An intelligent analysis and fault positioning system based on aircraft flight data

Abstract

The application relates to the technical field of data analysis and fault positioning, in particular to an intelligent analysis and fault positioning system based on aircraft flight data, which comprises a data acquisition module, an adaptive data analysis module, a data storage module, an intelligent analysis module, a fault positioning module, a report generation module and an auxiliary module, and is used for data acquisition, adaptive analysis, hybrid storage, intelligent analysis, accurate positioning and customized report generation, so as to solve the problems of poor adaptability, insufficient storage performance, insufficient analysis flexibility, low positioning accuracy and low report practicability in the prior art.

Description

Technical Field

[0001] This invention relates to the field of data analysis and fault location technology, specifically to an intelligent analysis and fault location system based on aircraft flight data. Background Technology

[0002] With the rapid development of the aviation industry, flight safety and operational efficiency have become core industry demands. The massive amounts of FDR and QAR flight data generated during aircraft flight have become the core basis for flight status analysis and fault diagnosis. Currently, various flight data-based processing and fault diagnosis technologies have emerged in the aviation field, covering key aspects such as data parsing, storage, analysis, and fault location. These technologies are widely used in scenarios such as airline operations and maintenance, aircraft manufacturer performance optimization, and aviation safety supervision.

[0003] The patent with publication number CN121009060A, entitled "A QAR data processing method, system, device and medium based on HDF5", describes the following technical solution: acquiring an HDF5 file generated from QAR data, dynamically parsing the file according to a preset event script to obtain configuration data and QAR parameter data, storing the QAR parameter data as a JSON file to form a JSON database, and then reading the corresponding file from the JSON database in parallel according to query requirements to obtain query results, thereby realizing real-time analysis, fault prediction and monitoring and troubleshooting of QAR data.

[0004] This solution has significant limitations: First, it lacks data parsing adaptability, designing parsing logic only for QAR data in HDF5 format, without considering the diverse storage formats (such as binary, ASCII, etc.) and different data transmission standards (such as ARINC429, ARINC664) of flight data from different aircraft models, thus failing to meet the data parsing needs of multiple aircraft models. Second, its data storage architecture is singular, using a JSON database to store parameter data, making it difficult to simultaneously address the differentiated storage needs of static data (such as basic aircraft information) and massive amounts of time-series flight data, resulting in limited performance for high-concurrency read / write and historical data query. Third, it lacks flexible parameter correlation analysis capabilities, failing to support user-configured multi-parameter correlation curve plotting, making it difficult to meet personalized analysis needs in different scenarios.

[0005] The patent with publication number CN120178947A, entitled "Method for Diagnosing Flight Faults of Aircraft", describes a technical solution that involves acquiring raw flight data of an aircraft through QAR, optimizing the raw data to remove irrelevant parameters, using an LSTM model to detect and identify abnormal points, and then using a decision tree fault diagnosis model to determine the fault type, thereby improving the reliability and accuracy of flight fault diagnosis.

[0006] The core shortcomings of this solution are: First, it lacks a comprehensive auxiliary database support system and a dedicated fault feature library to store the parameter characteristics and diagnostic rules corresponding to various faults. Fault localization relies solely on a single decision tree model, making it difficult to accurately determine the faulty component and the specific time point of the fault occurrence. Second, it does not involve automated report generation functionality, so fault diagnosis results cannot be output in a standardized format and do not support custom report content modules, which is not conducive to the rapid application of diagnostic results by maintenance personnel. Third, the data parsing stage only focuses on data optimization and deduplication and lacks the ability to self-learn parsing rules, making it unable to adapt to the flight data protocols of newly added aircraft models.

[0007] The patent with publication number CN119126765B, entitled "A Method and System for Self-Diagnosis of Faults in Avionics Systems", describes a technical solution that involves: dynamically sampling faulty modules of avionics systems, monitoring the correlation of time-series data between modules and generating correlation data, constructing a time-varying causal relationship network between modules, and realizing fault self-diagnosis and generating a report through fault confidence calculation.

[0008] This solution focuses on fault diagnosis of internal modules of avionics systems, but has limitations in scenario adaptability: First, the data acquisition range is narrow, and it does not support multiple interfaces (such as USB and Ethernet) to read FDR and QAR data, resulting in a single data acquisition method; second, the analysis objects are limited to avionics system module data, and do not cover comprehensive flight status parameters and flight command data such as altitude, speed, throttle commands, and control surface commands, resulting in incomplete analysis dimensions; third, it lacks the ability to self-learn parsing rules and dynamically update the algorithm model library, resulting in insufficient system scalability and difficulty in meeting the diagnostic needs of new fault types or new aircraft models.

[0009] The patent with publication number CN119201962A, entitled "A Method and System for Fusion Processing Airport Flight Data Based on Timestamps", describes the following technical solution: real-time capture of data streams from different data sources through a message queue and construction of a database; parsing and extracting key fields and preprocessing them; real-time processing and storage of data after synchronous sorting using timestamps; and data analysis and monitoring of the fusion results to identify potential problems.

[0010] This solution focuses on the fusion and processing of airport flight data, rather than on the specialized analysis and fault location of aircraft flight data. Its core shortcomings include: first, the data parsing lacks adaptive capabilities, failing to design parsing rules for the specific format and protocol of flight data, thus making it impossible to accurately extract flight parameters and flight command data; second, it lacks a dedicated intelligent analysis module, making it unable to perform correlation analysis between flight parameters and commands and automatically generate parameter curves; and third, it lacks fault location functionality, only able to detect potential problems, but unable to complete core fault diagnosis tasks such as fault type identification, fault component location, and fault time point determination.

[0011] Based on the above-mentioned existing technical solutions, it can be seen that the current technical solutions in the field of aviation data processing and fault diagnosis generally suffer from the following common shortcomings: First, the data parsing adaptability is poor, lacking adaptive parsing capabilities for multiple aircraft data formats and multiple transmission standards, as well as the self-learning capability of parsing rules; second, the data storage architecture is difficult to take into account the differentiated needs of static and time-series data, and the performance of high-concurrency read / write and historical data query is insufficient; third, the parameter analysis flexibility is insufficient, failing to support user-configured multi-parameter correlation analysis, making it difficult to meet personalized analysis needs; fourth, the fault location accuracy is limited, lacking comprehensive fault feature library support, making it difficult to accurately locate faulty components and fault time points; fifth, the automation and customization of report generation are low, which is not conducive to the efficient application of diagnostic results.

[0012] Therefore, there is an urgent need for an intelligent analysis and fault location system based on aircraft flight data, capable of data acquisition, adaptive parsing, hybrid storage, intelligent analysis, precise location, and customized report generation, to solve the problems of poor adaptability, insufficient storage performance, insufficient analysis flexibility, priority given to location accuracy, and low report usability of existing technologies. Summary of the Invention

[0013] The present invention aims to provide an intelligent analysis and fault location system based on aircraft flight data, which performs data acquisition, adaptive parsing, hybrid storage, intelligent analysis, precise location, and customized report generation, in order to solve the problems of poor adaptability, insufficient storage performance, insufficient analysis flexibility, priority given to location accuracy, and low report usability of existing technologies.

[0014] This invention provides the following basic solution: an intelligent analysis and fault location system based on aircraft flight data, comprising: a data acquisition module, an adaptive data parsing module, a data storage module, an intelligent analysis module, a fault location module, a report generation module, and an auxiliary module;

[0015] The data acquisition module is used to acquire the raw flight data stored during the aircraft's flight.

[0016] The adaptive data parsing module is used to parse the signal content of the raw flight data according to the flight data storage format and data interface protocol of different aircraft models, and obtain flight parameters.

[0017] The data storage module is used to classify and store the acquired static data and the parsed flight parameters.

[0018] The intelligent analysis module is used to analyze the correlation between flight parameters within a preset time range, and generate correlation analysis conclusions and correlation parameter curves through analysis algorithms; at the same time, it analyzes the influence relationship between flight parameters and identifies abnormal parameter fluctuations during flight.

[0019] The fault location module is used to perform fault analysis and obtain fault information based on the identified abnormal parameter fluctuations and the stored fault patterns through similarity matching and causal analysis.

[0020] The report generation module is used to generate an analysis report based on the flight parameter analysis results, correlation parameter curves, and fault information, according to a preset format. The preset format is a custom report template containing custom content modules. The report generation module contains all the analysis results from the intelligent analysis module for flight parameter analysis results.

[0021] The auxiliary module is used to pre-store parsing rules, analysis algorithms, and fault modes.

[0022] Beneficial effects: This system acquires raw flight data, and based on the preset flight data of different aircraft models, it parses the signal content of the raw flight data, obtains flight parameters, and stores them in categories;

[0023] The intelligent analysis module is constructed to perform intelligent analysis on flight parameters, including: analyzing the correlation between flight parameters within a preset time range, and generating correlation analysis conclusions and correlation parameter curves through analysis algorithms; at the same time, analyzing the influence relationship between flight parameters and identifying abnormal parameter fluctuations during flight.

[0024] Then, through the fault location module, based on the identified abnormal parameter fluctuations and combined with the stored fault patterns, fault analysis is performed through similarity matching and causal analysis to obtain fault information.

[0025] Finally, the report generation module synthesizes the above analysis results, generating an analysis report based on a preset format, which includes flight parameter analysis results, correlation parameter curves, and fault information. The preset format is a custom report template containing custom content modules. Additionally, an auxiliary module is set up for various data storage, specifically pre-storing parsing rules, analysis algorithms, and fault modes for easy retrieval.

[0026] The above modules form a complete technical architecture of data acquisition, adaptive parsing, hybrid storage, intelligent analysis, precise positioning, and customized reports. With the help of auxiliary modules, it can realize adaptive parsing of data from multiple models, differentiated data storage, personalized parameter correlation analysis, precise fault location, and customized report generation, comprehensively solving the problems of adaptability, storage performance, analysis flexibility, positioning accuracy, and report usability of existing technologies. Attached Figure Description

[0027] Figure 1 This is a logic block diagram of an embodiment of the intelligent analysis and fault location system based on aircraft flight data of the present invention;

[0028] Figure 2This is a schematic diagram of the workflow of the adaptive data parsing module in an embodiment of the intelligent analysis and fault location system based on aircraft flight data of the present invention;

[0029] Figure 3 This is a schematic diagram of the correlation parameter curves in an embodiment of the intelligent analysis and fault location system based on aircraft flight data of the present invention;

[0030] Figure 4 This is a schematic diagram illustrating the workflow of an embodiment of the intelligent analysis and fault location system based on aircraft flight data of the present invention. Detailed Implementation

[0031] The following detailed description illustrates the specific implementation method:

[0032] Example 1

[0033] This embodiment provides an intelligent analysis and fault location system based on aircraft flight data, as shown in the attached figure. Figure 1 As shown, it includes: a data acquisition module, an adaptive data parsing module, a data storage module, an intelligent analysis module, a fault location module, and a report generation module; among them, the adaptive data parsing module communicates bidirectionally with the data format adaptation library, the intelligent analysis module communicates bidirectionally with the algorithm model library, and the fault location module communicates bidirectionally with the fault feature library.

[0034] The data acquisition module is used to acquire raw flight data stored during the flight of the aircraft. In this embodiment, the data acquisition module supports reading data files in the flight data recorder (FDR) and quick access recorder (QAR) through various methods such as USB, Ethernet, and dedicated data interface, and is compatible with mainstream data transmission standards such as ARINC429 and ARINC664.

[0035] The adaptive data parsing module is used to parse the signal content of the raw flight data according to the flight data storage format and data interface protocol of different aircraft models to obtain flight parameters. The flight parameters include: aircraft flight parameter data, flight command data, and aircraft status data.

[0036] Specifically, such as Figure 2 As shown, by calling the preset parsing rules in the data format adaptation library of the auxiliary module, the system automatically identifies the format type (such as binary, ASCII, etc.) of the data file of the collected raw flight data, matches the corresponding parsing algorithm, and extracts the aircraft flight parameter data, flight command data, and aircraft status data.

[0037] The aircraft flight parameter data includes, but is not limited to: altitude, speed, heading, engine speed, and fuel consumption;

[0038] Flight command data, including but not limited to: throttle command, left turn command, right turn command, left control surface command, and right control surface command;

[0039] Aircraft status data, including but not limited to: the status of each control surface and the status of the doors;

[0040] In addition, the adaptive data parsing module also has the ability to learn parsing rules on its own, and can update the preset parsing rules to the data format adaptation library after analyzing the data protocols of newly added models.

[0041] The data storage module is used to classify and store the acquired static data and the parsed flight parameters.

[0042] This embodiment adopts a hybrid storage architecture, including: a relational database and a time-series database;

[0043] The relational database stores static data such as basic aircraft information, flight mission information, and interface parsing rule parameters. In this embodiment, the relational database uses a MySQL database. The basic aircraft information refers to design information such as aircraft weight and number of doors, which is manually entered by the user; if there is no information, it is left blank. The flight mission information consists of preset static data such as flight number, route, take-off and landing airports, and flight plan. These are not flight parameters and do not require parsing; they are directly entered and stored. The flight phase (take-off, level flight, etc.) is the result of the intelligent analysis module's analysis of flight parameters and can be used as an auxiliary dimension for correlation analysis. The interface parsing rule parameters store the parsing format and protocol parameters for different aircraft models, providing pre-support for the adaptive data parsing module to match the parsing algorithm.

[0044] The time-series database is used to store aircraft flight parameter data and flight command data, and supports high-concurrency read and write operations as well as fast historical data query functions; in this embodiment, the time-series database is an InfluxDB database.

[0045] Aircraft flight parameter data, aircraft command data, and aircraft status data are all time-series data.

[0046] The data storage module is the system's storage execution module, which is responsible for implementing data storage functions. It stores the parsed flight parameters and static data into the corresponding database according to their types.

[0047] The above parsing rules are custom-developed based on the flight data interface protocol files of different aircraft models. They are the core principles for adapting to the parsing of various raw flight data formats. The rule system includes six elements (six core elements), which work together to achieve accurate parsing of raw flight data, as detailed below:

[0048] Byte offset: Defines the starting byte position of the binary / ASCII data corresponding to each flight parameter in the file stream of the original flight data file, and clarifies the starting point for reading parameter data to avoid data reading misalignment;

[0049] Bit offset: For binary format flight data, further define the start and end bits of parameter data in the corresponding byte. It is suitable for scenarios where multiple parameters share a single byte, and realizes fine-grained bit-level data extraction.

[0050] Signal Name: Map the parsed raw data one by one with standardized flight parameter names to ensure the readability and standardization of the parsing results;

[0051] Signal type: Specifies the storage type of the raw parameter data, providing a basis for subsequent data format conversion and avoiding data type parsing errors;

[0052] Signal parsing methods: Based on the data protocol requirements of the device model, the conversion rules of the original data are defined, which mainly include three categories: two's complement parsing (applicable to parameters with positive and negative values), unsigned number parsing (applicable to non-negative parameters), and mapping relationship parsing (applicable to discrete parameters);

[0053] Signal data range: The normal value range of each flight parameter is preset. During the analysis process, the validity of the data is checked simultaneously, and invalid data that is out of range is removed.

[0054] The parsing algorithm is a dedicated parsing algorithm based on the aircraft flight data interface protocol. The algorithm uses the parsing rules mentioned above in the data format adaptation library as its core basis, and designs differentiated execution logic for different raw flight data formats (binary, ASCII, etc.) to achieve accurate conversion from raw data to standardized flight parameters. The core algorithm flow and specific implementation are as follows:

[0055] Format recognition-driven: The algorithm first reads the header information and data encoding identifier of the original flight data file, and combines them with the format features in the parsing rules to determine the file format type (binary / ASCII) and automatically matches the corresponding parsing branch;

[0056] Bit / byte level data extraction: Based on the byte offset + bit offset in the parsing rules, perform bitwise operations (bitwise AND / bitwise right shift) on binary format data and perform character truncation on ASCII format data to accurately extract the original binary / character data of each parameter;

[0057] Signal type conversion: According to the signal type in the parsing rules, the extracted raw data is converted into a standardized numerical type that the system can recognize, such as converting binary two's complement to signed floating point type and ASCII characters to integer type;

[0058] Protocol rule parsing and calculation: The core data conversion is completed according to the signal parsing method in the parsing rules. The two's complement parsing restores the true value by the one's complement + 1 operation. Unsigned numbers are directly converted into decimal values. The mapping relationship parsing completes the semantic conversion of discrete data through preset key-value pairs.

[0059] Data validity verification: Based on the signal data range in the parsing rules, the converted parameter values ​​are verified, abnormal data that exceeds the normal range is removed, and missing or duplicate data is marked.

[0060] Standardized mapping output: The verified valid data is standardized and mapped according to the signal name, and finally structured aircraft flight parameter data, flight command data, and aircraft status data are output for subsequent modules to process.

[0061] Interface parsing rule parameters refer to a set of standardized configurable parameters, based on the aircraft flight data interface protocol file, used to guide the conversion of raw flight data into standardized flight parameters. These parameters provide a direct basis for the adaptive data parsing module to perform parsing operations. They include six core parameters: byte offset, bit offset, signal name, signal type, signal parsing method, and signal data range. The definitions and functions of each parameter are as follows:

[0062] Byte offset: The starting byte position of the data block corresponding to each flight parameter in the original flight data file stream, used to locate the starting point for reading parameter data;

[0063] Bit offset: In binary format flight data, the start and end positions of each parameter data within the corresponding byte are used to achieve fine-grained bit-level data extraction;

[0064] Signal name: A standardized flight parameter name corresponding to the original data, used to achieve semantic mapping between parsed data and flight parameters;

[0065] Signal type: The data type in which the raw data is stored in the file, including integer, floating-point, Boolean, etc., used to guide the format conversion of the raw data;

[0066] Signal parsing methods: These are the specific rules for converting raw data into flight parameter values ​​with physical meaning, including two's complement parsing, unsigned number parsing, and mapping relationship parsing, which are used to complete the core numerical conversion of the raw data;

[0067] Signal data range: This is the normal physical value range of each flight parameter, including the minimum and maximum values. It is used to verify the validity of the parsed parameter values ​​and remove abnormal data.

[0068] The interface parsing rule parameters mentioned above are stored in the relational database of the data storage module according to the aircraft type. They can be updated synchronously with the interface protocol files of newly added aircraft types, providing rule support for the adaptive data parsing module to adapt to the flight data parsing of multiple aircraft types.

[0069] The intelligent analysis module is used to analyze the correlation between flight parameters within a preset time range, and generate correlation analysis conclusions and correlation parameter curves through analysis algorithms; at the same time, it analyzes the influence relationship between flight parameters and identifies abnormal parameter fluctuations during flight.

[0070] The analysis algorithms include time-series analysis algorithms and correlation analysis algorithms stored in the algorithm model library of the auxiliary module. The time-series analysis algorithms (linear regression, sliding window, etc.) divide flight data into different stages according to time and analyze the trend changes of parameters over time, specifically including: S103 (stage trend determination), S104 (dynamic trend determination), and S105 (visual analysis). The correlation analysis algorithms (Pearson, Spearman, etc.) calculate the correlation between parameters in different stages and determine whether the parameters are abnormal based on the time-series trend, including: S102 (correlation calculation) and S104 (dynamic correlation calculation).

[0071] Correlation parameter curves include, but are not limited to, altitude-time, speed-time, engine speed-fuel consumption, and other correlation parameter curves. Users can also configure and select multiple parameters to plot correlation curves. Figure 3 As shown, the relative pressure altitude-time and calculated vacuum velocity-time curves are used as examples to illustrate the changing trends of flight parameters over time.

[0072] Analyze the relationships between various flight parameters, uncover the inherent patterns of parameter changes, identify abnormal parameter fluctuations during flight, and extract mutually influencing flight parameter sets for users to refer to and select.

[0073] In this embodiment, the negative correlation trend analysis of two parameters, engine speed and fuel flow rate, is used as an example to adapt to the time series characteristics of flight data. The specific process is as follows:

[0074] S101. Preprocess the flight parameters to be analyzed to obtain a dataset of preprocessed flight parameters; the preprocessing includes: format standardization, time dimension alignment, outlier handling, and data annotation.

[0075] The specific processing procedure is as follows:

[0076] Format standardization: Obtain a dataset of several flight parameters to be analyzed (e.g., a dataset of two flight parameters, denoted as parameter A: engine high-pressure rotor speed N2, parameter B: engine fuel flow rate). The dataset includes a timestamp field (accurate to milliseconds) and parameter numerical fields (N2 unit: %RPM, fuel flow rate unit: kg / h); load the dataset into the intelligent analysis module and standardize the data format.

[0077] 1) The timestamp field is converted to UTC time data to remove time breaks caused by flight interruptions, data recording pauses, etc., to ensure the continuity of the time dimension;

[0078] 2) Convert the numerical fields of the timing parameters to floating-point numerical data and remove invalid non-numerical records such as "9999" and "NULL" caused by sensor failure;

[0079] 3) Data is stored in the InfluxDB time-series database. The integrity of core fields such as timestamp, N2 value, and fuel flow value is verified, and null and duplicate records are deleted.

[0080] Time dimension alignment: For the sampling frequency differences of several flight parameters (such as N2 sampling frequency of 10Hz and fuel flow sampling frequency of 5Hz), perform time axis unification processing;

[0081] In this embodiment, the specific details are as follows:

[0082] 1) If the sampling intervals of engine speed and fuel flow are inconsistent, use a linear interpolation algorithm to resample the low-frequency fuel flow data and unify the two time series parameters to the same time granularity (e.g., 1 second / sampling point).

[0083] 2) Generate flight time series with equal time intervals, so that each timestamp corresponds to a unique engine speed and fuel flow value, ensuring that the data points of the two parameters correspond one-to-one and meet the data matching requirements of correlation analysis.

[0084] Outlier handling: using The principle is to remove outliers from the dataset in order to eliminate the interference of abnormal data caused by transient sensor failures, electromagnetic interference, etc. during flight on the analysis results;

[0085] In this embodiment, the specific details are as follows:

[0086] 1) Calculate the average values ​​of engine speed (N2) and fuel flow rate. and standard deviation (Calibration based on historical data from the normal cruise phase);

[0087] 2) Set the outlier threshold range as follows: Remove time-series parameter values ​​that exceed the outlier threshold (such as N2 instantaneously jumping to 120% RPM, fuel flow suddenly dropping to 0, etc.).

[0088] 3) Clean up null values ​​in the dataset after removing outliers and retain valid data points (ensure that the number of data points in a single continuous analysis segment is not less than 100).

[0089] Data standardization: Standardize the flight parameters in the dataset to eliminate the influence of units;

[0090] In this embodiment, due to the significant differences in the dimensions and numerical ranges between engine speed (dimension: %RPM, numerical range: 20%-105%) and fuel flow rate (dimension: kg / h, numerical range: 500-2500 kg / h), Z-score normalization is performed.

[0091]

[0092] in, These are the original flight parameter values ​​(such as the N2 value or fuel flow rate value at a certain moment). This is the average value of this parameter during normal flight. The standard deviation of this parameter is denoted as 0. The standardized data has a mean of 0 and a standard deviation of 1, which eliminates the influence of the unit on the correlation calculation and ensures that the two parameters can be directly quantitatively compared and analyzed.

[0093] S102. Based on the preprocessed dataset, calculate the correlation coefficient and significance between flight parameters. The system analyzes the correlation between flight parameters and outputs the results; the correlation includes linear correlation and nonlinear correlation.

[0094] The specific processing procedure is as follows:

[0095] Linear correlation analysis:

[0096] S10201, Coefficient Calculation: Calculate the Pearson correlation coefficient based on the preprocessed dataset. and significance The values ​​are preprocessed datasets consisting of engine speed datasets and fuel flow datasets.

[0097]

[0098] in, , These are the engine speed and fuel flow rate, respectively. Data points, , These are the average values ​​of engine speed and fuel flow rate, respectively. The total number of valid data points for a single flight segment; the Pearson correlation coefficient is used to characterize the degree of linear correlation between the two parameters (in flight scenarios, under normal circumstances, the fuel flow should increase synchronously when the engine speed increases; if a negative correlation occurs, there may be a fuel system malfunction).

[0099] Significance The value was calculated using a two-tailed test:

[0100]

[0101] in, , for The distribution is greater than One-sided cumulative probability, The value represents the probability that the current correlation is observed to be caused by random fluctuations;

[0102] in The statistic follows a set of degrees of freedom. of Distribution, degrees of freedom :

[0103] .

[0104] S10202, Result Judgment:

[0105] like It falls within the first preset range. The result indicates a strong negative linear correlation (suggesting potential problems such as insufficient fuel pump supply or fuel line blockage). It falls within the second preset range. If it falls within the third preset range, it is determined to be a moderate negative linear correlation; It was determined to be a weak negative linear correlation;

[0106] like Less than the preset significance threshold, The correlation is determined to be statistically significant (excluding spurious correlations caused by random fluctuations in the data); if Greater than or equal to the preset significance threshold The correlation was not statistically significant.

[0107] Based on the judgment results, this embodiment outputs the conclusion that "engine speed and fuel flow are significantly negatively linearly correlated", providing data support for fuel system fault early warning.

[0108] Nonlinear correlation analysis:

[0109] S10203, Rank Transformation and Coefficient Calculation: The flight parameters are rank-transformed, replacing each data point with its rank within the single flight segment dataset. Based on the rank-transformed data, the Spearman rank correlation coefficient is calculated. and significance value;

[0110] In this embodiment, the values ​​of engine speed and fuel flow are rank-transformed to replace each data point with its ranking in the single flight segment dataset (the larger the value, the higher the rank). This is suitable for scenarios where flight parameters have non-linear fluctuations (such as abnormal parameter fluctuations (numerical mutation type) during takeoff and landing).

[0111] Calculate the Spearman rank correlation coefficient based on the rank-transformed data. and significance value:

[0112]

[0113] in, For engine speed and fuel flow rate The rank difference of each data point Given the total number of data points, the Spearman rank correlation coefficient can effectively avoid the interference of instantaneous changes in flight parameters on correlation analysis, and is more in line with the analysis needs of complex flight scenarios.

[0114] S10204. Result Judgment:

[0115] like It falls within the first preset range. This indicates a strong negative monotonic correlation (suggesting a possible persistent fault in the fuel system); if It falls within the second preset range. It is determined to be a moderate negative monotonic correlation; if It falls within the third preset range. It is determined to be a weak negative monotonic correlation;

[0116] like Less than the preset significance threshold, The monotonic correlation was determined to be statistically significant; if Greater than or equal to the preset significance threshold The result was determined to be statistically insignificant.

[0117] Based on the judgment results, this embodiment outputs the conclusion that "engine speed and fuel flow are significantly negatively monotonically correlated," which is applicable to correlation analysis in flight phases with obvious nonlinear trends, such as takeoff and landing.

[0118] Furthermore, nonlinear correlation analysis can also employ the mutual information method. Specifically, for flight scenarios with significant nonlinear trends, such as takeoff and go-around, the mutual information method can be used to analyze the correlation between engine speed and fuel flow, including:

[0119] The engine speed and fuel flow data are discretized (using equal-frequency binning, with each bin containing 50 data points);

[0120] Calculate the mutual information value. The larger the value, the stronger the nonlinear correlation between the two parameters (e.g., the mutual information value should be ≥0.6 during takeoff). This will help verify the correlation characteristics and avoid misjudging nonlinear scenarios by linear analysis methods.

[0121] The aforementioned linear correlation analysis and nonlinear analysis were performed in full as the foundational analyses. Pearson linear correlation analysis and Spearman rank correlation nonlinear analysis were then performed synchronously and indiscriminately on all preprocessed flight parameter datasets, and the correlation coefficients and significance of each type of analysis were calculated. The values ​​are calculated, and statistical significance is determined simultaneously. This allows for a comprehensive exploration of the correlation characteristics between parameters, avoiding misjudgments of parameter correlations by a single analysis method (such as linear correlations among parameters in some flight phases and monotonous nonlinear correlations in others).

[0122] Mutual information method, as a scenario-based supplement, provides targeted enhancement to nonlinear correlation analysis. However, it is not executed across all scenarios, but rather used to supplement and verify scenarios with significant nonlinear fluctuations in flight parameters, such as takeoff, landing, and go-around. For scenarios with stable parameter changes and obvious linear characteristics, such as cruise, the mutual information method can be used as an optional verification method, with Pearson + Spearman analysis results as the core, to further corroborate the nonlinear correlation characteristics.

[0123] The final correlation analysis results integrate multi-dimensional analysis results, including linear correlation, nonlinear correlation (full scenario), and mutual information method (nonlinear scenario). It also outputs coefficients, significance determinations, and scenario-specific features for various analyses, providing a complete parameter correlation basis for subsequent identification of abnormal parameter fluctuations.

[0124] S103. Divide the dataset into different stages according to the flight time dimension, calculate the slope of each stage, perform trend matching judgment based on the slope, determine whether the flight parameters of the stage conform to the preset trend, and output the result.

[0125] The specific process is as follows:

[0126] S10301, Trend Stage Division: The dataset is divided into two typical stages based on the flight time dimension (adapting to standard flight procedures), including: flat segment and changing segment; the trend is determined by calculating the slope of parameter time-value, and if the slope does not meet the preset range, it is determined to be an abnormal parameter fluctuation.

[0127] The flat section refers to the cruise phase (altitude > 9000 meters, Mach number stable at 0.78-0.82), a time interval in which engine speed and fuel flow show no obvious trend changes, and the absolute value of the slope is < the preset absolute value of the slope (parameters are stable). A slope exceeding this range is considered abnormal. In this embodiment, the preset absolute value of the slope is 0.1.

[0128] Variation segment: The descent phase (from cruising altitude to approach altitude) is the time interval in which engine speed shows an upward trend (throttle increases to maintain speed) and fuel flow shows a downward trend (decreasing altitude leads to a decrease in fuel consumption rate). This is a continuous data interval after the flat segment. It requires that the engine speed slope > preset slope (uphill) and the fuel flow slope < preset slope (downhill). Furthermore, the difference between |engine speed slope| / |fuel flow slope| and 1 is less than a preset difference range, that is, the ratio of the absolute values ​​of the two slopes is close to 1. If any condition is not met, it is considered abnormal. In this embodiment, the preset slope is 0.

[0129] S10302, Slope Calculation: Perform linear regression for each stage to calculate the time-numerical slope of the flight parameters; in this implementation, the time-numerical slope of engine speed and fuel flow rate is calculated, specifically including:

[0130] 1) Construct a linear regression model with flight time as the independent variable and parameter values ​​as the dependent variable. ( The slope (Intercept);

[0131] 2) Calculate the slope of engine speed in the cruise (flat section) and descent (changing section) segments respectively. Slope of fuel flow .

[0132] S10303, Trend Matching Determination: Determine whether the slope of each stage conforms to the preset slope range, so as to analyze whether the flight parameters of that stage conform to the preset trend, and output the result.

[0133] The criteria for determining the cruising section (straight section) are as follows: and This means that the engine speed and fuel flow slope are close to 0, which is in line with the expected trend of "parameter stability" during the cruise phase.

[0134] Criteria for determining the descending segment (change segment): (Engine speed increases) and (Fuel flow downhill), and simultaneously calculate the ratio of absolute slope values. The closer the ratio is to 1, the higher the degree of matching between the changes in the two parameters (which is consistent with the flight dynamics characteristics during the descent phase).

[0135] Output the conclusion of "whether the engine speed and fuel flow during the cruise / descent phase are in line with the expected trend". If they are not in line, mark them as abnormal parameter fluctuations (marked as abnormal during the XX flight phase).

[0136] S104. Based on the preprocessed dataset and the divided stages, set a sliding window to calculate the dynamic correlation coefficient and determine the dynamic trend.

[0137] The specific process is as follows:

[0138] S10401, Sliding window parameter setting: The sliding window parameter is set in combination with the flight parameter sampling frequency (1 second / sampling point); in this embodiment, the window size is 10 sampling points (covering 10 seconds of flight data, adapting to the dynamic response characteristics of engine parameters), and the step size is 2 sampling points (the window slides for 2 seconds each time). The window size can be adaptively adjusted according to the sampling frequency of different flight stages (such as adjusting the window size to 5 sampling points during the takeoff stage).

[0139] S10402, Calculation of dynamic correlation coefficient:

[0140] 1) Starting from the beginning of the single flight segment dataset, the sliding window sequentially covers consecutive flight parameter points, namely engine speed and fuel flow data points;

[0141] 2) Calculate the Pearson correlation coefficient for the data of several flight parameters within each window; in this embodiment, the data are for two flight parameters.

[0142] 3) Record the intermediate timestamp and corresponding correlation coefficient of each window to form a dynamic correlation sequence (reflecting the change in the degree of correlation between the two parameters at different flight times).

[0143] S10403, Dynamic Trend Determination: Calculate the mean value of the correlation coefficient of the sliding window within the cruise segment (flat segment) and descent segment (changing segment), determine the correlation of flight parameters based on the mean value, and output the result.

[0144] Specifically, the ideal result for the cruise segment is: the mean value is between -0.3 and 0.3, indicating that there is no obvious linear correlation between the two parameters (which is consistent with the characteristic of parameter stability during the cruise phase). If the result does not fall within the ideal range for the cruise segment, it indicates abnormal parameter fluctuations. If the result exceeds this range, it is considered abnormal.

[0145] Ideal result for the descent phase: mean ≤ -0.7, indicating a strong negative linear correlation (consistent with the change pattern of engine speed and fuel flow during the descent phase). If it does not belong to the ideal result for the descent phase, it indicates abnormal parameter fluctuation, that is, a value greater than this is abnormal.

[0146] Output the conclusion of "whether the dynamic correlation shows a trend reversal of 'weak correlation in the cruising segment and strong negative correlation in the descending segment'". If the trend reversal of "weak correlation in the cruising segment and strong negative correlation in the descending segment" is not shown, it is judged as abnormal parameter fluctuation (indicating faults in engine control logic / fuel system, etc.).

[0147] S105. Based on the flight parameters and the divided phases, draw a dual-axis time-series curve and a scatter plot of the flight parameters, and add a linear fitting line.

[0148] The specific content includes:

[0149] S10501. Draw a dual-axis time-series curve: the horizontal axis is flight time (UTC time), and the vertical axis is flight parameters; in this embodiment, the left vertical axis is engine speed (N2, unit: %RPM), and the right vertical axis is fuel flow rate (unit: kg / h). Mark the boundary points of different stages. In this embodiment, it is the boundary point of "cruise-descent". Visually show the trend matching degree of the two parameters in different flight stages, and provide visual support for maintenance personnel to quickly make judgments.

[0150] S10502, Scatter Plot and Linear Fitting: Plot a scatter plot of engine speed versus fuel flow rate and add a linear fitting line:

[0151] 1) If the scatter points are distributed downward from left to right (decreasing segment data) and the slope of the fitted line is negative, verify the negative correlation trend between the two parameters;

[0152] 2) The closer the coefficient of determination R² of the fitted line is to 1, the stronger the linear correlation (in the normal decreasing phase). (Should be ≥0.8), if If the value is less than 0.5, it is considered an abnormal parameter fluctuation, indicating the presence of abnormal interference factors.

[0153] S10503 and the above-mentioned visualization features (dual-axis time series curves and scatter plots of flight parameters) serve as auxiliary criteria for judging abnormal parameter fluctuations, and corroborate the quantitative results of S103 and S104.

[0154] The fault location module is used to perform fault analysis and obtain fault information based on the identified abnormal parameter fluctuations and the stored fault patterns through similarity matching and causal analysis.

[0155] The fault location module, based on the abnormal parameter fluctuations identified by the intelligent analysis module and combined with the fault patterns in the fault feature library of the auxiliary module (such as engine performance degradation, avionics system anomalies, control surface failure, abnormal door operation, etc.), uses algorithms such as similarity matching and causal analysis to locate the fault, determine the fault type, possible faulty components, and the specific time point of the fault occurrence, thereby completing the fault analysis and generating the fault location result as fault information.

[0156] The specific process is as follows:

[0157] S201, the fault location module receives abnormal parameter fluctuations output by the intelligent analysis module; the abnormal parameter fluctuations include abnormal parameter identifiers, time series characteristics of parameter fluctuations, and parameter anomaly levels; the intelligent analysis module completes the extraction, calculation, and classification of the above data during the analysis process of S101-S105, specifically, from the preprocessed flight parameter dataset with timestamps, stage trend determination results, and dynamic correlation analysis results, it extracts the main abnormal parameter (i.e., abnormal parameter fluctuations) as abnormal parameter identifiers, obtains the fluctuation start time, peak time, fluctuation amplitude, and trend change characteristics as time series characteristics of parameter fluctuations, and classifies the parameter anomaly levels as slight deviation, moderate anomaly, and severe over-limit according to the degree of parameter deviation from the preset threshold.

[0158] Meanwhile, the fault location module retrieves structured fault mode data from the fault feature library of the auxiliary module; the fault mode data includes: fault type labels, parameter fluctuation feature templates corresponding to typical faults, a list of fault-related components, and typical time / operating condition features of the fault occurrence; among them, fault type labels include: engine performance degradation, avionics system anomaly, control surface failure, and abnormal door operation.

[0159] S202. Preprocessing is performed on abnormal fluctuations in the input parameters; the preprocessing includes: time standardization, normalization and classification.

[0160] Specifically, the timestamps of abnormal parameter fluctuations are uniformly converted into the system's standard time format to ensure consistency in the time dimension;

[0161] Normalize the time series characteristics of parameter fluctuations (e.g., convert the fluctuation amplitude into a percentage relative to the benchmark value) to eliminate the differences in the dimensions of different parameters.

[0162] Based on the preset classification rules stored in the fault feature library of the auxiliary module, abnormal parameters are classified (such as power system parameters, avionics system parameters, control system parameters, and structural system parameters), and classification results are generated. The classification results are precisely matched with the fault mode classification system (the organizational framework of structured fault mode data) in the fault feature library, and then the structured fault mode data is retrieved under the corresponding classification framework to achieve targeted screening of fault modes.

[0163] Abnormal parameters are classified according to preset classification rules (such as power system parameters, avionics system parameters, control system parameters, and structural system parameters), and classification results are generated. The classification results are precisely matched with the fault mode classification system (the organizational framework of structured fault mode data) in the fault feature library, and then the structured fault mode data is retrieved under the corresponding classification framework to achieve targeted screening of fault modes.

[0164] The preset classification rules are stored in the fault feature library of the auxiliary module and are linked with the fault mode classification system to provide a unified basis for classifying abnormal parameters. The core source of abnormal parameter fluctuations is the abnormal parameter itself. The essence of classification is to define the system / category to which the fluctuating parameter belongs (e.g., engine speed belongs to the power system parameter, and control surface angle belongs to the control system parameter), rather than classifying the fluctuation itself. The fault mode classification system is the organizational framework of structured fault mode data, and the structured fault mode data is the specific data carrier of the fault mode classification system. The two are subordinate to each other, with the classification rule framework and the structured data set under the framework being subordinate to each other. Both are stored in the fault feature library of the auxiliary module, working together to support the rapid screening and matching of fault modes.

[0165] S203. Fault Feature Database Call and Matching Preparation: The indexing system of the fault feature database is pre-built and stored in the auxiliary module's fault feature database. The index dimensions include parameter type, fault type, component affiliation, and operating conditions. Based on the classification results of abnormal parameters in S202, fault mode data under the same category are selectively filtered from the structured fault mode data retrieved in S201 to form a candidate fault mode set, narrowing down the scope of subsequent similarity matching. For example, if the abnormal parameter is an engine thrust parameter, then power system-related fault modes such as "engine performance degradation" are filtered out, while irrelevant fault modes such as avionics system and control surface system are excluded.

[0166] S204. Feature Extraction and Quantization: Extract feature vectors of abnormal parameter fluctuations, including: time features, numerical features, and correlation features;

[0167] Timing characteristics of parameter anomalies:

[0168]

[0169] These are the start time of the fluctuation, the peak time, and the duration;

[0170] Numerical characteristics of the parameters:

[0171]

[0172] These are the magnitude of deviation from the benchmark value, the slope of the trend change, and the frequency of fluctuation, respectively.

[0173] Parameter correlation characteristics:

[0174]

[0175] That is, the abnormal correlation between this parameter and other parameters that are linked to it. For association features The first in The component represents the current abnormal parameter and the first component. Abnormal correlation between related parameters;

[0176] The core of association features is to quantify the degree of synchronous deviation of other parameters when the target parameter fluctuates abnormally, through temporal similarity / correlation measures. The specific acquisition process includes:

[0177] S20401, Abnormal Time Period Extraction: Based on Time Features Extract target parameters from full flight data Associated parameters with all candidates Including temporal subsequences;

[0178] S20402, Reference Normalization: For the time series subsequences of each parameter, standardization is performed with reference to the mean / reference value under normal operating conditions to eliminate dimensional differences.

[0179]

[0180] in For the first The parameters at time... The standardized value, For the first The parameters at time... The original observations, For the first The baseline mean of each parameter, For the first The benchmark standard deviation of each parameter;

[0181] S20403, Association Calculation: For the target parameter and each candidate association parameter, calculate the temporal similarity to obtain... The final combination is The temporal similarity is calculated using Euclidean distance or DTW (Dynamic Time Warping).

[0182] Simultaneously, typical feature vectors of each fault mode in the candidate fault mode set are extracted. This forms a fault characteristic template; among which The time feature is a typical feature vector. These are the numerical features in the typical feature vector. These are the associated features in the typical feature vector.

[0183] S205. Similarity Calculation: The weighted Euclidean distance algorithm is used to calculate the similarity between the feature vector of abnormal parameter fluctuations and the fault feature template. The calculation formula is as follows:

[0184]

[0185] Euclidean distance of time feature vectors:

[0186]

[0187] Euclidean distance for numerical eigenvectors:

[0188]

[0189] The Euclidean distance between the associated feature vectors;

[0190] , , These are the weighting coefficients for each feature dimension, which can be dynamically adjusted according to the fault type (e.g., numerical features have higher weights in engine faults, while time features have higher weights in door faults).

[0191] Similarity value , The closer it is to 1, the higher the degree of matching between the abnormal parameters and the failure mode.

[0192] S206. Initial Screening Results Output: Set Similarity Threshold Fault modes with similarity values ​​greater than a similarity threshold are selected to form a fault candidate list. In the subsequent causal analysis stage, the fault modes in this candidate list are precisely traced back to their source. A comprehensive score is calculated through a parameter-fault-component causal relationship network to determine the final fault type and component. If the similarity values ​​are all below the threshold, they are marked as "unknown fault modes". In the causal analysis stage, a full-dimensional investigation is carried out on this unknown anomaly. The root cause of the abnormal parameter fluctuations is traced back through the causal relationship network to uncover potential unrecorded fault modes or compound faults to avoid missed judgments. All judgment results are further investigated in the subsequent causal analysis stage.

[0193] S207. Construct a causal relationship network of parameters, faults, and components. The network nodes include abnormal parameters, candidate fault modes, and associated components. The edge weight is the causal relationship strength between the parameter and the fault / component (pre-calibrated based on historical fault data and system design manual). For example, the fault node "engine thrust reduction" is associated with component nodes such as "high pressure compressor", "fuel nozzle" and "turbine blade", and is also associated with parameter nodes such as "engine speed" and "fuel flow".

[0194] S208. Based on the causal relationship network, trace back to the root cause of abnormal parameter fluctuations;

[0195] Specifically, it includes:

[0196] 1) Starting with the abnormal parameter as the starting node, traverse the associated fault nodes and calculate the abnormal contribution of each fault node (i.e., the probability that the current parameter will become abnormal when the fault occurs).

[0197] 2) For each fault node, further traverse the associated component nodes and calculate the failure probability of each component (considering factors such as component failure rate, runtime, and environmental conditions).

[0198] 3) Construct a fault propagation path scoring model that integrates anomaly contribution, component failure probability, and parameter fluctuation feature matching degree to generate a comprehensive score for each fault-component combination. The fault propagation path scoring model is a weighted comprehensive scoring model that uses the causal correlation strength of fault propagation as the core weight basis. It establishes a three-dimensional weighted calculation structure of "anomaly contribution - component failure probability - parameter fluctuation feature matching degree". It integrates the three core indicators of anomaly contribution, component failure probability, and parameter fluctuation feature matching degree and generates a comprehensive score for each fault-component combination through a linear weighted algorithm. This enables quantitative tracing of fault propagation paths, significantly improves the accuracy and reliability of fault-component matching, and provides a quantitative basis for fault location.

[0199] Specifically, the fault propagation path scoring model is a three-level weighted quantization structure. It takes the fault-component combination as the scoring object and sets up a three-layer architecture of indicator layer, weight layer and calculation layer. The specific settings, quantification rules and data sources of each layer are deeply related to the modules mentioned above to ensure the feasibility of the model.

[0200] The indicator layer comprises three core quantitative indicators, all normalized values ​​between 0 and 1. Higher indicator values ​​indicate stronger support for fault identification in that dimension. The indicator data originates from the aforementioned intelligent analysis module or the fault feature library.

[0201] 1) Abnormal contribution : 0-1, the probability that a fault node will cause the current abnormal parameter fluctuation. The data comes from the statistical analysis of historical fault data in the fault feature library. The preset base value is determined by the system according to the fault type.

[0202] 2) Component failure probability : 0-1, the actual failure probability of the component associated with the fault node is calculated in real time by the system combining the component's basic failure rate (fault feature library) + component runtime (data storage module) + flight conditions (intelligent analysis module);

[0203] 3) Matching degree of parameter fluctuation characteristics : 0-1, the similarity between abnormal parameter fluctuations and fault mode feature templates, that is, the similarity value calculated by the weighted Euclidean distance algorithm in the fault location module S205, which is directly derived from the previous steps.

[0204] Weighting layer: Configures dynamic weighting coefficients for the three indicators ( , , ),satisfy The weighting coefficients are dynamically adjusted based on the system type to which the fault belongs. In this embodiment, the preset basic weights and dynamic adjustment rules are as follows:

[0205] 1) Basic weights: (Abnormal contribution) = 0.3 (Component failure probability) = 0.4 (Parameter fluctuation feature matching degree) = 0.3;

[0206] 2) Powertrain system failures (e.g., engine, fuel system): Increase the weight of feature matching. =0.5, =0.2、 =0.3;

[0207] 3) Avionics / control system failures (e.g., control surfaces, avionics voltage): Increase the weighting of component failure probability. =0.5, =0.2, ω3=0.3;

[0208] 4) Structural system failures (such as hatches, fuselage structure): Increase the weight of the anomaly contribution. =0.4, =0.3、 =0.3.

[0209] Calculation layer: Set up a linear weighted calculation core unit, configure an independent calculation channel for each fault-component combination, and output a comprehensive score value of 0-1. The higher the score value, the greater the possibility that the combination is the root cause of the fault.

[0210] 1) Comprehensive score calculation formula: ( For the overall score, 0≤ ≤1, the higher the score, the greater the likelihood of the root cause of the fault.

[0211] 2) Rules for judging scoring results: ≥0.8 indicates a high-confidence fault-component combination; 0.5≤ <0.8 indicates a medium confidence level. <0.5 indicates a low confidence level;

[0212] 3) Fault screening rules: Select only Combinations with a value ≥0.5 are included in the suspected fault set, ranking first and... A value ≥0.8 indicates a final faulty component combination.

[0213] The specific execution process involves extracting indicator data, matching dynamic weights, calculating comprehensive scores, and sorting scores. This process is executed synchronously with the traversal of the causal relationship network. Each fault-component combination is traversed to complete one score calculation, ensuring the real-time nature of fault location.

[0214] S209. Based on the comprehensive scores, sort the fault types from high to low, select the fault type with the highest score as the final fault type, and the corresponding component is the possible fault component; at the same time, combine the start timestamp of the abnormal parameter fluctuation to determine the specific time node when the fault occurred.

[0215] If there are multiple high-scoring fault-component combinations (score difference < 0.1), they are marked as a suspected fault set, and the confidence level of each candidate is indicated.

[0216] The report generation module is used to generate an analysis report based on the flight parameter analysis results, correlation parameter curves, and fault information according to a preset format (Word). It supports custom report templates and allows users to select the content modules to be included in the report (such as flight status overview, parameter trend analysis, fault details, etc.). Its flight parameter analysis results are a collective term for all quantitative conclusions, judgment results, and feature analysis results output by the intelligent analysis modules S101-S105 after analyzing the flight parameters. It is an overall summary of the core analysis results of the intelligent analysis modules mentioned above, including: quantitative judgment conclusions on parameter correlation, flight phase trends, dynamic correlation, and abnormal parameter fluctuation identification results.

[0217] The specific process is as follows:

[0218] S301. Receive the request and perform requirement parsing, including:

[0219] 1) The front-end interface receives user operation instructions as requests, which include three core parameters: target report template ID (standard template / custom template), flight data analysis task ID to be integrated (analysis results associated with specific flights), and custom content selection (select the content modules to be included, such as selecting only "flight status overview + parameter trend analysis", or selecting all modules).

[0220] 2) The backend module validates the request parameters to confirm that the template ID is valid, the task ID associated data is complete, and the selected content module is conflict-free (e.g., when the "fault details module" is selected, it verifies whether the task has fault location results). After the validation is successful, a unique report is generated to generate the task ID, and the process proceeds to the next step.

[0221] S302. Based on the parsing requirements, perform multi-source data aggregation and format standardization, including:

[0222] 1) Based on the task ID, retrieve target data from each module; specifically, obtain basic flight information and parsing rule version from the relational database; extract flight parameters from the time series database; obtain correlation parameter curves (in image format, converted to Base64 format for easy embedding in Word) and correlation analysis conclusion text from the intelligent analysis module; obtain fault information, i.e., fault details JSON data (marked as "no abnormal fault record" when there is no fault) from the fault location module.

[0223] 2) Standardize the retrieved data and aggregate it to generate a structured data dictionary. Specifically, standardize the aggregated data by retaining two decimal places for numerical data (e.g., a correlation coefficient of 0.876 is converted to 0.88), standardizing the time format to "YYYY-MM-DD HH:MM:SS" (e.g., converting the fault occurrence timestamp to a specific time), and removing special characters from text data (to avoid Word formatting errors). Finally, a structured data dictionary (key-value pair format, where the key is the content module name and the value is the corresponding standardized data) is generated.

[0224] S303. Based on the parsing requirements, perform dynamic assembly and filling of content based on templates, including:

[0225] 1) Load the corresponding template file (Word format) based on the template ID, parse the template's XML structure (using Java office document processing tools such as POI and Docx4j), and identify replaceable placeholders in the template (such as "{{flight number}}", "{{parameter curve image}}", "{{faulty component}}", etc.);

[0226] 2) Based on the user-selected content modules, fill in the content according to the chapter order of the content template; specifically, for text data (such as flight numbers, fault types), directly replace the corresponding placeholders; for image data (such as parameter curves), insert Base64 format images at the preset positions of the template, and automatically adjust the image size to fit the page (such as setting the width to 80% of the page width); for tabular data (such as abnormal parameter lists), dynamically generate Word tables based on the list data in the data dictionary (automatically match the number of columns, fill in the data, and set the table border style);

[0227] 3) If the user selects some content modules, the placeholders corresponding to the unselected modules will be automatically skipped, and the continuity of the chapter numbers will be adjusted (e.g., after skipping the "Fault Details Module", the subsequent chapter numbers will be sequentially extended).

[0228] S304. For the content template, perform text (Word) format rendering and style optimization, generating a temporary text file, including:

[0229] 1) Standardize the format of the assembled content and generate a temporary text file; specifically, standardize the heading styles of all chapters (e.g., first-level headings in bold, size 2; second-level headings in bold, size 3), body text styles (SimSun, size 4, line spacing 1.5), table styles (uniform border color and table header background color), and image styles (add image descriptions, such as "..."). Figure 3 Engine speed-fuel flow time sequence curve);

[0230] 2) Handle special formatting requirements; specifically, add highlight marks (such as yellow background) to key data (such as fault occurrence time, abnormal parameter values), add page breaks at the end of chapters to ensure chapter independence, insert report generation time and task ID in the header, and insert system name and copyright information in the footer.

[0231] S305. For temporary text files, perform report preview and output storage, including:

[0232] 1) Provide an online preview function for temporary text files (stored in the system's temporary directory) through the front-end interface, allowing users to view the completeness of the content and the correctness of the format;

[0233] 2) Upon receiving user confirmation (after previewing and confirming that everything is correct), output a temporary text file as a report. This embodiment provides multiple output methods, including but not limited to: local download (the front end returns a file download link, allowing users to select a local save path), server storage (stored to the system file server according to the naming rule of "flight number-generation time", and an access link is generated and associated with the user's account), and email push (allows users to enter their email address and automatically sends the report as an attachment).

[0234] 3) Generate report and log, recording information such as task ID, template ID, output method, storage path / email address, and generation time, to facilitate subsequent system operation and maintenance and traceability.

[0235] S306. Exception Handling and Retry Mechanism: If an exception occurs during report generation (such as data retrieval failure, template parsing error, or file storage failure), the exception handling process will be automatically triggered.

[0236] 1) Record exception information (including error code, error description, and steps in which it occurred);

[0237] 2) For recoverable exceptions (such as database connection timeout), automatically retry twice (with a 3-second retry interval);

[0238] If the retry fails, return a generation failure message to the front end, provide "Regenerate" or "Contact Operations" options, and push the exception log to the system operations and maintenance backend.

[0239] The auxiliary modules (data format adaptation library, algorithm model library, and fault feature library) are used to parse rules, analyze algorithms, fault modes, and supporting data or models.

[0240] The data format adaptation library stores information such as flight data format rules, interface protocol parameters, and parsing algorithms for different aircraft models.

[0241] Algorithm model library, storing various algorithm models and parameters such as time series analysis, correlation analysis, and curve fitting;

[0242] The fault feature library stores data such as common aircraft fault types, corresponding parameter characteristics, and fault diagnosis rules.

[0243] The hardware model and software type used in the specific implementation process in this embodiment are as follows:

[0244] Hardware selection: Industrial control computer (model: Advantech IPC-610L, configured with Intel Core i7 processor, 16GB memory, 1TB SSD + 4TB HDD); data acquisition card (model: NI PCIe-6570); Ethernet switch (model: Huawei S1720-28GWR-4P); storage server (model: Dell PowerEdge R750, configured with 24 8TB hard drives).

[0245] Software selection: Operating system is Windows 10; databases used are MySQL 8.0 (relational database) and InfluxDB 2.7 (time series database); front-end development framework is Qt 5.9.9; back-end development framework is Spring Boot 2.7; algorithm libraries used are Python Scikit-learn and TensorFlow (for some intelligent analysis algorithms); report generation components used are iText and POI.

[0246] The specific workflow of this system is as follows: Figure 4 As shown, its beneficial effects include:

[0247] Significantly Improved Data Adaptability and Parsing Accuracy: This system, through the bidirectional linkage between the adaptive data parsing module and the data format adaptation library, combined with the self-learning capability of parsing rules, can be compatible with FDR / QAR flight data from different aircraft types (such as Boeing 737, Airbus A320, and domestically produced large aircraft), different storage formats (binary, ASCII, etc.), and different transmission standards (ARINC429, ARINC664, etc.), enabling rapid adaptation to new aircraft types. This solves the pain points of existing technologies, such as narrow flight data parsing adaptation range, insufficient compatibility with new aircraft types and data formats, and the need for manual redevelopment of parsing rules. Actual testing has verified that the system achieves a parsing accuracy of over 99.8% for mainstream civil aviation aircraft flight data, and shortens the adaptation cycle for new aircraft types to one-third of existing technologies, significantly improving the system's versatility in aviation operations and maintenance scenarios.

[0248] Significantly improved data analysis efficiency and automation: This system achieves fully automated processing from flight data acquisition, adaptive parsing, and categorized storage to intelligent analysis, fault location, and report generation. Core tasks such as parameter curve plotting, anomaly identification, and fault tracing are completed without manual intervention. Compared to the traditional model of "manual data reading + basic tool analysis + manual report compilation," data analysis efficiency is improved by over 80%. Taking flight data with a single flight duration of 2 hours as an example, existing technologies require 2-3 hours to complete the entire analysis process, while this invention can complete all analysis work and generate a standardized report in just 20-30 minutes, significantly shortening the data analysis cycle for aviation operations and maintenance and saving time for rapid fault handling.

[0249] Significant improvements in fault location accuracy and latent fault identification capabilities: This system utilizes an intelligent analysis module to uncover the correlations between flight parameters (such as the dynamic correlation between engine speed and fuel flow). Combined with the synergistic effect of the fault location module and the fault feature database, and employing algorithms such as similarity matching and causal analysis, it can accurately locate fault types (such as engine performance degradation, fuel system blockage, control surface failure, etc.), faulty components (such as fuel injectors, fuel pumps, control surface actuators, etc.), and the specific time point of the fault occurrence (accurate to the second). This solves the problems of existing technologies relying on manual experience judgment, low fault location accuracy, and the tendency to miss latent faults (such as faults caused by potential abnormal parameter fluctuations). Verified through multiple real-world aviation maintenance cases, this system achieves a fault location accuracy rate of over 95% for common aircraft faults and improves the identification capability of latent faults by 60% compared to existing technologies, effectively reducing the risk of ineffective maintenance due to inaccurate fault location.

[0250] The system's versatility and flexibility have been significantly enhanced: Supporting multiple deployment options (local deployment, cloud-based collaborative deployment, and lightweight embedded deployment), it can meet the needs of different user groups. Large airlines can achieve centralized management and global analysis of multi-location, multi-flight data through cloud-based collaborative solutions; small and medium-sized airlines can reduce costs through local lightweight deployment; and frontline maintenance personnel can quickly conduct specialized troubleshooting through customized analysis functions. Simultaneously, the system supports user-defined combinations of analysis parameters (such as "engine speed-fuel flow" and "control surface commands-flight attitude angles"), personalized report template selection, and customized fault diagnosis models. Compared to the fixed functional modes of existing technologies, its flexibility is greatly improved, adapting to the differentiated needs of various scenarios such as aviation maintenance, aircraft development, and safety supervision.

[0251] The system effectively reduces aviation operation and maintenance costs and resource consumption. On the one hand, the hardware cost of the lightweight deployment solution is only 1 / 5 to 1 / 3 of that of existing commercial flight data analysis software. The cloud-based collaborative solution enables multi-user resource sharing, avoiding the waste of resources caused by repeated deployment of equipment and development of analysis tools by various units. On the other hand, the fully automated processing significantly reduces the labor costs required for manual analysis and reduces the cost of analysis errors caused by human operation. At the same time, accurate fault location avoids the additional costs of component replacement and labor time consumption caused by ineffective maintenance, significantly improving the economic benefits of aviation operation and maintenance.

[0252] Supporting both remote and centralized management to enhance overall operational coordination: This system's cloud-based collaborative deployment solution enables centralized storage and remote analysis of flight data from multiple airports and flights, facilitating comprehensive flight safety monitoring and operational coordination management for airlines and regulatory authorities. By aggregating flight parameter analysis results and fault location information from each flight in real time, it can quickly identify common potential faults in a particular aircraft model or component, providing data support for aircraft manufacturers to optimize product performance and airlines to develop targeted operational strategies, thereby helping to improve flight safety management across the entire aviation industry.

[0253] Stable and reliable data storage and query performance: This system adopts a hybrid storage architecture of "MySQL relational database + InfluxDB time-series database", which is adapted to the storage needs of static data (aircraft basic information, parsing rules, etc.) and massive time-series flight data (altitude, speed, engine parameters, etc.), supporting high-concurrency read and write operations and fast historical data queries. It solves the problems of low data storage efficiency and slow historical data query caused by the use of a single database in existing technologies, ensuring that the system can maintain stable operating performance when processing a large amount of historical flight data, and providing reliable support for long-term flight data traceability analysis.

[0254] Compared with existing technologies, this solution adopts an adaptive data parsing mechanism, combined with a data format adaptation library and self-learning capabilities, which greatly improves the adaptability to different models and data formats, and solves the problem of poor adaptability of existing technologies; while existing technologies mostly parse fixed formats, with limited adaptability.

[0255] The hybrid storage architecture selects an appropriate database for different types of data, which balances the structured storage of static data and the efficient storage and retrieval of time-series data, outperforming the performance of existing single storage architectures. Existing technologies mostly use a single database, which makes it difficult to achieve efficient storage of massive amounts of time-series data.

[0256] By integrating intelligent analysis and fault location algorithms, the system automates flight status analysis, parameter correlation mining, and precise fault location without human intervention, thus improving analysis efficiency and accuracy. Existing technologies often rely on manual analysis, which is inefficient and prone to errors.

[0257] It supports custom report templates, which can generate personalized analysis reports according to different user needs, making it more flexible; existing technology reports have fixed formats, which are difficult to meet diverse needs.

[0258] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. An intelligent analysis and fault location system based on aircraft flight data, characterized in that, include: The data acquisition module is used to acquire the raw flight data stored during the aircraft's flight. The adaptive data parsing module is used to parse the signal content of the raw flight data according to the flight data storage format and data interface protocol of different aircraft models, and obtain flight parameters. The data storage module is used to classify and store the acquired static data and the parsed flight parameters. The intelligent analysis module is used to analyze the correlation between flight parameters within a preset time range, and generate correlation analysis conclusions and correlation parameter curves through analysis algorithms; at the same time, it analyzes the influence relationship between flight parameters and identifies abnormal parameter fluctuations during flight. The fault location module is used to perform fault analysis and obtain fault information based on the identified abnormal parameter fluctuations and the stored fault patterns through similarity matching and causal analysis. The report generation module is used to generate analysis reports from flight parameter analysis results, correlation parameter curves, and fault information according to a preset format. The preset format is a custom report template, which includes custom content modules; the intelligent analysis module for flight parameter analysis results contains all analysis results. The auxiliary module is used to pre-store parsing rules, analysis algorithms, and fault modes; The intelligent analysis module performs analysis and identification, including: S101. Preprocess the flight parameters to be analyzed and obtain the dataset of preprocessed flight parameters; S102. Based on the preprocessed dataset, calculate the correlation coefficient and significance between flight parameters. Values, analyzing the correlation between flight parameters; S103. Divide the dataset into different stages according to the flight time dimension, calculate the slope of each stage, perform trend matching judgment based on the slope, determine whether the flight parameters of the stage conform to the preset trend, and output the result. S104. Based on the preprocessed dataset and the divided stages, set a sliding window to calculate the dynamic correlation coefficient and determine the dynamic trend. S105. Based on the flight parameters and the divided phases, draw a dual-axis time-series curve and a scatter plot of the flight parameters, and add a linear fitting line. The fault location module performs fault analysis, including: S201. The fault location module receives abnormal parameter fluctuations output by the intelligent analysis module; wherein the abnormal parameter fluctuations include: abnormal parameter identifier, time series characteristics of parameter fluctuations, and parameter abnormality level; Simultaneously, structured fault mode data is retrieved; the fault mode data includes: fault type labels, parameter fluctuation feature templates corresponding to typical faults, a list of fault-related components, and typical time characteristics / operating condition characteristics of fault occurrence. S202. Preprocessing is performed on the abnormal fluctuations of the input parameters; the preprocessing includes: classification, classifying the abnormal parameters according to preset classification rules, and generating classification results; S203. Based on the classification results of abnormal parameters, filter out the fault mode data under the same category from the retrieved structured fault mode data to form a candidate fault mode set. S204. Extract the feature vector of abnormal parameter fluctuations, extract the typical feature vector of each fault mode in the candidate fault mode set, and form a fault feature template. S205. The similarity between the feature vector of abnormal parameter fluctuations and the fault feature template is calculated using the weighted Euclidean distance algorithm. S206. Set a similarity threshold, filter out fault modes with similarity values ​​greater than the similarity threshold, and form a fault candidate list; if the similarity values ​​are all below the threshold, mark them as unknown fault modes. S207. Construct a causal relationship network of parameters, faults, and components. The network nodes include abnormal parameters, candidate fault modes, and associated components. The edge weights are the strength of the causal relationship between the parameters and the faults / components. S208. Based on the causal relationship network, trace back to the root cause of abnormal parameter fluctuations and obtain the comprehensive score of each fault-component combination. S209. Based on the comprehensive scores, sort from high to low, select the fault type with the highest score as the final fault type, and the corresponding component is the possible fault component; at the same time, combine the start timestamp of the abnormal parameter fluctuation to determine the specific time node of the fault occurrence; if there are multiple high-scoring fault-component combinations, mark them as a suspected fault set, and indicate the confidence level of each candidate.

2. The intelligent analysis and fault location system based on aircraft flight data according to claim 1, characterized in that, The flight parameters include: aircraft flight parameter data, flight command data, and aircraft status data; The aircraft flight parameter data includes: altitude, speed, heading, engine speed, and fuel consumption; Flight command data includes: throttle command, left turn command, right turn command, left control surface command, and right control surface command. Aircraft status data, including: the status of each control surface and the status of the doors; The adaptive data parsing module automatically identifies the format type of the data file of the collected raw flight data by calling preset parsing rules, matches the corresponding parsing algorithm, and extracts aircraft flight parameter data, flight command data, and aircraft status data. The adaptive data parsing module is also used to update the preset parsing rules to the auxiliary module after analyzing the data protocol of newly added models.

3. The intelligent analysis and fault location system based on aircraft flight data according to claim 2, characterized in that, The data storage module adopts a hybrid storage architecture, including: a relational database and a time-series database; The relational database is used to store static data, including: basic aircraft information, flight mission information, and interface parsing rule parameters; A time-series database is used to store the parsed flight parameters.

4. The intelligent analysis and fault location system based on aircraft flight data according to claim 1, characterized in that, The correlation includes: linear correlation and nonlinear correlation; Linear correlation analysis: S10201. Calculate the Pearson correlation coefficient based on the preprocessed dataset. and significance value; S10202, if If it falls within the first preset range, it is determined to be a strong negative linear correlation; if If it falls within the second preset range, it is determined to be a moderate negative linear correlation; if it falls within the third preset range, it is determined to be a weak negative linear correlation. like If the correlation is less than the preset significance threshold, it is considered statistically significant; if... If the correlation is greater than or equal to the preset significance threshold, the correlation is determined to be statistically insignificant. Combine the judgment results and output them; Nonlinear correlation analysis: S10203. Perform rank transformation on the flight parameters, replacing each data point with its rank within the single flight segment dataset. Calculate the Spearman rank correlation coefficient based on the rank-transformed data. and significance value; S10204, if If it falls within the first preset range, it is determined to be strongly negative monotonic correlated; if If it falls within the second preset range, it is determined to be of moderate negative monotonic correlation; if It falls within the third preset range and is determined to be weakly negative monotonic correlated. like If the value is less than the preset significance threshold, the monotonic correlation is considered statistically significant; if... If the value is greater than or equal to the preset significance threshold, it is determined that there is no statistical significance. Combine the judgment results and output them; The nonlinear correlation analysis also includes: employing the mutual information method, specifically including: Discretize the flight parameters; Calculate the mutual information value. The larger the value, the stronger the nonlinear correlation between the two flight parameters, which further verifies the correlation characteristics.

5. The intelligent analysis and fault location system based on aircraft flight data according to claim 1, characterized in that, S103 includes: S10301. Based on the flight time dimension, the dataset is divided into two typical stages, including: a flat segment and a changing segment; S10302. Perform linear regression for each stage to calculate the time-numerical slope of the flight parameters. S10303. Determine whether the slope of each stage meets the preset slope range, so as to analyze whether the flight parameters of that stage meet the preset trend, mark the corresponding abnormal parameter fluctuations, and output the results.

6. The intelligent analysis and fault location system based on aircraft flight data according to claim 1, characterized in that, S104 includes: S10401. Set the sliding window parameters based on the flight parameter sampling frequency; S10402, Calculation of dynamic correlation coefficient: Starting from the beginning of the single-segment dataset, the sliding window sequentially covers consecutive flight parameter points; Calculate the Pearson correlation coefficient for several flight parameter data within each window; Record the intermediate timestamp and corresponding correlation coefficient for each window to form a dynamic correlation sequence; S10403. Calculate the mean value of the correlation coefficient of the sliding window within the flat segment and the changing segment, determine the correlation of flight parameters based on the mean value, mark the corresponding abnormal parameter fluctuations, and output the results.

7. The intelligent analysis and fault location system based on aircraft flight data according to claim 1, characterized in that, S208 includes: Starting with the abnormal parameter as the starting node, traverse the associated fault nodes and calculate the abnormal contribution of each fault node. For each faulty node, further traverse the associated component nodes and calculate the failure probability of each component; A fault propagation path scoring model is constructed, which integrates the anomaly contribution, component failure probability, and parameter fluctuation feature matching degree to generate a comprehensive score for each fault-component combination. The fault propagation path scoring model is a weighted comprehensive scoring model, which integrates the anomaly contribution, component failure probability, and parameter fluctuation feature matching degree, and generates a comprehensive score for each fault-component combination through a linear weighting algorithm.

8. The intelligent analysis and fault location system based on aircraft flight data according to claim 1, characterized in that, The report generation module generates an analysis report, including: S301. Receive the request and perform requirement analysis; S302. Based on the parsing requirements, perform multi-source data aggregation and format standardization; S303. Based on the parsing requirements, perform dynamic assembly and filling of content based on templates; S304. Render the text format and optimize the style of the content template to generate a temporary text file; S305. Preview and output temporary text files for report storage; S306. If an exception occurs during the report generation process, the exception handling process will be automatically triggered.

Images