Fault detection and early warning method and system in drone inspection

Abstract

The present invention is applicable to the field of drone inspection, and provides a method and system for fault detection and early warning in drone inspection. The system includes: a data collection and feature extraction module. The method realizes intelligent planning of inspection paths by integrating multi-objective optimization, adaptive adjustment and innovative algorithms for priority of key areas, thereby significantly improving the efficiency, flexibility and reliability of the entire system. This method not only improves the efficiency and coverage of inspections, but also enhances the system's ability to identify and warn of potential faults, while optimizing resource allocation and reducing operation and maintenance costs. These improvements have jointly promoted the development of power equipment operation and maintenance in a more intelligent, accurate and proactive direction, and have made important contributions to improving the safety and reliability of power grids. Through this innovative method, the drone inspection system can monitor the equipment status more comprehensively and timely, effectively prevent faults from occurring, and ultimately achieve safe, stable and efficient operation of the power system.

Description

Technical Field

[0001] The present invention belongs to the field of unmanned aerial vehicle inspection, and in particular relates to a fault detection and early warning method and system in unmanned aerial vehicle inspection. Background Art

[0002] Fault inspection using drones is a modern equipment monitoring and maintenance method, which is widely used in industries such as electricity, petroleum, and transportation. With its flexibility and efficiency, drones can complete large-scale inspection tasks in a short time. Equipped with sensors such as high-resolution cameras, infrared thermal imagers, and lidar, they can collect equipment operating status and environmental data in real time, significantly improving inspection efficiency.

[0003] The advantages of drone fault inspections are that they improve efficiency, reduce manual work in dangerous areas, reduce safety risks, save manpower and material costs, and reduce equipment downtime. In the future, with the continuous advancement of technology, drone inspections will gradually integrate more advanced artificial intelligence and big data analysis capabilities to improve the accuracy and real-time performance of fault detection, and are expected to be more widely used in various industries. Overall, drone fault inspections are an efficient, safe, and economical monitoring method that provides a strong guarantee for the safe operation of various industries.

[0004] Traditional inspection paths are often preset fixed routes that lack flexibility and use the same inspection frequency for all areas, resulting in waste of resources or insufficient attention to high-risk areas. They fail to pay sufficient attention to high-risk or fault-prone areas, and inspection paths may also contain duplicate or unnecessary inspections, reducing overall efficiency. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for fault detection and early warning in drone inspection, aiming to solve the technical problems existing in the prior art identified in the background technology.

[0006] The present invention is implemented as follows: a method for fault detection and early warning in drone inspection, the method comprising:

[0007] Collect fault records, operating parameters and environmental status data of existing equipment, extract data features and form feature values;

[0008] Establish a frequency model based on the historical and current status data of the equipment, adjust the detection frequency of the equipment and synchronize it to the drone inspection strategy;

[0009] Collect and annotate historical inspection data and establish a fault symptom database;

[0010] Establish a fault identification model to identify the relationship between fault categories and characteristic manifestations;

[0011] Obtain the current equipment status data obtained by real-time monitoring by drones, determine whether the equipment has potential faults, calculate the detection frequency of the equipment through the frequency model, and optimize and adjust the inspection path of the drone;

[0012] The real-time monitoring data collected by the drone from the marked area is analyzed again, and a threshold is set to determine the abnormality of the real-time monitoring data. If there is an actual abnormality, an early warning signal is issued;

[0013] Collect on-site feedback data, evaluate the inspection effect and the accuracy of the early warning, set inspection effect indicators, and if no feedback is received, judge the effectiveness of the inspection and whether the effect indicators are achieved, and analyze whether the inspection strategy needs to be adjusted based on the judgment results.

[0014] As a further solution of the present invention, the frequency model is established based on the historical and current status data of the equipment, the detection frequency of the equipment is adjusted and synchronized to the drone inspection strategy, specifically including:

[0015] Analyze the historical detection data of the equipment under different working conditions, identify the impact of different working conditions on the operating status of the equipment, and evaluate the current operating status of the equipment based on the data collected by the drone to determine what working condition it is in;

[0016] A frequency model is established based on the historical detection data of the equipment. Based on the rule that equipment with high-risk working conditions and high failure rates requires high-frequency detection, a corresponding detection frequency weight is assigned to each working condition, and the optimal detection frequency under different working conditions is calculated;

[0017] The detection frequency of the target area is dynamically adjusted according to the usage frequency and status of the equipment, and the initial inspection route of the drone is formulated based on the detection frequency.

[0018] As a further solution of the present invention, the collection and marking of historical inspection data and establishment of a fault symptom database specifically include:

[0019] Collect all fault data and normal operation data recorded in previous drone inspections, determine the fault category based on the fault records, and annotate each piece of fault data and record its corresponding characteristic performance features;

[0020] Analyze the characteristic manifestations of different fault categories, identify common features and abnormal features, and establish a fault symptom database to store all the acquired data.

[0021] As a further solution of the present invention, the establishment of a fault identification model to identify the relationship between the fault category and the characteristic performance specifically includes:

[0022] Analyze the correlation between the characteristic values ​​and the fault categories, and select the fault-related characteristic values ​​based on the correlation;

[0023] Establish a fault identification model and train the model through the collected historical inspection data, define the rule logic of fault identification, and identify the relationship between fault category and characteristic performance.

[0024] As a further solution of the present invention, the determination of whether a device has a potential fault and calculating the detection frequency of the device through a frequency model specifically include:

[0025] Collect current equipment status data from the sensors onboard the drone in real time, pre-process the data, and extract feature values ​​related to fault identification;

[0026] Load the trained fault identification model and input the characteristic value to identify and analyze the current fault;

[0027] Obtain the equipment failure probability and failure category output by the fault identification model, evaluate the severity of the failure, classify the existing failures into failure levels, and mark the area where the equipment exists according to the failure level;

[0028] According to the severity of the fault, the frequency model is used to calculate the detection frequency required for the area. Specifically:

[0029] ;

[0030] in, Indicates at time The required detection frequency is in times / time unit; represents the initial detection frequency; Represents the attenuation constant, the unit is: 1 / time unit; A numerical representation indicating the severity of the fault;

[0031] According to the calculation results of the fault degree and frequency model, the patrol mode is defined and the UAV inspection path is calculated.

[0032] As a further solution of the present invention, the method of calculating the inspection path of the drone specifically includes:

[0033] Get the initial inspection path of the drone , and calculate the fitness value :

[0034] ;

[0035] in:

[0036] Indicates the total distance of the initial inspection path;

[0037] ;

[0038] Indicates the total inspection time required for the initial inspection path;

[0039] ;

[0040] represents the penalty term for satisfying the frequency demand;

[0041] ;

[0042] in, Indicates the number of areas to be detected; Indicates the required inspection frequency for each area; Indicates the actual detection frequency of drones in each area;

[0043] Read the area where the annotation exists , select the path to The designated position for additional detection, the designated position means: After the detection of the area outside is completed, return again ;

[0044] Traverse the current drone initial inspection path , and read the detection frequency, based on the detection frequency and the initial inspection path of the drone The total number of areas in the initial inspection path Divide equally and insert the pair at each dividing point The inspection command is used to obtain the new inspection path of the drone. ;

[0045] Re-evaluate the fitness value of the new path :

[0046] ;

[0047] Compare and ,like , then accept the new path .

[0048] As a further solution of the present invention, the real-time monitoring data collected by the drone from the marked area is analyzed again, and a threshold is set to determine the abnormality of the real-time monitoring data, specifically including:

[0049] Obtain real-time monitoring data of the marked area obtained by the drone based on the new route inspection;

[0050] Set the threshold of monitoring data according to the fault severity of the marked area, compare the monitoring data with the set threshold, and determine whether there is anomaly that deviates from the fault stability;

[0051] Generate early warning signals for deviations from the failure stable region.

[0052] Another object of the present invention is to provide a fault detection and early warning system in drone inspection, the system comprising:

[0053] The data collection and feature extraction module is used to collect the fault records, operating parameters and environmental status data of existing equipment, and extract data features to form feature values;

[0054] The frequency model and equipment grouping module is used to establish a frequency model based on the historical detection and current status of different equipment working conditions, and adaptively adjust the detection frequency of the target area according to the uneven use of equipment;

[0055] A fault symptom database establishment module is used to collect historical inspection data, including fault data and normal operation data, annotate the collected data, identify the fault categories appearing in the data and their corresponding characteristic performance characteristics, and establish a fault symptom database based on the characteristic data;

[0056] The fault identification model training module is used to establish a fault identification model, train the fault identification model through the collected historical inspection data, and identify the relationship between the fault category and the characteristic performance;

[0057] The real-time monitoring and path optimization module is used to obtain the current equipment status data obtained by the drone sensor in real-time monitoring, analyze the current equipment status data using the fault identification model, and determine whether the equipment has potential faults. If so, the area is marked according to the degree of fault, and a patrol mode is defined for the marked area. The inspection path of the drone is adjusted using path optimization;

[0058] The anomaly detection and warning module is used to judge the real-time monitoring data collected by the drone from the marked area again, set a threshold to judge the abnormality of the real-time monitoring data, and issue a warning signal if there is an actual abnormality;

[0059] The feedback evaluation and strategy adjustment module is used to collect on-site feedback data, evaluate the inspection effect and the accuracy of the early warning, set the inspection effect indicators, and if no feedback is received, judge the effectiveness of the inspection and whether the effect indicators are achieved, and analyze whether the inspection strategy needs to be adjusted based on the judgment results.

[0060] As a further solution of the present invention, the fault identification model training module includes:

[0061] The characteristic value correlation analysis unit is used to analyze the correlation between the characteristic value and the fault category, and select the fault-related characteristic value according to the correlation;

[0062] The fault identification model training unit is used to establish a fault identification model and train the model through the collected historical inspection data, define the rule logic of fault identification, and identify the relationship between fault category and characteristic performance.

[0063] As a further solution of the present invention, the real-time monitoring and path optimization module includes:

[0064] The real-time data collection and preprocessing unit is used to collect the current equipment status data from the sensors carried by the drone in real time, preprocess the data, and extract the characteristic values ​​related to fault identification;

[0065] The fault identification model loading and feature value input unit is used to load the trained fault identification model and input feature values ​​to identify and analyze the current fault;

[0066] A fault probability and category evaluation unit is used to obtain the fault probability and fault category of the equipment output by the fault identification model, evaluate the severity of the fault, classify the existing faults into fault levels, and mark the area where the equipment exists according to the fault level;

[0067] A detection frequency estimation unit is used to estimate the detection frequency required for the area using a frequency model according to the severity of the fault;

[0068] The patrol mode definition and path optimization unit is used to define the patrol mode and calculate the UAV inspection path according to the calculation results of the fault degree and frequency model.

[0069] The beneficial effects of the present invention are:

[0070] This method can achieve the best balance between inspection distance, time and frequency requirements, which not only improves resource utilization efficiency, but also ensures the overall improvement of inspection quality. The adaptive adjustment mechanism gives the system the ability to dynamically optimize inspection strategies according to real-time conditions, making the entire fault detection and early warning process more agile and accurate. The design concept of key area priority further enhances the system's monitoring of high-risk areas and greatly improves the early detection and early warning capabilities of potential faults.

[0071] Not only can routine inspection tasks be completed more efficiently, but also the monitoring of key areas can be strengthened in a targeted manner, thereby improving the accuracy and timeliness of fault detection. At the same time, the efficient design of the algorithm ensures that the system can respond quickly in a real-time environment and adapt to complex and changing inspection scenarios. In addition, the scalability of this step provides a solid foundation for the continuous optimization of the system, allowing the entire method to continuously adapt to new technological developments and inspection needs.

[0072] This method not only improves the efficiency and coverage of inspections, but also enhances the system's ability to identify and warn of potential faults, while optimizing resource allocation and reducing operation and maintenance costs. These improvements have jointly promoted the development of power equipment operation and maintenance in a more intelligent, precise and proactive direction, making an important contribution to improving the safety and reliability of the power grid. Through this innovative method, the drone inspection system can monitor the equipment status more comprehensively and timely, effectively prevent faults from occurring, and ultimately achieve safe, stable and efficient operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 A flow chart of a method for fault detection and early warning in drone inspection provided by an embodiment of the present invention;

[0074] Figure 2 A flowchart of establishing a frequency model based on the historical and current status data of the device, adjusting the detection frequency of the device and synchronizing it to the drone inspection strategy provided by an embodiment of the present invention;

[0075] Figure 3 A flowchart for collecting and marking historical inspection data and establishing a fault symptom database provided by an embodiment of the present invention;

[0076] Figure 4 A flowchart of establishing a fault identification model and identifying the relationship between fault categories and characteristic manifestations provided by an embodiment of the present invention;

[0077] Figure 5 A flowchart for determining whether a device has a potential fault and calculating the detection frequency of the device through a frequency model provided by an embodiment of the present invention;

[0078] Figure 6 A flowchart of an embodiment of the present invention for analyzing the real-time monitoring data collected by the drone from the marked area again and setting a threshold to determine the abnormality of the real-time monitoring data;

[0079] Figure 7 A structural block diagram of a fault detection and early warning system for drone inspection provided by an embodiment of the present invention;

[0080] Figure 8 A structural block diagram of a fault identification model training module provided in an embodiment of the present invention;

[0081] Fig. 9 A structural block diagram of the real-time monitoring and path optimization module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0082] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0083] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0084] Figure 1 A flowchart of a method for fault detection and early warning in drone inspection provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method includes:

[0085] S100, collecting fault records, operating parameters and environmental status data of existing equipment, and extracting data features to form feature values;

[0086] S200, establishes a frequency model based on the historical and current status data of the device, adjusts the detection frequency of the device and synchronizes it to the drone inspection strategy;

[0087] By collecting and analyzing the historical detection data of the equipment under various working conditions, using machine learning algorithms such as cluster analysis and decision trees to identify the characteristics of different working conditions, establishing a working condition-equipment status impact model, and quantifying the degree of influence of various working conditions on equipment parameters. Subsequently, based on the data collected by the drone in real time, combined with the working condition identification model, the specific working condition of the equipment is evaluated. On this basis, the relationship between the frequency of failures and working conditions in the historical detection data is analyzed, and a risk-based frequency allocation model is established to allocate higher detection frequencies to high-risk working conditions and high-failure rate equipment. Using methods such as time series analysis, an optimal detection frequency prediction model is established for each working condition, and optimization is performed considering factors such as equipment importance and maintenance cost. In addition, by real-time monitoring of the equipment usage frequency, analyzing the relationship between usage intensity and detection requirements, an equipment status evaluation model is established, and the detection frequency is dynamically adjusted according to the state change trend. At the same time, the frequency is fine-tuned considering the impact of environmental factors such as season and weather on the equipment status, and the upper and lower limits of adjustment are set to avoid excessive or insufficient detection. Finally, based on the adjusted detection frequency and combined with geographic information system data, heuristic algorithms such as ant colony algorithm and genetic algorithm are applied to optimize the inspection path. At the same time, constraints such as drone endurance and charging station distribution are considered to formulate the initial inspection path and reserve space for dynamic adjustment.

[0088] This step refines the identification of multiple working conditions and quantifies their impact, introduces risk assessment into the frequency model, and achieves optimal allocation of resources. At the same time, multiple factors such as equipment usage intensity, state changes, and environmental factors are considered to achieve real-time optimization of detection frequency. By combining GIS and intelligent algorithms, path optimization is performed based on frequency to improve inspection efficiency. In addition, through continuous evaluation and adjustment, an adaptive detection strategy is formed, and historical data and real-time data are fully utilized to achieve scientific decision-making based on data. These advances make inspection strategies more accurate, flexible, and efficient, and can better adapt to complex and changing actual situations, improve the timeliness and accuracy of fault detection, and optimize resource utilization, providing a more advanced and reliable foundation for fault detection and early warning methods in drone inspections.

[0089] like Figure 2 As shown, the frequency model is established based on the historical and current status data of the equipment, the detection frequency of the equipment is adjusted and synchronized to the drone inspection strategy, specifically including:

[0090] S210, analyzing historical detection data of the equipment under different working conditions, identifying the impact of different working conditions on the operating state of the equipment, and evaluating the current operating state of the equipment based on the data collected by the drone to determine what working condition it is in;

[0091] S220, based on the historical detection data of the equipment, a frequency model is established, with the rule that equipment with high risk working conditions and high failure rates requires high frequency detection, a corresponding detection frequency weight is assigned to each working condition, and the optimal detection frequency under different working conditions is calculated;

[0092] S230, dynamically adjust the detection frequency of the target area according to the usage frequency and status of the equipment, and formulate an initial inspection path of the drone based on the detection frequency.

[0093] S300, collects and annotates historical inspection data and builds a fault symptom database;

[0094] This step first comprehensively collects all fault data and normal operation data recorded in previous drone inspections, including but not limited to multi-dimensional information such as equipment operating parameters, environmental conditions, images and video data. Subsequently, based on expert knowledge and data mining technology, the fault data is deeply analyzed to determine the type and severity of the fault. Each piece of fault data is accurately labeled and its corresponding characteristic performance is recorded, such as abnormal vibration, abnormal temperature, surface defects, etc. This process not only involves traditional manual labeling, but also introduces semi-supervised learning methods to improve labeling efficiency and accuracy.

[0095] Further analysis of the characteristic performance under different fault categories, using machine learning algorithms such as cluster analysis and association rule mining, to identify the common characteristics and unique abnormal characteristics of various faults. This step also includes time series data analysis to reveal the dynamic characteristics of fault development and potential fault evolution laws. On this basis, a comprehensive fault symptom database is established, which not only contains the characteristics of known faults, but also reserves expansion space for unknown fault types. The database adopts a flexible data structure, supports the storage and rapid retrieval of multimodal data, and is dynamically updated with real-time data streams.

[0096] In addition, this step also introduces a data quality assessment mechanism to analyze the reliability and representativeness of the collected data, remove noise data, and ensure the high quality of the database. At the same time, a data version control system is established to record the evolution history of the data and support data backtracking and model iteration. Finally, strict data security and privacy protection measures are implemented to ensure the safe storage and compliant use of sensitive information.

[0097] like Figure 3 As shown, the collection and annotation of historical inspection data and establishment of a fault symptom database specifically include:

[0098] S310, collect all fault data and normal operation data recorded in previous drone inspections, determine the fault category based on the fault records, and annotate each piece of fault data and record its corresponding characteristic performance features;

[0099] S320, analyzing the characteristic manifestations under different fault categories, identifying common characteristics and abnormal characteristics, and establishing a fault symptom database, in which all acquired data are retained.

[0100] S400, establishing a fault identification model to identify the relationship between the fault category and the characteristic performance;

[0101] like Figure 4 As shown, the establishment of a fault identification model to identify the relationship between fault categories and characteristic manifestations specifically includes:

[0102] S410, analyzing the correlation between the characteristic value and the fault category, and selecting the fault-related characteristic value according to the correlation;

[0103] S420, establish a fault identification model, and train the model through the collected historical inspection data, define the rule logic of fault identification, and identify the relationship between the fault category and the characteristic performance.

[0104] S500, obtaining the current equipment status data obtained by the drone in real-time monitoring, determining whether the equipment has potential faults, calculating the detection frequency of the equipment through the frequency model, and optimizing and adjusting the inspection path of the drone;

[0105] This step integrates real-time data collection, fault identification, risk assessment and inspection strategy optimization to achieve intelligent equipment status monitoring and early warning. First, the system collects multi-dimensional data such as temperature, vibration, sound, and appearance of the equipment in real time through a variety of high-precision sensors carried by drones, such as infrared thermal imagers, high-resolution cameras, and acoustic sensors. These raw data are converted into feature vectors that can be used by the fault identification model after preprocessing steps such as noise reduction, standardization, and feature extraction. Next, the system calls the fault identification model trained in S4, which may be a complex integrated learning system containing multiple sub-models such as deep neural networks, support vector machines, and random forests. The model analyzes the feature vector and outputs the fault probability distribution and possible fault categories of the equipment. The system then evaluates the severity of the fault. This process not only considers the fault probability, but also combines factors such as the importance of the equipment and the potential impact range, and uses a multi-criteria decision-making method to comprehensively assess the fault level. Based on the evaluation results, the system uses geographic information system (GIS) technology to accurately mark the fault area to provide a spatial reference for subsequent key monitoring.

[0106] After determining the severity of the fault, the system uses a dynamic frequency model to calculate the required inspection frequency. This model not only takes into account the severity of the fault, but also introduces a time decay factor to reflect the changing trend of fault risk over time. This model can adaptively adjust the inspection frequency according to the dynamic changes of the fault, ensuring timely monitoring of high-risk areas and avoiding excessive inspection of low-risk areas, thereby optimizing resource allocation. Based on the calculated inspection frequency, the system defines different patrol modes, such as regular inspection, key inspection and emergency inspection, each of which corresponds to different flight altitudes, speeds and data collection densities.

[0107] Finally, the system uses advanced path planning algorithms, combined with multi-objective optimization technology, to generate the best inspection path for the drone by comprehensively considering multiple factors such as detection frequency, patrol mode, battery life, and weather conditions. This path not only meets the inspection requirements of all marked areas, but also maximizes battery efficiency and minimizes flight risks.

[0108] like Figure 5 As shown, the determination of whether a device has a potential fault and calculating the detection frequency of the device through a frequency model specifically include:

[0109] S510, collecting current equipment status data from sensors carried by the drone in real time, and preprocessing the data to extract feature values ​​related to fault identification;

[0110] S520, loading the trained fault identification model and inputting the characteristic value to perform current fault identification and analysis;

[0111] S530, obtaining the fault probability and fault category of the device output by the fault identification model, evaluating the severity of the fault, classifying the existing faults into fault levels, and marking the area where the device exists according to the fault level;

[0112] S540, according to the severity of the fault, use the frequency model to calculate the required detection frequency of the area, specifically:

[0113] ;

[0114] in, Indicates at time The required detection frequency is in times / time unit; represents the initial detection frequency; Represents the attenuation constant, the unit is: 1 / time unit; A numerical representation indicating the severity of the fault;

[0115] S550, defines the patrol mode and calculates the UAV patrol path according to the calculation results of the fault degree and frequency model.

[0116] In the embodiment of the present invention, the calculating of the inspection path of the drone specifically includes:

[0117] Get the initial inspection path of the drone , and calculate the fitness value :

[0118] ;

[0119] in:

[0120] Indicates the total distance of the initial inspection path;

[0121] ;

[0122] Indicates the total inspection time required for the initial inspection path;

[0123] ;

[0124] represents the penalty term for satisfying the frequency demand;

[0125] ;

[0126] in, Indicates the number of areas to be detected; Indicates the required inspection frequency for each area; Indicates the actual detection frequency of drones in each area;

[0127] Read the area where the annotation exists , select the path The designated position for additional detection, the designated position means: After the detection of the area outside is completed, return again ;

[0128] Traverse the current drone initial inspection path , and read the detection frequency, based on the detection frequency and the initial inspection path of the drone The total number of areas in the initial inspection path Divide equally and insert the pair at each dividing point The inspection command is used to obtain the new inspection path of the drone. ;

[0129] Re-evaluate the fitness value of the new path :

[0130] ;

[0131] Compare and ,like , then accept the new path .

[0132] This step uses an innovative path optimization algorithm that combines the ideas of multi-objective optimization and adaptive adjustment to achieve efficient, flexible inspection path planning that meets complex constraints. First, the system generates an initial inspection path based on previous inspection experience and current equipment distribution. Then, the quality of this path is evaluated through a comprehensive fitness function. This fitness function cleverly combines multiple key factors: total inspection distance, total inspection time, and a penalty term that reflects the degree to which frequency requirements are met.

[0133] The design of fitness function reflects the idea of ​​multi-objective optimization. , and To balance the importance of different goals. The total distance directly affects the energy consumption of the drone, the total time is related to the inspection efficiency, and the penalty term ensures that sufficient attention is paid to high-risk areas. In particular, the design of the penalty term F quantifies the degree to which the path meets the frequency requirements by comparing the difference between the actual detection frequency and the required frequency of each area. This design enables the system to prioritize the detection needs of high-risk areas while ensuring overall efficiency.

[0134] Next, the algorithm introduces an innovative step: selecting additional inspection locations for annotated areas (usually areas of high risk or requiring special attention). This step demonstrates the algorithm's flexibility and ability to adapt to special needs. By inserting additional inspections of these areas in the path, the system is able to increase the frequency of monitoring of key areas without significantly increasing the total inspection time and distance.

[0135] The algorithm then divides the initial path equally based on the detection frequency requirements and inserts inspection commands for the marked areas at the equalization points. This approach cleverly balances the needs of routine inspections and key inspections, ensuring that high-risk areas receive timely attention throughout the inspection process rather than being concentrated in a certain period of time.

[0136] Finally, the algorithm re-evaluates the fitness value of the newly generated path and decides whether to accept the new path by comparison. This process can be regarded as a local search step, and the path quality can be continuously optimized through iteration.

[0137] S600, analyzing the real-time monitoring data collected by the drone from the marked area again, setting a threshold to determine the abnormality of the real-time monitoring data, and issuing a warning signal if there is an actual abnormality;

[0138] like Figure 6 As shown, the real-time monitoring data collected by the drone from the marked area is analyzed again, and a threshold is set to determine the abnormality of the real-time monitoring data, specifically including:

[0139] S610, obtaining real-time monitoring data of the marked area obtained by the drone according to the new path inspection;

[0140] S620, setting a threshold of the monitoring data according to the fault severity of the marked area, comparing the monitoring data with the set threshold, and determining whether there is an anomaly that deviates from the fault stability;

[0141] S630, generating a warning signal for deviation from the fault stable region.

[0142] S700 collects on-site feedback data, evaluates the inspection effect and the accuracy of the early warning, sets inspection effect indicators, and if no feedback is received, determines the effectiveness of the inspection and whether the effect indicators are achieved, and analyzes whether the inspection strategy needs to be adjusted based on the judgment results.

[0143] Figure 7 The structural block diagram of the fault detection and early warning system in the UAV inspection provided by the embodiment of the present invention is as follows: Figure 7 As shown, the system comprises:

[0144] The data collection and feature extraction module 100 is used to collect the fault records, operating parameters and environmental status data of the existing equipment, and extract the data features to form feature values;

[0145] The frequency model and equipment grouping module 200 is used to establish a frequency model based on the historical detection and current status of different working conditions of the equipment, and adaptively adjust the detection frequency of the target area according to the uneven use of the equipment;

[0146] The fault symptom database establishment module 300 is used to collect historical inspection data, including fault data and normal operation data, annotate the collected data, identify the fault categories appearing in the data and their corresponding characteristic performance characteristics, and establish a fault symptom database based on the characteristic data;

[0147] The fault identification model training module 400 is used to establish a fault identification model, train the fault identification model through the collected historical inspection data, and identify the relationship between the fault category and the characteristic performance;

[0148] The real-time monitoring and path optimization module 500 is used to obtain the current equipment status data obtained by the drone sensor in real-time monitoring, analyze the current equipment status data using the fault identification model, determine whether the equipment has a potential fault, and if so, mark the area according to the fault degree, define a patrol mode for the marked area, and adjust the drone's patrol path using path optimization;

[0149] The abnormality detection and warning module 600 is used to judge the real-time monitoring data collected by the drone from the marked area again, set a threshold to judge the abnormality of the real-time monitoring data, and issue a warning signal if there is an actual abnormality;

[0150] The feedback evaluation and strategy adjustment module 700 is used to collect on-site feedback data, evaluate the inspection effect and the accuracy of the early warning, set the inspection effect indicators, and if no feedback is received, judge the effectiveness of the inspection and whether the effect indicators are achieved, and analyze whether the inspection strategy needs to be adjusted based on the judgment results.

[0151] like Figure 8 As shown, the fault identification model training module includes:

[0152] The characteristic value correlation analysis unit 410 is used to analyze the correlation between the characteristic value and the fault category, and select the fault-related characteristic value according to the correlation;

[0153] The fault identification model training unit 420 is used to establish a fault identification model and train the model through the collected historical inspection data, define the rule logic of fault identification, and identify the relationship between the fault category and the characteristic performance.

[0154] like Fig. 9 As shown, the real-time monitoring and path optimization module includes:

[0155] The real-time data collection and preprocessing unit 510 is used to collect the current equipment status data from the sensors carried by the drone in real time, and preprocess the data to extract the feature values ​​related to fault identification;

[0156] The fault identification model loading and feature value input unit 520 is used to load the trained fault identification model and input feature values ​​to perform current fault identification and analysis;

[0157] The fault probability and category evaluation unit 530 is used to obtain the fault probability and fault category of the device output by the fault identification model, evaluate the severity of the fault, classify the existing faults into fault levels, and mark the area where the device exists according to the fault level;

[0158] The detection frequency calculation unit 540 is used to calculate the detection frequency required for the area using a frequency model according to the severity of the fault;

[0159] The patrol mode definition and path optimization unit 550 is used to define the patrol mode and calculate the drone inspection path according to the calculation results of the fault degree and frequency model.

[0160] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0161] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0162] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0163] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0164] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A fault detection and early warning method in drone inspection, characterized in that: The method comprises: Collect fault records, operating parameters and environmental status data of existing equipment, extract data features and form feature values; Establish a frequency model based on the historical and current status data of the equipment, adjust the detection frequency of the equipment and synchronize it to the drone inspection strategy; Collect and annotate historical inspection data and establish a fault symptom database; Establish a fault identification model to identify the relationship between fault categories and characteristic manifestations; Obtain the current equipment status data obtained by real-time monitoring by drones, determine whether the equipment has potential faults, calculate the detection frequency of the equipment through the frequency model, and optimize and adjust the inspection path of the drone; The real-time monitoring data collected by the drone from the marked area is analyzed again, and a threshold is set to determine the abnormality of the real-time monitoring data. If there is an actual abnormality, an early warning signal is issued; Collect on-site feedback data, evaluate the inspection effect and the accuracy of early warning, set inspection effect indicators, and if no feedback is received, judge the effectiveness of the inspection and whether the effect indicators are achieved, and analyze whether the inspection strategy needs to be adjusted based on the judgment results; The frequency model is established based on the historical and current status data of the equipment, the detection frequency of the equipment is adjusted and synchronized to the drone inspection strategy, specifically including: Analyze the historical detection data of the equipment under different working conditions, identify the impact of different working conditions on the operating status of the equipment, and evaluate the current operating status of the equipment based on the data collected by the drone to determine what working condition it is in; A frequency model is established based on the historical detection data of the equipment. Based on the rule that equipment with high-risk working conditions and high failure rates requires high-frequency detection, a corresponding detection frequency weight is assigned to each working condition, and the optimal detection frequency under different working conditions is calculated; Dynamically adjust the detection frequency of the target area according to the usage frequency and status of the equipment, and formulate the initial inspection path of the drone based on the detection frequency; The establishment of a fault identification model to identify the relationship between the fault category and the characteristic performance specifically includes: Analyze the correlation between the characteristic values ​​and the fault categories, and select the fault-related characteristic values ​​based on the correlation; Establish a fault identification model and train the model through the collected historical inspection data, define the rule logic of fault identification, and identify the relationship between fault category and characteristic performance; The determining whether the device has a potential fault and calculating the detection frequency of the device through a frequency model specifically includes: Collect current equipment status data from the sensors onboard the drone in real time, pre-process the data, and extract feature values ​​related to fault identification; Load the trained fault identification model and input the characteristic value to identify and analyze the current fault; Obtain the equipment failure probability and failure category output by the fault identification model, evaluate the severity of the failure, classify the existing failures into failure levels, and mark the area where the equipment exists according to the failure level; According to the severity of the fault, the frequency model is used to calculate the detection frequency required for the area. Specifically: ; in, Indicates at time The required detection frequency is in times / time unit; represents the initial detection frequency; Represents the attenuation constant, the unit is: 1 / time unit; A numerical representation indicating the severity of the fault; Based on the calculated results of the fault degree and frequency model, the inspection mode is defined and the inspection path of the drone is calculated; The calculating of the inspection path of the drone specifically includes: Get the initial inspection path of the drone , and calculate the fitness value : ; in: Indicates the total distance of the initial inspection path; ; ; represents the penalty term for satisfying the frequency demand; ; in, Indicates the number of areas to be detected; Indicates the required inspection frequency for each area; Indicates the actual detection frequency of drones in each area; Read the area where the annotation exists Select the path The designated position for additional detection, the designated position means: After the detection of the area outside is completed, return again ; Traverse the current drone initial inspection path , and read the detection frequency, based on the detection frequency and the initial inspection path of the drone The total number of areas in the initial inspection path Divide equally and insert the pair at each dividing point The inspection command is used to obtain the new inspection path of the drone. ; Re-evaluate the fitness value of the new path : ； in, is the weight coefficient; Compare , then accept the new path .

2. The method according to claim 1, characterized in that The collecting and marking of historical inspection data and establishing a fault symptom database specifically include: Collect all fault data and normal operation data recorded in previous drone inspections, determine the fault category based on the fault records, and annotate each piece of fault data and record its corresponding characteristic performance features; Analyze the characteristic manifestations of different fault categories, identify common features and abnormal features, and establish a fault symptom database to store all the acquired data.

3. The method according to claim 2, characterized in that The real-time monitoring data collected by the drone from the marked area is analyzed again, and a threshold is set to determine the abnormality of the real-time monitoring data, specifically including: Obtain real-time monitoring data of the marked area obtained by the drone based on the new route inspection; Set the threshold of monitoring data according to the fault severity of the marked area, compare the monitoring data with the set threshold, and determine whether there is anomaly that deviates from the fault stability; Generate early warning signals for deviations from the failure stable region.

4. A fault detection and early warning system in drone inspection, used to implement the fault detection and early warning method in drone inspection according to any one of claims 1 to 3, characterized in that: The system comprises: The data collection and feature extraction module is used to collect the fault records, operating parameters and environmental status data of existing equipment, and extract data features to form feature values; The frequency model and equipment grouping module is used to establish a frequency model based on the historical detection and current status of different equipment working conditions, and adaptively adjust the detection frequency of the target area according to the uneven use of equipment; A fault symptom database establishment module is used to collect historical inspection data, including fault data and normal operation data, annotate the collected data, identify the fault categories appearing in the data and their corresponding characteristic performance characteristics, and establish a fault symptom database based on the characteristic data; The fault identification model training module is used to establish a fault identification model, train the fault identification model through the collected historical inspection data, and identify the relationship between the fault category and the characteristic performance; The real-time monitoring and path optimization module is used to obtain the current equipment status data obtained by the drone sensor in real-time monitoring, analyze the current equipment status data using the fault identification model, and determine whether the equipment has potential faults. If so, the area is marked according to the degree of fault, and a patrol mode is defined for the marked area. The inspection path of the drone is adjusted using path optimization; The anomaly detection and warning module is used to judge the real-time monitoring data collected by the drone from the marked area again, set a threshold to judge the abnormality of the real-time monitoring data, and issue a warning signal if there is an actual abnormality; The feedback evaluation and strategy adjustment module is used to collect on-site feedback data, evaluate the inspection effect and the accuracy of the early warning, set the inspection effect indicators, and if no feedback is received, judge the effectiveness of the inspection and whether the effect indicators are achieved, and analyze whether the inspection strategy needs to be adjusted based on the judgment results.

5. The system according to claim 4, characterized in that The fault identification model training module includes: The characteristic value correlation analysis unit is used to analyze the correlation between the characteristic value and the fault category, and select the fault-related characteristic value according to the correlation; The fault identification model training unit is used to establish a fault identification model and train the model through the collected historical inspection data, define the rule logic of fault identification, and identify the relationship between fault category and characteristic performance.

6. The system according to claim 5, characterized in that The real-time monitoring and path optimization module includes: The real-time data collection and preprocessing unit is used to collect the current equipment status data from the sensors carried by the drone in real time, preprocess the data, and extract the characteristic values ​​related to fault identification; The fault identification model loading and feature value input unit is used to load the trained fault identification model and input feature values ​​to identify and analyze the current fault; A fault probability and category evaluation unit is used to obtain the fault probability and fault category of the equipment output by the fault identification model, evaluate the severity of the fault, classify the existing faults into fault levels, and mark the area where the equipment exists according to the fault level; A detection frequency estimation unit is used to estimate the detection frequency required for the area using a frequency model according to the severity of the fault; The patrol mode definition and path optimization unit is used to define the patrol mode and calculate the UAV inspection path according to the calculation results of the fault degree and frequency model.

Images