based on the BERT large model, after the fault signal is sent, the fault signal is extracted and analyzed by the BERT large model, and the corresponding fault cause and solution are output, thereby speeding up the analysis speed of the fault cause and the output speed of the solution, thereby reducing the time required for troubleshooting. The fault cause should at least include the fault location and the cause of occurrence, thereby improving the fault processing speed. DETAILED DESCRIPTION

[0027] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] The method for tracing the source of underground equipment faults based on a large language model disclosed in an embodiment of the present invention includes:

[0029] Step S1, sorting out fault signals and their corresponding fault causes and solutions according to the maintenance ledger;

[0030] Step S2: predicting possible fault signals and their corresponding fault causes and solutions;

[0031] Step S3: Establish a fault tracing database based on fault signals, fault causes and solutions;

[0032] Step S4: Establish a pre-training model based on the BERT large model and perform pre-training through the fault tracing database;

[0033] Step S5, establishing a fault tracing network, and connecting the fault signal of the downhole equipment to the fault tracing network;

[0034] Step S6: The fault tracing network is used as an input terminal to access the trained BERT large model;

[0035] Step S7: When the downhole equipment sends a fault signal, the BERT large model outputs the fault cause and solution accordingly.

[0036] Specifically, by analyzing the maintenance records and predicting possible failures, a fault tracing database is established through the data. The fault signal is used as input, and a pre-trained model is established based on the BERT big model. The model is pre-trained through the fault tracing database, and then a BERT big model for underground equipment fault tracing is established. After the fault tracing network receives the fault signal, it is analyzed through the BERT big model to output the cause of the fault and the solution.

[0037] In this way, by connecting the fault signal to the fault tracing network based on the BERT large model, after the fault signal is sent, the fault signal is extracted and analyzed by the BERT large model, and the corresponding fault cause and solution are output, thereby speeding up the analysis speed of the fault cause and the output speed of the solution, thereby reducing the time required for troubleshooting. The fault cause should at least include the fault location and the cause of occurrence, thereby improving the fault processing speed.

[0038] As a specific implementation of the method for tracing the source of downhole equipment faults based on a large language model provided in the application, the fault signal includes an independent signal and a combined signal when multiple independent signals occur synchronously;

[0039] The step S2 also includes combining independent signals, screening possible combined signals and predicting corresponding fault causes and solutions.

[0040] In general, fault signals may come from a single device or involve multiple devices. By combining different fault signals and predicting possible combined signals, the amount of data in the fault tracing database is increased, making it easier to deal with different fault situations.

[0041] Furthermore, the step S3 also includes sorting the fault signals according to the control levels of the relevant equipment;

[0042] The step S7 also includes, when multiple fault signals alarm and the combined signal database does not include this situation, analyzing the fault signals with a high control level first.

[0043] It should be understood that in power grids and equipment, equipment or power grids with higher control levels affect the operation of multiple devices. However, when multiple fault signals appear, on the one hand, the probability of failure of the upper-level equipment pointed to by multiple fault signals is higher. On the other hand, the impact caused by the failure of the upper-level equipment is greater. Prioritizing the analysis of the upper-level equipment will help reduce subsequent impacts.

[0044] As another specific implementation of the downhole equipment fault tracing method based on a large language model provided in the application, step S7 includes, when the fault signal includes multiple fault causes, the BERT large model reversely traces the fault cause to other fault signal sources that may cause the fault and detects them.

[0045] In combination with specific usage scenarios, a fault signal may correspond to multiple fault causes, and a fault cause may also correspond to multiple fault signals. However, when the fault cause is unclear, active detection is performed on other fault monitoring locations of the displayed fault cause to more clearly determine the main cause and facilitate faster troubleshooting.

[0046] Furthermore, the step S7 includes, based on the detection of other fault signal sources, sorting the fault causes according to the likelihood, and outputting the maintenance order according to the likelihood.

[0047] It should be understood that when there are multiple causes for a fault signal and the specific location cannot be determined through other fault monitoring location detection, the possibility of the fault cause is sorted and the location with the greatest possibility is repaired first, so as to facilitate faster troubleshooting. The possibility comparison can be performed through relevant values ​​such as the fluctuation amplitude of the sensor value.

[0048] As a specific implementation method of the underground equipment fault tracing method based on a large language model provided in the application, it also includes step S8, establishing an emergency address book for the emergency group and synchronizing it in the solution. When the solution is output, the fault tracing network emergency address book notifies the corresponding emergency group.

[0049] It should be understood that after tracing the fault and finding the fault location and cause, the fault needs to be eliminated. By setting up an emergency address book, the emergency team members responsible for the corresponding area are matched with the solutions. After finding the fault location and cause, the corresponding personnel are notified through the fault tracing network to reduce the communication time and improve the response speed.

[0050] Furthermore, the step S8 also includes that the emergency team uploads the fault log after troubleshooting, supplements the fault causes and solutions that are not involved, and provides feedback on the fault causes and solutions that are analyzed incorrectly.

[0051] It should be understood that by establishing feedback through fault logs, the fault causes and solutions added during the actual maintenance process are supplemented, and timely feedback is provided for fault signals that are incorrectly analyzed by the BERT large model, so that the BERT large model continues to grow and improves the accuracy of the BERT large model.

[0052] As a specific implementation of the method for tracing the source of a downhole equipment fault based on a large language model provided in the application, the fault signal includes a narrative signal formed by narrative speech or text;

[0053] The step S3 also includes extracting narration signal keywords and establishing a keyword database;

[0054] The step S7 includes extracting the keywords of the narrative signal by the BERT large model, comparing them with the keyword database, and outputting the fault cause and solution accordingly.

[0055] Combined with specific usage scenarios, fault signals include peripheral sensor signals, equipment fault signals, and manual inspection broadcast signals. When broadcast manually, they are usually voice clips or text clips. By establishing a keyword database, when broadcasting manually, the BERT large model is used to analyze the keywords, retrieve the keyword fragments, and output the cause of the fault and the solution.

[0056] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for tracing the source of underground equipment faults based on a large language model, characterized in that: include: Step S1, sorting out fault signals and their corresponding fault causes and solutions according to the maintenance ledger; Step S2: predicting possible fault signals and their corresponding fault causes and solutions; Step S3: Establish a fault tracing database based on fault signals, fault causes and solutions; Step S4: Establish a pre-training model based on the BERT large model and perform pre-training through the fault tracing database; Step S5, establishing a fault tracing network, and connecting the fault signal of the downhole equipment to the fault tracing network; Step S6: The fault tracing network is used as an input terminal to access the trained BERT large model; Step S7: When the downhole equipment sends a fault signal, the BERT large model outputs the fault cause and solution accordingly.

2. The method for tracing the source of underground equipment faults based on a large language model according to claim 1, characterized in that: Fault signals include independent signals and combined signals when multiple independent signals occur simultaneously; The step S2 also includes combining independent signals, screening possible combined signals and predicting corresponding fault causes and solutions.

3. The method for tracing the source of underground equipment faults based on a large language model according to claim 1, characterized in that: The step S7 includes, when the fault signal includes multiple fault causes, the BERT large model reversely traces the fault cause to other fault signal sources that may cause the fault and detects them.

4. The method for tracing the source of underground equipment faults based on a large language model according to claim 3 is characterized in that: The step S7 includes, based on the detection of other fault signal sources, sorting the fault causes according to the likelihood, and outputting the maintenance order according to the likelihood.

5. The method for tracing the source of underground equipment faults based on a large language model according to claim 1, characterized in that: The method further includes step S8, establishing an emergency contact list of the emergency group and synchronizing it in the solution. When outputting the solution, the fault tracing network notifies the corresponding emergency group through the emergency contact list.

6. The method for tracing the source of underground equipment faults based on a large language model according to claim 5 is characterized in that: The step S8 also includes that the emergency team uploads the fault log after troubleshooting, supplements the fault causes and solutions that are not involved, and provides feedback on the fault causes and solutions that are analyzed incorrectly.

7. The method for tracing the source of underground equipment faults based on a large language model according to claim 1, characterized in that: The fault signal includes a narrative signal formed by narrative voice or text; The step S3 also includes extracting narration signal keywords and establishing a keyword database; The step S7 includes extracting the keywords of the narrative signal by the BERT large model, comparing them with the keyword database, and outputting the fault cause and solution accordingly.

8. The method for tracing the source of underground equipment faults based on a large language model according to claim 2, characterized in that: The step S3 also includes sorting the fault signals according to the control levels of the relevant equipment; The step S7 also includes, when multiple fault signals alarm and the combined signal database does not include this situation, analyzing the signal with a high control level first.

9. The method for tracing the source of underground equipment faults based on a large language model according to any one of claims 1 to 8, characterized in that: The cause of the fault at least includes the fault location and the cause.