A gas turbine maintenance tool full life cycle intelligent supervision method and system

Abstract

This invention provides a method and system for intelligent monitoring of the entire lifecycle of gas turbine maintenance tools, belonging to the field of maintenance tool management technology. The method includes: acquiring historical full lifecycle data of gas turbine maintenance tools; determining the spatiotemporal compliance of the historical full lifecycle data; if the determination fails, training a spatiotemporal early warning model based on the corresponding historical full lifecycle data; if the determination passes, training a fault risk model based on the corresponding historical full lifecycle data; and performing risk analysis and monitoring of real-time local lifecycle data of gas turbine maintenance tools based on the spatiotemporal early warning model and the fault risk model. This invention's intelligent lifecycle monitoring method and system for gas turbine maintenance tools, in self-service retrieval and return scenarios, can promptly detect abnormal usage conditions such as overdue returns and unauthorized movement, and can also accurately predict tool status and promptly remind users to replace them, making monitoring more intelligent.

Description

Technical Field

[0001] This invention relates to the field of maintenance tool management technology, and in particular to a method and system for intelligent monitoring of the entire life cycle of gas turbine maintenance tools. Background Technology

[0002] Gas turbine maintenance is a crucial aspect of ensuring the normal operation of power plant production equipment. In existing technologies, the daily management of maintenance tools, a vital task at the maintenance site, relies primarily on manual registration, basic RFID tag tracking, or simple sensor monitoring systems. These methods typically focus on static inventory management or single-dimensional condition monitoring (such as usage count or battery level), failing to fully utilize tool lifecycle data (including historical records of manufacturing, use, maintenance, and disposal). This presents significant shortcomings: Firstly, it cannot effectively identify spatiotemporal anomalies, such as tools not being returned on time, unauthorized movement, or use in non-compliant environments (such as high temperature or high humidity), leading to frequent tool loss, damage, or safety incidents. Secondly, the prediction of inherent tool failure risks (such as material fatigue and accuracy drift) depends on empirical thresholds or periodic calibration, lacking dynamic modeling based on historical data. This results in tools being replaced too early (increasing costs) or too late (causing sudden failures), leading to unplanned gas turbine downtime, decreased maintenance efficiency, and accumulated safety risks.

[0003] In view of this, there is an urgent need for a method and system for intelligent monitoring of the entire life cycle of gas turbine maintenance tools, in order to at least address the above-mentioned shortcomings. Summary of the Invention

[0004] One objective of this invention is to provide a method and system for intelligent monitoring of the entire lifecycle of gas turbine maintenance tools. This method incorporates historical lifecycle data of gas turbine maintenance tools, performs spatiotemporal compliance checks on this data, uses historical lifecycle data that does not meet the compliance criteria to train a spatiotemporal early warning model, and uses historical lifecycle data that does meet the compliance criteria to train a fault risk model. The spatiotemporal early warning model and the fault risk model are then coordinated, and risk analysis and monitoring are performed based on real-time local lifecycle data of the gas turbine maintenance tools. In self-service retrieval and return scenarios for gas turbine maintenance tools, this invention can promptly detect abnormal usage such as overdue returns and unauthorized movement, and can also accurately predict the tool status and promptly remind users to replace them, making monitoring more intelligent.

[0005] This invention provides a method for intelligent monitoring of the entire lifecycle of gas turbine maintenance tools, comprising:

[0006] Step 1: Obtain historical full lifecycle data for various gas turbine maintenance tools;

[0007] Step 2: Determine the spatiotemporal compliance of historical full lifecycle data;

[0008] Step 3: If the judgment fails, train a spatiotemporal early warning model based on the corresponding historical full life cycle data;

[0009] Step 4: If the judgment is successful, train the fault risk model based on the corresponding historical full life cycle data;

[0010] Step 5: Based on the spatiotemporal early warning model and the fault risk model, conduct risk analysis on the real-time local life cycle data of gas turbine maintenance tools, identify risk items, and implement corresponding supervision;

[0011] Step 4: If the judgment is successful, train a fault risk model based on the corresponding historical full lifecycle data, including:

[0012] If the judgment is successful, extract the tool type, purpose of each use, duration of each use, and status after each use;

[0013] The fault risk model is trained by taking the type of tool, the purpose of each use, and the duration of each use as inputs to the second neural network model, and taking the state after each use as the output of the second neural network model.

[0014] Preferably, step 2: determine the spatiotemporal compliance of historical full lifecycle data, including:

[0015] The historical full life cycle data is unfolded on the full life cycle timeline of the corresponding gas turbine maintenance tool to determine the application information at multiple timeline points;

[0016] Traverse the timeline points and use the spatiotemporal point constraints corresponding to the application information of the timeline point being traversed as the standard spatiotemporal point constraints of the historical spatiotemporal points of the gas turbine maintenance tools obtained within the first target duration after the timeline point being traversed.

[0017] Determine whether historical spatiotemporal points meet the corresponding standard spatiotemporal point constraints;

[0018] If all conditions are met, the test is passed; otherwise, it is failed.

[0019] Preferably, step 3: If the judgment fails, train a spatiotemporal early warning model based on the corresponding historical full lifecycle data, including:

[0020] If the application fails, extract the application information, the historical spatiotemporal points within the first target time after the application information, and the first tool abnormal usage annotation information after the application information;

[0021] The application information and historical spatiotemporal locations within the first target time period after the application information are used as input to the first neural network model, and the tool abnormal use annotation information is used as output to train the spatiotemporal early warning model.

[0022] Preferably, the fault risk model is trained by using the tool type, purpose of each use, and duration of each use as inputs to the second neural network model, and the state after each use as the output of the second neural network model, including:

[0023] Based on the type of tool, determine the set of critical wear locations for the target tool;

[0024] Based on the relative positional relationship between each critical wear location in the critical wear location set and the target tool, each critical wear location is marked in the first tool model accordingly;

[0025] Construct a model point anchoring network based on the labeled location set;

[0026] After converting the output state of the second neural network model into the second tool model, the anchoring result is obtained using the model point anchoring network. The anchoring result is then compared with the preset standard anchoring result, and the output of the fault risk model is determined based on the comparison result.

[0027] Preferably, the fault risk model is trained by using the tool type, purpose of each use, and duration of each use as inputs to the second neural network model, and the state after each use as the output of the second neural network model. The training also includes:

[0028] During the process of reading the anchoring results based on the model point anchoring network, the standard wear morphology auxiliary verification relationship between different marked positions is used to determine whether the process anchoring results meet the triggering conditions of any standard wear morphology auxiliary verification relationship.

[0029] If satisfied, update the model point anchoring network according to the model point anchoring network update rules corresponding to the wear morphology auxiliary verification relationship of the corresponding standard.

[0030] Preferably, the standard wear morphology auxiliary verification relationship between different marked locations is derived from historical fault data statistics.

[0031] Preferably, step 5: Based on the spatiotemporal early warning model and the fault risk model, perform risk analysis on the real-time local lifecycle data of gas turbine maintenance tools, identify risk items, and implement corresponding supervision, including:

[0032] When a risky project involves violations of time and space regulations, the violating tools should be traced.

[0033] When a risk item is identified as a malfunction, locate the faulty tool and replace it.

[0034] Preferably, when the risk item is a spatiotemporal violation, the violation tool should be traced, including:

[0035] When a risky project involves a violation of spatiotemporal rules, the violation dimension must be determined.

[0036] When the violation is time-based, obtain the subsequent appointment information for the corresponding violation tool;

[0037] Based on subsequent appointment information, determine the time-based warning information for the relevant violators;

[0038] After issuing an early warning based on the warning information, the credit downgrade value of the violator is quantified according to the duration of the violation and the delay in subsequent appointments, and the credit downgrade is carried out accordingly.

[0039] When the violation dimension is spatial, obtain the sudden events within the second target duration of the violation space;

[0040] Based on the severity of the emergency, the matching results between the type of tool used in violation and the type of emergency in the preset emergency matching rule base, and the distance between the violation space and the standard space, the credit downgrade value of the violator is quantified and the corresponding credit downgrade is carried out.

[0041] This invention provides an intelligent monitoring system for the entire lifecycle of gas turbine maintenance tools, comprising:

[0042] The historical data acquisition module is used to acquire historical full lifecycle data of various gas turbine maintenance tools;

[0043] The spatiotemporal compliance determination module is used to determine the spatiotemporal compliance of historical full lifecycle data.

[0044] The first training module is used to train a spatiotemporal early warning model based on the corresponding historical full life cycle data if the judgment fails.

[0045] The second training module is used to train a fault risk model based on the corresponding historical full life cycle data if the judgment is passed.

[0046] The monitoring module is used to perform risk analysis on real-time local life cycle data of gas turbine maintenance tools based on spatiotemporal early warning models and fault risk models, identify risk items, and carry out corresponding monitoring.

[0047] The second training module performs the following operations:

[0048] If the judgment is successful, extract the tool type, purpose of each use, duration of each use, and status after each use;

[0049] The fault risk model is trained by taking the type of tool, the purpose of each use, and the duration of each use as inputs to the second neural network model, and taking the state after each use as the output of the second neural network model.

[0050] The beneficial effects of this invention are as follows:

[0051] This invention introduces historical full lifecycle data of gas turbine maintenance tools, performs spatiotemporal compliance assessment on this data, uses historical full lifecycle data that does not meet the spatiotemporal compliance assessment as training a spatiotemporal early warning model, and uses historical full lifecycle data that meets the spatiotemporal compliance assessment as training a fault risk model. The spatiotemporal early warning model and the fault risk model are coordinated, and risk analysis and supervision are performed based on real-time local lifecycle data of the gas turbine maintenance tools. In the scenario of self-service retrieval and return of gas turbine maintenance tools, this invention can promptly detect abnormal usage situations such as overdue returns and unauthorized movement, and can also accurately predict the tool status and promptly remind users to replace them, making supervision more intelligent.

[0052] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0053] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0054] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0055] Figure 1 This is a schematic diagram of an intelligent monitoring method for the entire life cycle of gas turbine maintenance tools according to an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram of a gas turbine maintenance tool full life cycle intelligent monitoring system according to an embodiment of the present invention. Detailed Implementation

[0057] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0058] This invention provides a method for intelligent monitoring of the entire lifecycle of gas turbine maintenance tools, such as... Figure 1 As shown, it includes:

[0059] Step 1: Obtain historical full life cycle data for various gas turbine maintenance tools.

[0060] In this embodiment, historical full life cycle data is obtained by connecting to the tool data management library corresponding to the tool container of the gas turbine maintenance tool; historical full life cycle data refers to the location, user, purpose, duration of use, and tool images before and after borrowing and returning of the gas turbine maintenance tool from the start of its use until it is removed from the tool data management library.

[0061] Step 2: Determine the spatiotemporal compliance of historical full lifecycle data. Specifically, Step 2 includes:

[0062] Step 21: Expand the historical full life cycle data on the full life cycle timeline of the corresponding gas turbine maintenance tool to determine the application information at multiple timeline points.

[0063] In this embodiment, the full lifecycle timeline is a timeline representing the full lifecycle time periods corresponding to historical full lifecycle data. The application information is the application and usage information for gas turbine maintenance tools, including the application period, application purpose, and application area.

[0064] Step 22: Traverse the timeline points and use the spatiotemporal point constraints corresponding to the application information of the timeline point being traversed as the standard spatiotemporal point constraints for the historical spatiotemporal points of the gas turbine maintenance tools obtained within the first target duration after the timeline point being traversed.

[0065] In this embodiment, the spatiotemporal location constraint is the standard that the gas turbine maintenance tool should adhere to in time and space. The first target duration is obtained by parsing the application information of the timeline point being traversed, and is the duration from the corresponding time of the timeline point being traversed to the end time of the corresponding application period. The historical spatiotemporal location is the actual spatiotemporal coordinate of the gas turbine maintenance tool obtained within the first target duration, for example: 2023-05-10 14:00, power plant Unit 3 maintenance area. The standard spatiotemporal location constraint is the spatiotemporal standard for comparison with the historical spatiotemporal location, that is, the spatiotemporal standard that the historical spatiotemporal location should adhere to, for example: constraining tool A to be geographically located in the power plant Unit 3 maintenance area from 14:00 to 15:00 on 2023-05-10.

[0066] Step 23: Determine whether the historical spatiotemporal points meet the corresponding standard spatiotemporal point constraints.

[0067] Step 24: If all conditions are met, the test is passed; otherwise, the test is failed.

[0068] Step 3: If the judgment fails, train a spatiotemporal early warning model based on the corresponding historical full life cycle data.

[0069] In this embodiment, if the historical spatiotemporal location does not meet the standard spatiotemporal location constraints, it indicates that the corresponding gas turbine maintenance tool was not used according to its intended purpose. If historical data of tools not used according to their intended purpose is used to train the failure risk model, it will lead to unreliable model predictions and inaccurate failure risk warnings (e.g., unreported use will cause the failure risk model to predict a shorter tool lifespan). Therefore, a spatiotemporal warning model is trained using this data. The spatiotemporal warning model is used to monitor the spatiotemporal location of the tool in real time and issue a warning when potential violations are detected (such as entering a prohibited area or exceeding the permitted usage time). Specifically, step 3 includes:

[0070] Step 31: If the application fails, extract the application information, the historical spatiotemporal points within the first target time after the application information, and the first tool abnormal usage annotation information after the application information.

[0071] In this embodiment, the abnormal tool usage annotation information is the abnormal usage information manually annotated in the historical full life cycle data that does not meet the spatiotemporal compliance judgment, such as: entering a prohibited area or exceeding the time limit for use.

[0072] Step 32: Use the application information and the historical spatiotemporal locations within the first target time period after the application information as input to the first neural network model, and use the tool abnormal use annotation information as output to train the spatiotemporal early warning model.

[0073] In this embodiment, the first neural network model is a preset CNN neural network.

[0074] Step 4: If the judgment is successful, train the fault risk model based on the corresponding historical full life cycle data.

[0075] In this embodiment, the failure risk model is a deep learning model used to analyze inherent failure factors (such as wear and tear, aging) from spatiotemporal compliance data. Specifically, step 4 includes:

[0076] Step 41: If the judgment is successful, extract the tool type, purpose of each use, duration of each use, and status after each use.

[0077] In this embodiment, the state after each use is the degree of tool wear and aging.

[0078] Step 42: Use the tool type, purpose of each use, and duration of each use as input to the second neural network model, and use the state after each use as output to train the failure risk model.

[0079] In this embodiment, the second neural network model is a preset CNN neural network.

[0080] Step 5: Based on the spatiotemporal early warning model and the fault risk model, conduct risk analysis on the real-time local life cycle data of gas turbine maintenance tools, identify risk items, and implement corresponding supervision.

[0081] In this embodiment, the real-time local lifecycle data includes the lifecycle data of the currently used tool (the tool's location, user, purpose, duration of use, and images of the tool before and after borrowing / returning). Risk analysis refers to analyzing spatiotemporal violations and malfunctions. Risk items are spatiotemporal violations (e.g., exceeding time limits, entering prohibited areas) and malfunctions (e.g., wear exceeding wear thresholds, remaining lifespan less than lifespan thresholds). During supervision, the corresponding supervision strategies are implemented based on the risk items; for example, tool tracing is performed for spatiotemporal violations, and tool replacement is performed for malfunctions.

[0082] The working principle and beneficial effects of the above technical solution are as follows:

[0083] This invention introduces historical full lifecycle data of gas turbine maintenance tools, performs spatiotemporal compliance assessment on this data, uses historical full lifecycle data that does not meet the spatiotemporal compliance assessment as training a spatiotemporal early warning model, and uses historical full lifecycle data that meets the spatiotemporal compliance assessment as training a fault risk model. The spatiotemporal early warning model and the fault risk model are coordinated, and risk analysis and supervision are performed based on real-time local lifecycle data of the gas turbine maintenance tools. In the scenario of self-service retrieval and return of gas turbine maintenance tools, this invention can promptly detect abnormal usage situations such as overdue returns and unauthorized movement, and can also accurately predict the tool status and promptly remind users to replace them, making supervision more intelligent.

[0084] In one embodiment, step 42: Training a failure risk model by taking the tool type, purpose of each use, and duration of each use as inputs to the second neural network model, and the state after each use as the output of the second neural network model, includes:

[0085] Step 421: Determine the set of critical wear locations for the target tool based on the tool type.

[0086] In this embodiment, the target tool is a gas turbine maintenance tool. The critical wear location set is the set of points or areas on the target tool that are most prone to wear or failure. For example, the critical wear location set of a clamp is: {jaw teeth tips, hydraulic cylinder piston rod connection hole, stress sensing plate attachment area}.

[0087] Step 422: Based on the relative positional relationship between each critical wear location in the critical wear location set and the target tool, mark each critical wear location in the first tool model.

[0088] In this embodiment, the first tool model is the three-dimensional model corresponding to the target tool.

[0089] Step 423: Construct the model point anchoring network based on the labeled location set.

[0090] In this embodiment, the model point anchoring network is a neural network for 3D feature point extraction, employing the PointNet++ architecture to automatically identify 3D regions corresponding to key wear locations on the tool model. This network receives 3D point cloud data of the tool as input, extracts features through a multilayer perceptron and max pooling operations, and outputs the 3D coordinates and confidence scores of the key wear locations. For example, it sequentially captures the coordinates of model regions corresponding to the anchor jaw tips, hydraulic cylinder piston rod connection holes, and stress-sensing sheet attachment areas within the tool model.

[0091] Step 424: After converting the output state of the second neural network model into the second tool model, the anchoring result is obtained using the model point anchoring network. The anchoring result is compared with the preset standard anchoring result, and the output of the fault risk model is determined based on the comparison result.

[0092] In this embodiment, the output state of the second neural network model is the tool usage state predicted by the model. The second tool model is the 3D model corresponding to the tool in the predicted tool usage state. The anchoring result is the result obtained after the anchoring point grasping logic of the model point anchoring network grasps the model points of the second tool model. For example, the model areas corresponding to the jaw teeth, hydraulic cylinder piston rod connection hole, and stress sensing sheet attachment area in the tool model of the grasped model predicting the clamp usage state. The preset standard anchoring result is: the ideal tool state benchmark, for example: the jaw teeth wear depth and connection hole crack length within the preset safety threshold range. The output of the fault risk model is determined according to the comparison results: if the comparison results show that all anchoring results are within the safety threshold range, then no fault is output; if any anchoring result is not within the safety threshold range, then a fault is output.

[0093] Step 425: During the process of reading the anchoring results based on the model point anchoring network, determine whether the process anchoring results meet the triggering conditions of any standard wear morphology auxiliary verification relationship based on the standard wear morphology auxiliary verification relationship between different marked positions.

[0094] In this embodiment, the standard wear morphology auxiliary verification relationship is the wear association rule for different tool locations. For example, based on historical fault data statistics, if the wear depth of the jaw teeth tip is >0.4mm, then the piston rod connecting hole will inevitably have radial cracks ≥1mm. Therefore, the wear depth of the jaw teeth tip >0.4mm and the radial cracks ≥1mm in the piston rod connecting hole constitute a pair of standard wear morphology auxiliary verification relationships. The process anchoring result is the anchoring result identified during the process of reading all anchoring results according to the model point anchoring network. For example, during the process of sequentially identifying the model areas of the jaw teeth tip, the hydraulic cylinder piston rod connecting hole, and the stress-sensing sheet attachment area, the process anchoring result can be a jaw teeth tip wear depth of 0.5mm. The triggering condition for the standard wear morphology-assisted verification relationship is: the process anchoring result matches any wear morphology that has a standard wear morphology-assisted verification relationship. For example, if the process anchoring result satisfies that the jaw tooth tip wear depth is >0.4mm, then the triggering condition for the standard wear morphology-assisted verification relationship is satisfied, which consists of the jaw tooth tip wear depth >0.4mm and the appearance of radial cracks ≥1mm in the piston rod connecting hole.

[0095] Step 426: If satisfied, update the model point anchoring network according to the model point anchoring network update rule corresponding to the standard wear morphology auxiliary verification relationship.

[0096] In this embodiment, the model point anchoring network update rule is: the self-optimization rule of the model point anchoring strategy, for example: optimizing the model point anchoring process of the anchoring piston rod connection hole so that the anchoring strategy is suitable for the anchoring of radial cracks greater than or equal to 1 mm.

[0097] The working principle and beneficial effects of the above technical solution are as follows:

[0098] Different types of tools experience wear on different parts during use. After a fault detection model predicts a tool's condition, it's necessary to further confirm whether it can still be used. However, directly outputting the current wear amount of key parts of different tools is not accurate enough (for example, unclear boundaries between parts may cause wear belonging to part A to be counted as part B; or a crack belonging to part A may radiate to part B, and the system can only determine the degree of fault based on the local crack in part A, which is not accurate enough). Therefore, detailed identification of wear parts is required. Specifically, this invention introduces a set of key wear locations for the target tool of a tool type, and marks each key wear location in the first tool model. Based on the marked location set, a model point anchoring network is constructed. The model point anchoring network is used to identify the second tool model output by the second neural network model to obtain the anchoring result. During the identification process, based on the standard wear morphology auxiliary verification relationship between different marked locations, it is determined whether the process anchoring result meets the triggering condition of any standard wear morphology auxiliary verification relationship. If so, the model point anchoring network is updated according to the update rule of the model point anchoring network corresponding to the standard wear morphology auxiliary verification relationship. By comparing all identified anchoring results with standard anchoring results, and determining the output of the fault risk model based on the comparison results, detailed identification of wear areas is achieved, making it more intelligent.

[0099] In one embodiment, step 5: Based on the spatiotemporal early warning model and the fault risk model, perform risk analysis on the real-time local lifecycle data of gas turbine maintenance tools, identify risk items, and implement corresponding supervision, including:

[0100] Step 51: When a risk item involves a spatiotemporal violation, trace the violating tools. Specifically, Step 51 includes:

[0101] Step 511: When the risk item is a spatiotemporal violation, determine the violation dimension.

[0102] In this embodiment, the violation dimension refers to the specific type of spatiotemporal violation, including the time dimension and the space dimension.

[0103] Step 512: When the violation dimension is time, obtain the subsequent appointment information of the corresponding violation tool.

[0104] In this embodiment, the subsequent appointment information is: the usage plan that the violation tool has been arranged after the violation occurred, including the appointment time, task type and associated personnel.

[0105] Step 513: Based on subsequent appointment information, determine the warning information for the time of the corresponding violation.

[0106] In this embodiment, the time alert is a real-time warning for time violations, reminding personnel of potential consequences. The alert information is a notification containing specific time warning details (including specific impacts and action suggestions).

[0107] Step 514: After issuing a warning based on the warning information, quantify the credit downgrade value of the violator according to the duration of the violation and the delay in subsequent appointments, and make the corresponding credit downgrade.

[0108] In this embodiment, the time violation duration is defined as the duration of the violation (e.g., the number of minutes exceeding the time limit). The delay in subsequent appointments is defined as the actual delay in subsequent tasks caused by the violation (e.g., the number of minutes) and the task's importance weight. When quantifying the credit downgrade value of violators based on the time violation duration and the delay in subsequent appointments, the longer the violation duration and the larger the product of the actual delay in subsequent tasks and the corresponding task importance weight, the larger the corresponding quantified credit downgrade value. The specific quantification ratio is preset by staff.

[0109] Step 515: When the violation dimension is space, obtain the sudden events within the second target duration of the violation space.

[0110] In this embodiment, the second target duration is the time between the determination of a spatiotemporal violation and the start of the violating tool's current use. The violation space is the physical location where the tool is used improperly. The unexpected event is an accidental event occurring within the violation space, such as a safety incident or equipment malfunction.

[0111] Step 516: Based on the severity of the emergency, the matching results between the type of the violation tool and the type of the emergency in the preset emergency matching rule library, and the distance between the violation space and the standard space, quantify the credit downgrade value of the violator and make the corresponding credit downgrade.

[0112] In this embodiment, the severity of the emergency is a quantified result of the emergency's severity. The matching result between the tool type of the prohibited tool and the emergency type in the preset emergency matching rule base includes match and non-match. For example, if the tool type is a fire extinguisher and the emergency type is a fire, the matching result is a match. The preset emergency matching rule base is exemplarily shown in Table 1:

[0113] Table 1. Example table of emergency matching rule base

[0114]

[0115] The regulated space is the space where the violating tool should be used. The distance between the violating space and the regulated space is defined as the distance between the center points of the two spaces. When quantifying the credit downgrade value of violators based on severity, matching results, and distance, in the absence of unforeseen events, the greater the distance, the larger the credit downgrade value; in the presence of unforeseen events and when the tool is irrelevant (matching result is non-matching), the greater the distance, the larger the credit downgrade value; in the presence of unforeseen events and when the tool is relevant (matching result is matching), the lower the severity and the greater the distance, the larger the credit downgrade value. The specific quantification ratio is preset by the staff.

[0116] Step 52: When the risk item is a fault, locate the faulty tool and replace it.

[0117] The working principle and beneficial effects of the above technical solution are as follows:

[0118] This invention classifies risky projects for appropriate supervision. When a risky project involves a time-space violation, it is handled according to the violation dimension. Specifically, when the violation dimension is time, subsequent appointment information of the corresponding violation tool is introduced. A time-based warning is issued based on this appointment information. After the warning, the credit downgrade value of the violator is quantified based on the duration of the time violation and the delay in subsequent appointments. In the quantification process, the longer the violation duration and the larger the product of the actual delay in subsequent tasks and the corresponding task importance weight, the larger the corresponding quantified credit downgrade value. When the violation dimension is spatial, an emergency event within the second target time period of the violation space is introduced. Based on the severity of the emergency event, the matching result between the type of the violating tool and the type of the emergency event in the preset emergency matching rule base, and the distance between the violation space and the regulated space, the credit downgrade value of the violating personnel is quantified. During quantification, in the absence of an emergency event, the greater the distance, the larger the credit downgrade value; in the presence of an emergency event but the tool is unrelated (matching result is non-matching), the greater the distance, the larger the credit downgrade value; in the presence of an emergency event but the tool is related (matching result is matching), the lower the severity and the greater the distance, the larger the credit downgrade value. Credit quantification factors for corresponding tool users are extracted according to different spatiotemporal violation scenarios, and adaptive credit adjustments are made by integrating the gas turbine maintenance scenario. This avoids unreasonable credit penalties for tool users when tools temporarily support other emergency operations, making it more intelligent. When the risk item is a malfunction, the location of the malfunctioning tool is located and the tool is replaced. This improves the comprehensiveness and suitability of maintenance tool risk supervision.

[0119] This invention provides an intelligent monitoring system for the entire lifecycle of gas turbine maintenance tools, such as... Figure 2 As shown, it includes:

[0120] Historical data acquisition module 1 is used to acquire historical full life cycle data of various gas turbine maintenance tools;

[0121] Spatiotemporal compliance determination module 2 is used to determine the spatiotemporal compliance of historical full lifecycle data;

[0122] The first training module 3 is used to train a spatiotemporal early warning model based on the corresponding historical full life cycle data if the judgment fails.

[0123] The second training module 4 is used to train a fault risk model based on the corresponding historical full life cycle data if the judgment passes.

[0124] The supervision module 5 is used to perform risk analysis on the real-time local life cycle data of gas turbine maintenance tools based on the spatiotemporal early warning model and the fault risk model, identify risk items and carry out corresponding supervision.

[0125] The second training module performs the following operations:

[0126] If the judgment is successful, extract the tool type, purpose of each use, duration of each use, and status after each use;

[0127] Based on the type of tool, determine the set of critical wear locations for the target tool;

[0128] Based on the relative positional relationship between each critical wear location in the critical wear location set and the target tool, each critical wear location is marked in the first tool model accordingly;

[0129] Construct a model point anchoring network based on the labeled location set;

[0130] After converting the output state of the second neural network model into the second tool model, the anchoring result is obtained by using the model point anchoring network. The anchoring result is compared with the preset standard anchoring result, and the output of the fault risk model is determined based on the comparison result.

[0131] During the process of reading the anchoring results based on the model point anchoring network, the standard wear morphology auxiliary verification relationship between different marked positions is used to determine whether the process anchoring results meet the triggering conditions of any standard wear morphology auxiliary verification relationship.

[0132] If satisfied, update the model point anchoring network according to the model point anchoring network update rules corresponding to the wear morphology auxiliary verification relationship of the corresponding standard.

[0133] Among them, the standard wear morphology auxiliary verification relationship between different marked locations was obtained based on historical fault data statistics;

[0134] The monitoring module performs the following operations:

[0135] When a risky project involves violations of time and space regulations, the violating tools should be traced.

[0136] When the risk item is a fault, locate the faulty tool and replace it.

[0137] When a risk item involves a violation of spatiotemporal regulations, the relevant tools used to commit the violation will be traced, including:

[0138] When a risky project involves a violation of spatiotemporal rules, the violation dimension must be determined.

[0139] When the violation is time-based, obtain the subsequent appointment information for the corresponding violation tool;

[0140] Based on subsequent appointment information, determine the time-based warning information for the relevant violators;

[0141] After issuing an early warning based on the warning information, the credit downgrade value of the violator is quantified according to the duration of the violation and the delay in subsequent appointments, and the credit downgrade is carried out accordingly.

[0142] When the violation dimension is spatial, obtain the sudden events within the second target duration of the violation space;

[0143] Based on the severity of the emergency, the matching results between the type of tool used in violation and the type of emergency in the preset emergency matching rule base, and the distance between the violation space and the standard space, the credit downgrade value of the violator is quantified and the corresponding credit downgrade is carried out.

[0144] 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 method for intelligent monitoring of the entire lifecycle of gas turbine maintenance tools, characterized in that, include: Step 1: Obtain historical full lifecycle data for various gas turbine maintenance tools; Step 2: Determine the spatiotemporal compliance of historical full lifecycle data; Step 3: If the judgment fails, train a spatiotemporal early warning model based on the corresponding historical full life cycle data; Step 4: If the judgment is successful, train the fault risk model based on the corresponding historical full life cycle data; Step 5: Based on the spatiotemporal early warning model and the fault risk model, conduct risk analysis on the real-time local life cycle data of gas turbine maintenance tools, identify risk items, and implement corresponding supervision; Step 4: If the judgment is successful, train a fault risk model based on the corresponding historical full lifecycle data, including: If the judgment is successful, extract the tool type, purpose of each use, duration of each use, and status after each use; The fault risk model is trained by using the tool type, purpose of each use, and duration of each use as inputs to the second neural network model, and the state after each use as the output of the second neural network model, including: Based on the type of tool, determine the set of critical wear locations for the target tool; Based on the relative positional relationship between each critical wear location in the critical wear location set and the target tool, each critical wear location is marked in the first tool model accordingly; Construct a model point anchoring network based on the labeled location set; After converting the output state of the second neural network model into the second tool model, the anchoring result is obtained by using the model point anchoring network. The anchoring result is compared with the preset standard anchoring result, and the output of the fault risk model is determined based on the comparison result. During the process of reading the anchoring results based on the model point anchoring network, the standard wear morphology auxiliary verification relationship between different marked positions is used to determine whether the process anchoring results meet the triggering conditions of any standard wear morphology auxiliary verification relationship. If satisfied, update the model point anchoring network according to the model point anchoring network update rules corresponding to the wear morphology auxiliary verification relationship of the corresponding standard. The standard wear morphology auxiliary verification relationship between different marked locations was derived from historical fault data statistics.

2. The intelligent monitoring method for the entire life cycle of gas turbine maintenance tools as described in claim 1, characterized in that, Step 2: Determine the spatiotemporal compliance of historical full lifecycle data, including: The historical full life cycle data is unfolded on the full life cycle timeline of the corresponding gas turbine maintenance tool to determine the application information at multiple timeline points; Traverse the timeline points and use the spatiotemporal point constraints corresponding to the application information of the timeline point being traversed as the standard spatiotemporal point constraints of the historical spatiotemporal points of the gas turbine maintenance tools obtained within the first target duration after the timeline point being traversed. Determine whether historical spatiotemporal points meet the corresponding standard spatiotemporal point constraints; If all conditions are met, the test is passed; otherwise, it is failed.

3. The intelligent monitoring method for the entire life cycle of gas turbine maintenance tools as described in claim 1, characterized in that, Step 3: If the judgment fails, train a spatiotemporal early warning model based on the corresponding historical full lifecycle data, including: If the application fails, extract the application information, the historical spatiotemporal points within the first target time after the application information, and the first tool abnormal usage annotation information after the application information; The application information and historical spatiotemporal locations within the first target time period after the application information are used as input to the first neural network model, and the tool abnormal use annotation information is used as output to train the spatiotemporal early warning model.

4. The intelligent monitoring method for the entire life cycle of gas turbine maintenance tools as described in claim 1, characterized in that, Step 5: Based on the spatiotemporal early warning model and the fault risk model, conduct risk analysis on the real-time local lifecycle data of gas turbine maintenance tools, identify risk items, and implement corresponding supervision, including: When a risky project involves violations of time and space regulations, the violating tools should be traced. When a risk item is identified as a malfunction, locate the faulty tool and replace it.

5. The intelligent monitoring method for the entire life cycle of gas turbine maintenance tools as described in claim 4, characterized in that, When a risky project involves a violation of spatiotemporal regulations, the violation tools will be traced, including: When a risky project involves a violation of spatiotemporal rules, the violation dimension must be determined. When the violation is time-based, obtain the subsequent appointment information for the corresponding violation tool; Based on subsequent appointment information, determine the time-based warning information for the relevant violators; After issuing an early warning based on the warning information, the credit downgrade value of the violator is quantified according to the duration of the violation and the delay in subsequent appointments, and the credit downgrade is carried out accordingly. When the violation dimension is spatial, obtain the sudden events within the second target duration of the violation space; Based on the severity of the emergency, the matching results between the type of tool used in violation and the type of emergency in the preset emergency matching rule base, and the distance between the violation space and the standard space, the credit downgrade value of the violator is quantified and the corresponding credit downgrade is carried out.

6. A smart monitoring system for the entire lifecycle of gas turbine maintenance tools, characterized in that, include: The historical data acquisition module is used to acquire historical full lifecycle data of various gas turbine maintenance tools; The spatiotemporal compliance determination module is used to determine the spatiotemporal compliance of historical full lifecycle data. The first training module is used to train a spatiotemporal early warning model based on the corresponding historical full life cycle data if the judgment fails. The second training module is used to train a fault risk model based on the corresponding historical full life cycle data if the judgment is passed. The monitoring module is used to perform risk analysis on real-time local life cycle data of gas turbine maintenance tools based on spatiotemporal early warning models and fault risk models, identify risk items, and carry out corresponding monitoring. The second training module performs the following operations: If the judgment is successful, extract the tool type, purpose of each use, duration of each use, and status after each use; The fault risk model is trained by using the tool type, purpose of each use, and duration of each use as inputs to the second neural network model, and the state after each use as the output of the second neural network model, including: Based on the type of tool, determine the set of critical wear locations for the target tool; Based on the relative positional relationship between each critical wear location in the critical wear location set and the target tool, each critical wear location is marked in the first tool model accordingly; Construct a model point anchoring network based on the labeled location set; After converting the output state of the second neural network model into the second tool model, the anchoring result is obtained by using the model point anchoring network. The anchoring result is compared with the preset standard anchoring result, and the output of the fault risk model is determined based on the comparison result. During the process of reading the anchoring results based on the model point anchoring network, the standard wear morphology auxiliary verification relationship between different marked positions is used to determine whether the process anchoring results meet the triggering conditions of any standard wear morphology auxiliary verification relationship. If satisfied, update the model point anchoring network according to the model point anchoring network update rules corresponding to the wear morphology auxiliary verification relationship of the corresponding standard. The standard wear morphology auxiliary verification relationship between different marked locations was derived from historical fault data statistics.

Images