A knowledge graph-based unmanned aerial vehicle fault diagnosis method, device, equipment and medium
By using a knowledge graph-based method for UAV fault diagnosis, a target knowledge graph model is constructed using data features and reference data to search for fault propagation paths. This solves the problem of low efficiency in existing technologies and achieves efficient and accurate fault diagnosis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- AVIC (CHENGDU) UAS CO LTD
- Filing Date
- 2026-02-05
- Publication Date
- 2026-05-12
AI Technical Summary
Existing UAV fault diagnosis methods are inefficient in complex relationships and scenarios with obscure knowledge, have poor automation effects, incomplete knowledge systems, complex model construction, and high requirements for sample data, making it difficult to meet the needs of efficient diagnosis.
A knowledge graph-based approach is adopted to obtain data features by interpreting UAV parameters, construct a target knowledge graph model, use symptom vectors and reference data to search for fault propagation paths, calculate fault probability, and generate a diagnostic report.
It improves the accuracy and efficiency of UAV fault diagnosis, simplifies the model building process, reduces reliance on expert knowledge bases, and enhances the level of automation in diagnosis.
Smart Images
Figure CN121637263B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned aerial vehicle (UAV) fault diagnosis technology, and in particular to a UAV fault diagnosis method, apparatus, equipment, and medium based on knowledge graphs. Background Technology
[0002] Currently, the typical method for diagnosing drone faults involves: an onboard flight data recorder (FDR) records the flight parameters of various systems during flight; after landing, the FDR data is downloaded from the recorder; then, specialized flight data interpretation software is used to manually read and analyze the data, and fault diagnosis is performed based on experience, troubleshooting manuals, or knowledge bases. However, in actual aircraft system fault diagnosis, diagnostic knowledge is characterized by nested relationships, complex logic, and multi-directional connections. This makes traditional methods time-consuming, inefficient, and overly reliant on expert knowledge bases when analyzing fault relationships.
[0003] Existing UAV ground fault diagnosis technologies mostly employ framework-based production rules and relational databases, or rely on statistical models, aiming to build structured, hierarchical automated diagnostic models to support fault reasoning. However, when dealing with scenarios like UAV systems where relationships are complex and knowledge is cryptic, these technologies exhibit the following significant drawbacks: incomplete knowledge systems, poor automated diagnostic performance; difficulty in representing complex knowledge logic; complex model construction processes and high sample data requirements; and low reasoning efficiency, making it difficult to meet the needs of efficient diagnosis.
[0004] In summary, improving the accuracy and efficiency of UAV fault diagnosis is a problem that needs to be addressed. Summary of the Invention
[0005] In view of this, the purpose of this invention is to provide a method, apparatus, device, and medium for UAV fault diagnosis based on knowledge graphs, which can improve the accuracy and efficiency of UAV fault diagnosis. The specific solution is as follows:
[0006] Firstly, this application discloses a method for diagnosing unmanned aerial vehicle (UAV) faults based on knowledge graphs, including:
[0007] The parameters of the UAV are interpreted to obtain data features, and symptom vectors are integrated based on the data features; the symptom vector contains several reference symptoms; the reference symptoms include reference symptom parameters and the probability of occurrence of each reference symptom parameter;
[0008] A target knowledge graph model is constructed based on data characteristics and reference data; the reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of UAVs.
[0009] Extract target symptoms with a probability of occurrence greater than the activation threshold from the symptom vector, and activate the symptom nodes mapped to the target knowledge graph model to obtain the activated knowledge graph model.
[0010] The activated knowledge graph model performs path search to obtain several target fault propagation paths for each target fault, and calculates the fault probability of the target fault based on the path information and symptom information of the several target fault propagation paths; the symptom information includes the occurrence probability and number of related symptom nodes.
[0011] The target diagnostic report is obtained by integrating the target fault and fault probability according to the preset diagnostic report template.
[0012] Optionally, the calculation of the failure probability of the target failure based on path information and symptom information of several target failure propagation paths includes:
[0013] The importance of the target path in the target fault propagation path is assessed based on the importance calculation formula and the path weight, path length, probability of occurrence of relevant symptom nodes, and number of relevant symptom nodes in the target fault propagation path.
[0014] The failure probability of a target failure is calculated based on the number of relevant symptom nodes corresponding to several target failure propagation paths and the importance of the target path.
[0015] The formula for calculating importance is as follows:
[0016] ;
[0017] in, This indicates the importance of the target path in the target fault propagation path; This represents the weight of the k-th path in the target fault propagation path; This represents the probability of the occurrence of the i-th related symptom node in the target fault propagation path; This indicates the number of relevant symptom nodes in the target fault propagation path.
[0018] Optionally, calculating the failure probability of the target failure based on the number of relevant symptom nodes corresponding to several target failure propagation paths and the importance of the target path includes:
[0019] Determine the maximum path importance of several target fault propagation paths corresponding to the target fault;
[0020] When the maximum path importance is less than the preset importance, the symptom quantity factor corresponding to the target fault is calculated by using the number of relevant symptom nodes corresponding to the target fault determined based on several target fault propagation paths.
[0021] The probability of target failure is calculated based on the maximum path importance and the number of symptoms.
[0022] Optionally, calculating the failure probability of the target failure based on the number of relevant symptom nodes corresponding to several target failure propagation paths and the importance of the target path includes:
[0023] Determine the maximum path importance of several target fault propagation paths corresponding to the target fault;
[0024] When the maximum path importance is not less than the preset importance, the maximum path importance is logarithmically scaled to obtain the processing result. The number of related symptom nodes corresponding to the target fault determined based on several target fault propagation paths is used to calculate the symptom quantity factor corresponding to the target fault, and the average path importance corresponding to several target fault propagation paths corresponding to the target fault is determined.
[0025] The probability of the target fault is calculated based on the average path importance, the number of symptoms, and the treatment results.
[0026] Optionally, the step of performing path search based on the activated knowledge graph model to obtain several target fault propagation paths for each target fault includes:
[0027] Path search is performed based on the activated knowledge graph model to obtain several temporary fault propagation paths;
[0028] Based on the temporary fault propagation path, we can count the number of target fault propagation paths corresponding to each target fault in the temporary fault propagation path.
[0029] Optionally, the step of interpreting the UAV parameters to obtain data features includes:
[0030] The system performs type analysis on UAV parameters to identify their parameter types, system analysis on UAV parameters to identify their parameter sources, and status analysis on discrete UAV parameters to determine triggering conditions. Parameter types include discrete and continuous types. Discrete types include fault codes and switch states; continuous types include timing types.
[0031] Data characteristics are determined based on parameter type, parameter source, and triggering conditions;
[0032] Accordingly, the integration of symptom vectors based on data features includes:
[0033] The triggered discrete-type drone parameters are identified as discrete-type reference symptom parameters, and the occurrence probability of the reference symptom parameters is determined based on the triggering situation.
[0034] The out-of-limit parameters are identified as continuous reference symptom parameters, and the probability of occurrence of the reference symptom parameters is determined based on the out-of-limit range; the out-of-limit parameters are extracted by performing out-of-limit analysis on continuous UAV parameters.
[0035] The symptom vector is integrated based on discrete and continuous reference symptom parameters and the probability of occurrence of the reference symptom parameters.
[0036] Optionally, the step of interpreting the UAV parameters to obtain data features includes:
[0037] The raw parameter values are obtained by parsing the UAV parameters, and the engineered parameter values are calculated based on the product of the scaling factor and the target parameter value. The target parameter value is the difference between the raw parameter value and the offset. The offset and scaling factor are the core parameters of the sensor that collects UAV parameters when performing data conversion.
[0038] The continuity of the timestamps corresponding to the parameter engineering values is checked and time alignment is performed to obtain a complete dataset under a unified timeline;
[0039] The complete dataset is analyzed to obtain data features.
[0040] Secondly, this application discloses a knowledge graph-based unmanned aerial vehicle (UAV) fault diagnosis device, comprising:
[0041] The vector integration module is used to interpret UAV parameters to obtain data features and integrate symptom vectors based on the data features. The symptom vector contains several reference symptoms. The reference symptoms include reference symptom parameters and the probability of occurrence of each reference symptom parameter.
[0042] The model building module is used to construct a target knowledge graph model based on data characteristics and reference data; the reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of UAVs.
[0043] The node activation module is used to extract target symptoms with a probability of occurrence greater than the activation threshold from the symptom vector, and activate the symptom nodes mapped to the target knowledge graph model to obtain the activated knowledge graph model.
[0044] The path search module is used to perform path search based on the activated knowledge graph model to obtain several target fault propagation paths for each target fault, and calculate the fault probability of the target fault based on the path information and symptom information of the several target fault propagation paths; the symptom information includes the occurrence probability and number of related symptom nodes.
[0045] The diagnostic report integration module is used to integrate target faults and fault probabilities based on preset diagnostic report templates to obtain target diagnostic reports.
[0046] Thirdly, this application discloses an electronic device, comprising:
[0047] Memory, used to store computer programs;
[0048] A processor is used to execute the computer program to implement the aforementioned knowledge graph-based UAV fault diagnosis method.
[0049] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned knowledge graph-based UAV fault diagnosis method.
[0050] As can be seen, this application interprets UAV parameters to obtain data features and integrates symptom vectors based on these features. The symptom vectors contain several reference symptoms. Each reference symptom includes reference symptom parameters and the probability of occurrence for each parameter. A target knowledge graph model is constructed based on the data features and reference data. The reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of the UAV. Target symptoms with a probability of occurrence greater than an activation threshold are extracted from the symptom vectors, and the symptom nodes mapped to these symptoms in the target knowledge graph model are activated to obtain the activated knowledge graph model. Path search is performed based on the activated knowledge graph model to obtain several target fault propagation paths for each type of target fault. The fault probability of the target fault is calculated based on the path information and symptom information of these propagation paths. The symptom information includes the probability and number of occurrences of relevant symptom nodes. A target diagnostic report is obtained by integrating the target fault and fault probability according to a preset diagnostic report template. Therefore, this application constructs a target knowledge graph model by combining feature data obtained from interpreting UAV parameters with reference data, fully characterizing knowledge relationships. Directly using a knowledge graph model simplifies the construction process. Furthermore, feature data is structured data, while reference data is unstructured data. The target knowledge graph model achieves the fusion and synergy of these two types of data, utilizing both the computability of structured data and the richness of knowledge in unstructured data. This solves the problems of incomplete knowledge systems and difficulty in characterizing complex relationships in traditional diagnostic methods, thus improving accuracy. In the subsequent calculation of fault probability, this application proposes calculation based on path information of the target fault propagation path and the probability and number of occurrences of related symptom nodes. Path information is unstructured data, while the probability and number of occurrences of related symptom nodes are structured data. Using both types of data in the diagnostic process improves diagnostic accuracy. This application directly searches the target fault propagation path, calculates the fault probability of the target fault based on path information and the probability and number of occurrences of related symptom nodes, and generates a diagnostic report accordingly. This eliminates the need for production rules and relational databases in the framework, and avoids reliance on statistical models, thereby improving diagnostic efficiency. Attached Figure Description
[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0052] Figure 1 This is a flowchart of a knowledge graph-based UAV fault diagnosis method disclosed in this application;
[0053] Figure 2 This is a schematic diagram of a knowledge graph-based drone fault diagnosis process disclosed in this application;
[0054] Figure 3 This is a schematic diagram of a specific knowledge graph-based UAV fault diagnosis process disclosed in this application;
[0055] Figure 4 This is a schematic diagram of the structure of a knowledge graph-based unmanned aerial vehicle (UAV) fault diagnosis device disclosed in this application.
[0056] Figure 5 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation
[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0058] Existing UAV ground fault diagnosis technologies mostly employ framework-based production rules and relational databases, or rely on statistical models, aiming to build structured, hierarchical automated diagnostic models to support fault reasoning. However, when dealing with scenarios like UAV systems where relationships are complex and knowledge is cryptic, these technologies exhibit the following significant drawbacks: incomplete knowledge systems, poor automated diagnostic performance; difficulty in representing complex knowledge logic; complex model construction processes and high sample data requirements; and low reasoning efficiency, making it difficult to meet the needs of efficient diagnosis.
[0059] Therefore, this application proposes a knowledge graph-based UAV fault diagnosis scheme, which can improve the accuracy and efficiency of UAV fault diagnosis.
[0060] This application discloses a knowledge graph-based method for diagnosing drone faults. See [link to relevant documentation]. Figure 1 As shown, the method includes:
[0061] Step S11: Interpret the UAV parameters to obtain data features, and integrate the symptom vector based on the data features; the symptom vector contains several reference symptoms; the reference symptoms include reference symptom parameters and the probability of occurrence of each reference symptom parameter.
[0062] In this embodiment, the UAV parameters (flight parameters) specifically include data such as power supply system current, voltage, generator switching alarm, power supply temperature, and battery over-temperature alarm. Flight parameters: flight parameters refer to data collected by airborne sensors that reflect flight status and system operating conditions.
[0063] In this embodiment, when interpreting UAV parameters, the UAV parameters need to be parsed and preprocessed before interpretation can be performed. Specifically, the interpretation of UAV parameters to obtain data features includes: parsing the UAV parameters to obtain the original parameter values, and calculating the parameter engineering values based on the product of the scaling factor and the parameter target value; the parameter target value is the difference between the original parameter value and the offset; the offset and scaling factor are the core parameters of the sensor that collects UAV parameters when performing data conversion; performing continuity checks on the timestamps corresponding to the parameter engineering values and performing time alignment to obtain a complete dataset under a unified time axis; and interpreting the complete dataset to obtain data features.
[0064] It should be noted that the process of parsing binary flight parameter data is as follows: The start position of a complete data frame is located by searching for a specific frame header synchronization word. Based on a preset parameter definition table, the raw binary data stream is identified and parsed. The parsed raw parameter values (usually integers) are converted into engineered values with actual physical meaning. For each parameter, a linear transformation formula is applied according to its calibration rules: Engineered value = (Raw value - Offset) × Scaling factor; where the offset and scaling factor are the core parameters for sensor data conversion. The offset is responsible for adjusting the zero point, resolving the reference difference between the sensor output and the actual physical value, while the scaling factor solves the numerical scaling problem, mapping the digital range of the sensor output to the actual physical quantity range.
[0065] It should be noted that the preprocessing process for the parsed flight parameter data is as follows: The timestamp parameters are checked for continuity to detect and handle anomalies such as dropped frames or timestamp jumps; multiple flight parameter data streams from different subsystems or recording periods are time-aligned to form a complete dataset with a unified timeline; the preprocessed flight parameter data is then interpreted to extract symptom vectors. Specifically, the process for detecting and handling dropped frames or timestamp jumps is as follows: Time difference statistics are performed on the dropped frames or timestamp jump portions, and a threshold is set. Dropped frames that do not exceed the threshold use the data from the previous frame; portions exceeding the threshold are padded with +9999999 data and used as markers.
[0066] In this embodiment, the step of interpreting UAV parameters to obtain data features includes: interpreting the type of UAV parameters to identify their parameter type; interpreting the system of UAV parameters to identify their parameter source; and interpreting the status of discrete UAV parameters to determine the triggering condition. Parameter types include discrete and continuous types. Discrete types include fault codes and switch states; continuous types include timing types. Data features are determined based on parameter type, parameter source, and triggering condition. Correspondingly, the step of integrating symptom vectors based on data features includes: determining triggered discrete UAV parameters as discrete reference symptom parameters and determining the occurrence probability of the reference symptom parameters based on the triggering condition; determining out-of-limit parameters as continuous reference symptom parameters and determining the occurrence probability of the reference symptom parameters based on the out-of-limit range; out-of-limit parameters are extracted by performing out-of-limit analysis on continuous UAV parameters; and integrating symptom vectors based on discrete and continuous reference symptom parameters and the occurrence probability of the reference symptom parameters.
[0067] It should be noted that the specific process of interpreting the preprocessed flight parameter data and extracting the symptom vector is as follows: First, the flight parameter data undergoes type interpretation, system interpretation, and status interpretation. Type interpretation includes identifying whether the flight parameter data is a discrete quantity such as a fault code or switch status, or a continuous quantity such as time-series data; system interpretation refers to identifying the source of the flight parameter data; and status interpretation refers to trigger interpretation of discrete data. Exceedance analysis is performed on the time-series data, quantifying the probability of occurrence by calculating the deviation of the actual value from the exceedance threshold. The formula is as follows: ;in, This represents the threshold at which an anomaly begins; exceeding this value indicates the onset of abnormal behavior. The steepness of the control curve is determined by expert experience. This represents the current drone parameters; additionally, during trigger detection, the probability of occurrence is 1 if triggered and 0 if not triggered. The trigger fault codes, timing data exceeding limits, and corresponding occurrence probabilities from the flight parameter interpretation are integrated into a flight parameter interpretation symptom vector.
[0068] In one specific embodiment, the symptom vector is {overcurrent (0.85), undervoltage (0.72), abnormal power supply temperature (0.63), generator switching alarm (1), and battery over-temperature alarm (0)}.
[0069] Step S12: Construct a target knowledge graph model based on data characteristics and reference data; the reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of the UAV.
[0070] It should be noted that knowledge graphs are a knowledge representation method that uses a graph structure to represent entities and their relationships, suitable for reasoning and analysis of complex relationships. The process of constructing a target knowledge graph model based on data characteristics and reference data is as follows: First, knowledge modeling is performed. The specific functions and scope of the knowledge graph are clarified, and relevant data is defined: 1. Entity definition. The entity types in the graph are determined, including flight parameters, aircraft components, fault types, fault symptoms, etc. Simultaneously, their attribute names, descriptions, identifiers, severity, etc., are constructed. The specific structure is defined as follows: (Entity A {Name: name, Description: description, Identifier: node_001, Severity: High, Others: Unique attributes of different nodes}); 2. Relationship definition. The possible relationships and attributes between entity types are defined, such as causal relationships, association relationships, etc., and relationship attributes such as weights; Second, knowledge graph construction is performed. Based on the defined entity types, the specific entities, relationships, and attributes of the knowledge graph are constructed: 1. Entity extraction or mapping. 1. Extract or map specific entity nodes from data such as the system component architecture of the UAV, data features extracted from flight parameter interpretation, and troubleshooting manuals for UAV fault analysis and diagnosis; 2. Establish or map entity relationships and attributes. Based on the troubleshooting manuals for UAV fault analysis and diagnosis and expert experience, establish or map specific relationships and attributes and store them in an Excel spreadsheet. The format for establishing entity node relationships is as follows: (Fault {Name}) - [Cause {Weight}] → (Symptom {Name}). Then, integrate the Neo4j graph database and store the constructed knowledge graph model in the Neo4j graph database.
[0071] It should be noted that Neo4j is a graph-based database management system that uses the Cypher query language, suitable for highly connected data storage and retrieval; Cypher is the declarative graph query language used by Neo4j for graph construction, querying, and updating. Nodes: The basic units in a knowledge graph, representing entities (such as systems, components, or symptoms); Relationships: Directed connections between nodes, representing logical or physical associations between entities. Flight parameter interpretation: The process of feature extraction, state identification, and anomaly detection of flight parameters.
[0072] Step S13: Extract target symptoms with a probability of occurrence greater than the activation threshold from the symptom vector, and activate the symptom nodes mapped to the target knowledge graph model to obtain the activated knowledge graph model.
[0073] In this embodiment, the symptom vector is represented as: ,in Indicates the first The probability of occurrence of symptoms interpreted by each flight parameter (within the range of 0 to 1). This refers to the total number of symptoms interpreted by flight parameters; setting an activation threshold: setting an activation threshold. This threshold can be adjusted based on historical data and expert experience. A symptom node is activated only when the combination of symptoms in the symptom vector exceeds the threshold. The resulting set of activated symptoms is as follows: The symptom vectors extracted by the flight parameter interpretation are mapped to the corresponding symptom nodes in the knowledge graph and activated, resulting in the set of activated symptoms that are determined to participate in subsequent reasoning. ;in, Indicates the first The probability of occurrence of similar symptoms, This indicates the set activation threshold. The set of activated symptom nodes constitutes the initial inference starting point, which is marked as active in the knowledge graph to initiate the subsequent inference process.
[0074] In one specific embodiment, when the symptom vector is {overcurrent (0.85), undervoltage (0.72), abnormal power supply temperature (0.63), generator switching alarm (1), battery overtemperature alarm (0)}, the activation threshold is set to 0.5, and the activated symptom set is: {overcurrent (0.85), undervoltage (0.72), abnormal power supply temperature (0.63), generator switching alarm (1)}.
[0075] Step S14: Perform path search based on the activated knowledge graph model to obtain several target fault propagation paths for each target fault, and calculate the fault probability of the target fault based on the path information and symptom information of the several target fault propagation paths; the symptom information includes the occurrence probability and number of related symptom nodes.
[0076] In this embodiment, the step of performing path search based on the activated knowledge graph model to obtain several target fault propagation paths for each target fault includes: performing path search based on the activated knowledge graph model to obtain several temporary fault propagation paths; and statistically analyzing several target fault propagation paths corresponding to each target fault in the temporary fault propagation paths.
[0077] In one specific embodiment, the temporary fault propagation path is as follows: Temporary fault propagation path 1: Main generator fault (weight: 0.9) → generator switching alarm (it can also be represented as (fault {main generator fault}) - [cause {weight}] → (symptom {name}) based on (fault {name}) - [cause {weight}] → (symptom {generator switching alarm})); Temporary fault propagation path 2: Power distribution unit fault (weight: 0.75) → line short circuit (weight: 0.8) → overcurrent; Temporary fault propagation path 3: Voltage regulation fault (weight: 0.7) → undervoltage; Temporary fault propagation path 4: Cooling system fault (weight: 0.6) → abnormal power supply temperature.
[0078] Based on the above temporary fault propagation paths, a line short circuit can be extracted as a target fault propagation path, namely the target fault propagation path 5 below.
[0079] In one specific embodiment, based on the above-mentioned temporary fault propagation path, the target fault propagation path is as follows: Target fault propagation path 1: Main generator fault (weight: 0.9) → Generator switching alarm; Target fault propagation path 2: Power distribution unit fault (weight: 0.75) → Line short circuit (weight: 0.8) → Overcurrent; Target fault propagation path 3: Voltage regulation fault (weight: 0.7) → Undervoltage; Target fault propagation path 4: Cooling system fault (weight: 0.6) → Abnormal power supply temperature; Target fault propagation path 5: Line short circuit (weight: 0.8) → Overcurrent.
[0080] In this embodiment, the calculation of the failure probability of a target failure based on the path information and symptom information of several target failure propagation paths includes: assessing the importance of the target path of the target failure propagation path according to the importance calculation formula, and based on the path weight, path length, occurrence probability of related symptom nodes and the number of related symptom nodes in the target failure propagation path; and calculating the failure probability of the target failure based on the number of related symptom nodes corresponding to several target failure propagation paths and the importance of the target path.
[0081] The formula for calculating importance is as follows:
[0082] ; This indicates the importance of the target path in the target fault propagation path; Let L represent the weight of the k-th path in the target fault propagation path, and L represent the number of path weights in the target fault propagation path. This represents the probability of the occurrence of the i-th related symptom node in the target fault propagation path; This indicates the number of relevant symptom nodes in the target fault propagation path.
[0083] In one specific embodiment, the target importance of the target fault propagation path is as follows: Target fault propagation path 1 (target fault propagation path when the target fault is a main generator fault): Target fault propagation path 2 (target fault propagation path when the target fault is a power distribution unit fault): Target fault propagation path 3 (target fault propagation path when the target fault is a voltage regulation fault): Target fault propagation path 4 (target fault propagation path when the target fault is a heat dissipation system fault): ; Target fault propagation path 5 (target fault propagation path when the target fault is a line short circuit): ;
[0084] In this embodiment, the step of calculating the failure probability of a target fault based on the number of relevant symptom nodes corresponding to several target fault propagation paths and the importance of the target path includes: determining the maximum path importance corresponding to several target fault propagation paths corresponding to the target fault; when the maximum path importance is less than a preset importance, calculating the symptom quantity factor corresponding to the target fault using the number of relevant symptom nodes corresponding to the target fault determined based on several target fault propagation paths; and calculating the failure probability of the target fault based on the maximum path importance and the symptom quantity factor.
[0085] In this embodiment, the step of calculating the failure probability of a target fault based on the number of relevant symptom nodes corresponding to several target fault propagation paths and the importance of the target path includes: determining the maximum path importance corresponding to several target fault propagation paths corresponding to the target fault; when the maximum path importance is not less than a preset importance, performing logarithmic scaling on the maximum path importance to obtain a processing result, and using the number of relevant symptom nodes corresponding to the target fault determined based on several target fault propagation paths to calculate the symptom quantity factor corresponding to the target fault, and determining the average path importance corresponding to several target fault propagation paths corresponding to the target fault; and calculating the failure probability of the target fault based on the average path importance, the symptom quantity factor, and the processing result.
[0086] It should be noted that the formula for determining the importance of the maximum path is as follows: Where M represents the number of target fault propagation paths corresponding to a target fault; the formula for determining the total importance of paths is as follows: The formula for determining the average path importance is as follows: The formula for logarithmically scaling the importance of the maximum path is as follows: ; Formula for calculating the symptom quantity factor: ; The symptom quantity factor represents the target fault. This represents the number of related symptoms corresponding to the target fault. The formula for calculating the fault probability is as follows: .
[0087] It should be noted that the calculated failure probability needs to be thresholded. When the calculated failure probability is greater than 1, it is set to 1. It should also be noted that the coefficients in the failure probability calculation formula (e.g., 2, 0.7, 0.3 and 1) and the coefficients in the logarithmic scaling formula (e.g., 10, 1 and 11) can be changed according to the actual situation, and no specific restrictions are made here.
[0088] It should be noted that logarithmic scaling prevents extremely high values from dominating the calculation results and maintains the reasonableness of the probability distribution. The use of the symptom count factor can avoid over-reliance on the number of symptoms. The coefficient 0.1 represents the contribution weight of each symptom, and the upper limit of 0.5 is designed to prevent the number of symptoms from excessively affecting the probability calculation. Depending on the actual situation, 0.5 can also be updated to other values.
[0089] In one specific embodiment, the only path for the target fault to be a main generator fault is target fault propagation path 1. The calculation of the fault probability of the main generator fault is as follows: ; ; ; ; ; The failure probability is greater than 1, therefore the final failure probability is taken as 1. The calculation process for the failure probability of other failures will not be illustrated here.
[0090] Step S15: Integrate the target fault and fault probability according to the preset diagnostic report template to obtain the target diagnostic report.
[0091] In this embodiment, the target diagnostic report may include the name of the target fault, the fault probability (i.e., the diagnosis probability), the ranking, and the corresponding related symptoms. The target diagnostic report can be visualized.
[0092] In one specific embodiment, the target diagnostic report is shown in Table 1 below:
[0093] Table 1 Target Diagnostic Report
[0094]
[0095] See details Figure 2 The diagram illustrates a knowledge graph-based UAV fault diagnosis process, encompassing UAV parameter (flight parameter) parsing and preprocessing, interpretation, knowledge graph construction and storage, activation, path search, path importance calculation, fault probability calculation, and result integration to obtain a target diagnostic report. First, the UAV's binary flight parameter data is parsed and converted into readable and processable engineering values. The parsed flight parameter data undergoes preprocessing and interpretation; the interpretation function primarily extracts flight parameter data features relevant to fault diagnosis. Combining the UAV's system component architecture, the data features extracted from flight parameter interpretation, and data from UAV fault analysis and diagnosis troubleshooting manuals and expert experience, a knowledge graph is constructed. Supported by the Neo4j graph database, a knowledge graph reasoning model is established to perform logical reasoning about the causal relationships of aircraft symptoms and the probabilities of fault root causes. Finally, the obtained fault diagnosis reasoning results are processed to output the final diagnostic result. Specific details have been described above and will not be repeated here.
[0096] See Figure 3 The diagram illustrates a specific knowledge graph-based UAV fault diagnosis process. It shows the relationship between symptoms, faults, and components determined based on reference data (UAV fault analysis and diagnosis troubleshooting manuals, expert experience, and UAV system component architecture), the interpretation results based on UAV parameters (flight parameters, including power supply system flight parameter data, fuel system flight parameter data, ring system flight parameter data, landing gear system flight parameter data, etc.), and the knowledge graph constructed based on the relationship diagram and interpretation results. Subsequently, the fault probability is calculated based on the knowledge graph, and the diagnosis results are integrated.
[0097] In summary, this application overcomes the shortcomings of traditional methods in separating unstructured and structured data, effectively alleviates the difficulties in multi-source information fusion in traditional methods, makes full use of unstructured health management knowledge in the UAV field, can more naturally express complex fault correlations, improves the data objectivity of UAV ground fault diagnosis process, reduces the time and human and material costs in the diagnosis process, and promotes the automation and intelligence of UAVs.
[0098] As can be seen, this application interprets UAV parameters to obtain data features and integrates symptom vectors based on these features. The symptom vectors contain several reference symptoms. Each reference symptom includes reference symptom parameters and the probability of occurrence for each parameter. A target knowledge graph model is constructed based on the data features and reference data. The reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of the UAV. Target symptoms with a probability of occurrence greater than an activation threshold are extracted from the symptom vectors, and the symptom nodes mapped to these symptoms in the target knowledge graph model are activated to obtain the activated knowledge graph model. Path search is performed based on the activated knowledge graph model to obtain several target fault propagation paths for each type of target fault. The fault probability of the target fault is calculated based on the path information and symptom information of these propagation paths. The symptom information includes the probability and number of occurrences of relevant symptom nodes. A target diagnostic report is obtained by integrating the target fault and fault probability according to a preset diagnostic report template. Therefore, this application constructs a target knowledge graph model by combining feature data obtained from interpreting UAV parameters with reference data, fully characterizing knowledge relationships. Directly using a knowledge graph model simplifies the construction process. Furthermore, feature data is structured data, while reference data is unstructured data. The target knowledge graph model achieves the fusion and synergy of these two types of data, utilizing both the computability of structured data and the richness of knowledge in unstructured data. This solves the problems of incomplete knowledge systems and difficulty in characterizing complex relationships in traditional diagnostic methods, thus improving accuracy. In the subsequent calculation of fault probability, this application proposes calculation based on path information of the target fault propagation path and the probability and number of occurrences of related symptom nodes. Path information is unstructured data, while the probability and number of occurrences of related symptom nodes are structured data. Using both types of data in the diagnostic process improves diagnostic accuracy. This application directly searches the target fault propagation path, calculates the fault probability of the target fault based on path information and the probability and number of occurrences of related symptom nodes, and generates a diagnostic report accordingly. This eliminates the need for production rules and relational databases in the framework, and avoids reliance on statistical models, thereby improving diagnostic efficiency.
[0099] Accordingly, this application also discloses a knowledge graph-based UAV fault diagnosis device, see [link to relevant documentation]. Figure 4 As shown, the device includes:
[0100] The vector integration module 11 is used to interpret the UAV parameters to obtain data features and integrate symptom vectors based on the data features; the symptom vector contains several reference symptoms; the reference symptoms include reference symptom parameters and the probability of occurrence of each reference symptom parameter;
[0101] Model building module 12 is used to build a target knowledge graph model based on data features and reference data; the reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of UAVs.
[0102] The node activation module 13 is used to extract target symptoms with a probability of occurrence greater than the activation threshold from the symptom vector, and activate the symptom nodes mapped to the target knowledge graph model to obtain the activated knowledge graph model.
[0103] The path search module 14 is used to perform path search based on the activated knowledge graph model to obtain several target fault propagation paths for each target fault, and to calculate the fault probability of the target fault based on the path information and symptom information of the several target fault propagation paths; the symptom information includes the occurrence probability and number of related symptom nodes.
[0104] The diagnostic report integration module 15 is used to integrate the target fault and fault probability according to the preset diagnostic report template to obtain the target diagnostic report.
[0105] The more specific working process of each of the above modules can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.
[0106] As can be seen, this application interprets UAV parameters to obtain data features and integrates symptom vectors based on these features. The symptom vectors contain several reference symptoms. Each reference symptom includes reference symptom parameters and the probability of occurrence for each parameter. A target knowledge graph model is constructed based on the data features and reference data. The reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of the UAV. Target symptoms with a probability of occurrence greater than an activation threshold are extracted from the symptom vectors, and the symptom nodes mapped to these symptoms in the target knowledge graph model are activated to obtain the activated knowledge graph model. Path search is performed based on the activated knowledge graph model to obtain several target fault propagation paths for each type of target fault. The fault probability of the target fault is calculated based on the path information and symptom information of these propagation paths. The symptom information includes the probability and number of occurrences of relevant symptom nodes. A target diagnostic report is obtained by integrating the target fault and fault probability according to a preset diagnostic report template. Therefore, this application constructs a target knowledge graph model by combining feature data obtained from interpreting UAV parameters with reference data, fully characterizing knowledge relationships. Directly using a knowledge graph model simplifies the construction process. Furthermore, feature data is structured data, while reference data is unstructured data. The target knowledge graph model achieves the fusion and synergy of these two types of data, utilizing both the computability of structured data and the richness of knowledge in unstructured data. This solves the problems of incomplete knowledge systems and difficulty in characterizing complex relationships in traditional diagnostic methods, thus improving accuracy. In the subsequent calculation of fault probability, this application proposes calculation based on path information of the target fault propagation path and the probability and number of occurrences of related symptom nodes. Path information is unstructured data, while the probability and number of occurrences of related symptom nodes are structured data. Using both types of data in the diagnostic process improves diagnostic accuracy. This application directly searches the target fault propagation path, calculates the fault probability of the target fault based on path information and the probability and number of occurrences of related symptom nodes, and generates a diagnostic report accordingly. This eliminates the need for production rules and relational databases in the framework, and avoids reliance on statistical models, thereby improving diagnostic efficiency.
[0107] Furthermore, embodiments of this application also provide an electronic device. Figure 5 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.
[0108] Figure 5This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a display screen 23, an input / output interface 24, a communication interface 25, a power supply 26, and a communication bus 27. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the knowledge graph-based UAV fault diagnosis method disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.
[0109] In this embodiment, the power supply 26 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 25 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 24 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.
[0110] Furthermore, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk, or optical disk, etc. The resources stored thereon can include computer programs 221, and the storage method can be temporary storage or permanent storage. The computer programs 221 may include, in addition to computer programs capable of performing the knowledge graph-based UAV fault diagnosis method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, computer programs capable of performing other specific tasks.
[0111] Furthermore, embodiments of this application also disclose a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned knowledge graph-based UAV fault diagnosis method.
[0112] The specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.
[0113] The various embodiments in this application are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. For the same or similar parts between the various embodiments, refer to each other. As for the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and relevant parts can be referred to in the method section.
[0114] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0115] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0116] Finally, it should be noted that in this document, relational terms such as "first" and "first" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0117] The above provides a detailed description of a knowledge graph-based method, apparatus, device, and storage medium for UAV fault diagnosis. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for diagnosing unmanned aerial vehicle (UAV) faults based on knowledge graphs, characterized in that, include: The parameters of the drone are interpreted to obtain data features, and symptom vectors are integrated based on the data features; The symptom vector contains several reference symptoms; refer to Symptoms include reference symptom parameters and the probability of occurrence for each reference symptom parameter; A target knowledge graph model is constructed based on data characteristics and reference data; the reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of UAVs. Extract target symptoms with a probability of occurrence greater than the activation threshold from the symptom vector, and activate the symptom nodes mapped to the target knowledge graph model to obtain the activated knowledge graph model. The activated knowledge graph model performs path search to obtain several target fault propagation paths for each target fault, and calculates the fault probability of the target fault based on the path information and symptom information of the several target fault propagation paths; the symptom information includes the occurrence probability and number of related symptom nodes. A target diagnostic report is generated by integrating the target fault and fault probability based on a preset diagnostic report template. The calculation of the failure probability of a target failure based on path information and symptom information of several target failure propagation paths includes: The importance of the target path in the target fault propagation path is assessed based on the importance calculation formula and the path weight, path length, probability of occurrence of relevant symptom nodes, and number of relevant symptom nodes in the target fault propagation path. The failure probability of a target failure is calculated based on the number of relevant symptom nodes corresponding to several target failure propagation paths and the importance of the target path. The formula for calculating importance is as follows: ； in, This indicates the importance of the target path in the target fault propagation path; This represents the weight of the k-th path in the target fault propagation path; This represents the probability of the occurrence of the i-th related symptom node in the target fault propagation path; This indicates the number of relevant symptom nodes in the target fault propagation path.
2. The UAV fault diagnosis method based on knowledge graph according to claim 1, characterized in that, The calculation of the failure probability of a target failure based on the number of relevant symptom nodes corresponding to several target failure propagation paths and the importance of the target path includes: Determine the maximum path importance of several target fault propagation paths corresponding to the target fault; When the maximum path importance is less than the preset importance, the symptom quantity factor corresponding to the target fault is calculated by using the number of relevant symptom nodes corresponding to the target fault determined based on several target fault propagation paths. The probability of target failure is calculated based on the maximum path importance and the number of symptoms.
3. The UAV fault diagnosis method based on knowledge graph according to claim 1, characterized in that, The calculation of the failure probability of a target failure based on the number of relevant symptom nodes corresponding to several target failure propagation paths and the importance of the target path includes: Determine the maximum path importance of several target fault propagation paths corresponding to the target fault; When the maximum path importance is not less than the preset importance, the maximum path importance is logarithmically scaled to obtain the processing result. The number of related symptom nodes corresponding to the target fault determined based on several target fault propagation paths is used to calculate the symptom quantity factor corresponding to the target fault, and the average path importance corresponding to several target fault propagation paths corresponding to the target fault is determined. The probability of the target fault is calculated based on the average path importance, the number of symptoms, and the treatment results.
4. The UAV fault diagnosis method based on knowledge graph according to claim 1, characterized in that, The step of performing path search based on the activated knowledge graph model to obtain several target fault propagation paths for each target fault includes: Path search is performed based on the activated knowledge graph model to obtain several temporary fault propagation paths; Based on the temporary fault propagation path, we can count the number of target fault propagation paths corresponding to each target fault in the temporary fault propagation path.
5. The knowledge graph-based UAV fault diagnosis method according to claim 1, characterized in that, The process of interpreting UAV parameters to obtain data features includes: The system performs type analysis on UAV parameters to identify their parameter types, system analysis on UAV parameters to identify their parameter sources, and status analysis on discrete UAV parameters to determine triggering conditions. Parameter types include discrete and continuous types. Discrete types include fault codes and switch states; continuous types include timing types. Data characteristics are determined based on parameter type, parameter source, and triggering conditions; Accordingly, the integration of symptom vectors based on data features includes: The triggered discrete-type drone parameters are identified as discrete-type reference symptom parameters, and the occurrence probability of the reference symptom parameters is determined based on the triggering situation. The out-of-limit parameters are identified as continuous reference symptom parameters, and the probability of occurrence of the reference symptom parameters is determined based on the out-of-limit range; the out-of-limit parameters are extracted by performing out-of-limit analysis on continuous UAV parameters. The symptom vector is integrated based on discrete and continuous reference symptom parameters and the probability of occurrence of the reference symptom parameters.
6. The knowledge graph-based UAV fault diagnosis method according to any one of claims 1 to 5, characterized in that, The process of interpreting UAV parameters to obtain data features includes: The raw parameter values are obtained by parsing the UAV parameters, and the engineered parameter values are calculated based on the product of the scaling factor and the target parameter value. The target parameter value is the difference between the raw parameter value and the offset. The offset and scaling factor are the core parameters of the sensor that collects UAV parameters when performing data conversion. The continuity of the timestamps corresponding to the parameter engineering values is checked and time alignment is performed to obtain a complete dataset under a unified timeline; The complete dataset is analyzed to obtain data features.
7. A knowledge graph-based unmanned aerial vehicle (UAV) fault diagnosis device, characterized in that, include: The vector integration module is used to interpret UAV parameters to obtain data features and integrate symptom vectors based on the data features; The symptom vector contains several reference symptoms; each reference symptom includes reference symptom parameters and the probability of occurrence for each reference symptom parameter. The model building module is used to construct a target knowledge graph model based on data characteristics and reference data; the reference data includes troubleshooting manuals for UAV fault analysis and diagnosis, expert experience, and the system component architecture of UAVs. The node activation module is used to extract target symptoms with a probability of occurrence greater than the activation threshold from the symptom vector, and activate the symptom nodes mapped to the target knowledge graph model to obtain the activated knowledge graph model. The path search module is used to perform path search based on the activated knowledge graph model to obtain several target fault propagation paths for each target fault, and calculate the fault probability of the target fault based on the path information and symptom information of the several target fault propagation paths; the symptom information includes the occurrence probability and number of related symptom nodes. The diagnostic report integration module is used to integrate the target fault and fault probability according to the preset diagnostic report template to obtain the target diagnostic report; The path search module further includes: The path importance assessment unit is used to assess the importance of the target path based on the importance calculation formula and the path weight, path length, probability of occurrence of related symptom nodes, and number of related symptom nodes in the target fault propagation path. The failure probability calculation unit is used to calculate the failure probability of a target failure based on the number of relevant symptom nodes corresponding to several target failure propagation paths and the importance of the target path. The formula for calculating importance is as follows: ； in, This indicates the importance of the target path in the target fault propagation path; This represents the weight of the k-th path in the target fault propagation path; This represents the probability of the occurrence of the i-th related symptom node in the target fault propagation path; This indicates the number of relevant symptom nodes in the target fault propagation path.
8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the knowledge graph-based UAV fault diagnosis method as described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, Used to store computer programs; wherein, when the computer programs are executed by a processor, they implement the knowledge graph-based UAV fault diagnosis method as described in any one of claims 1 to 6.