A Fault Diagnosis Method and System for Aircraft Electrical Boxes Based on Multidimensional Feature and Rule Matching

Abstract

This invention belongs to the field of aviation equipment testing technology, specifically involving a fault diagnosis method and system for aircraft electrical boxes based on multi-dimensional feature and rule matching. It includes acquiring multi-dimensional operating parameters such as coil voltage, coil current, contact resistance, and pull-in / release time through a global clock synchronization mechanism. Based on massive amounts of normal data, a normal parameter space model is constructed using principal component analysis. Simultaneously, based on fault tree theory, a mapping rule base is established between typical fault modes and abnormal combinations of original parameters. During the diagnosis process, the T² statistic and Q statistic are used to comprehensively capture the collaborative anomaly characteristics between parameters. Even if a single parameter in the original data does not exceed the standard, it can be identified as long as the multi-dimensional parameter linkage deviates from the normal parameter space model, thus overcoming the limitations of traditional single-parameter judgment. This invention fundamentally solves the problem of parameter collaborative anomaly failure in traditional methods, significantly reducing the risk of missed latent faults and providing reliable assurance for flight safety.

Description

Technical Field

[0001] This invention belongs to the field of aviation equipment testing technology, specifically relating to a method and system for diagnosing aircraft electrical boxes based on multi-dimensional feature and rule matching. Background Technology

[0002] Currently, in the field of aviation maintenance, fault diagnosis of traditional electrical boxes generally adopts a threshold-based "single-parameter judgment" method. This method typically measures only a single key parameter of the relay (such as contact resistance, coil voltage, or coil resistance) and determines its condition based on an empirically set fixed threshold. This simplified diagnostic mode exposes two significant technical bottlenecks in practical applications:

[0003] First, the high rate of missed detection of latent faults poses a serious safety hazard. The operating state of a relay is a comprehensive reflection of the dynamic coupling of its multi-dimensional electrical parameters (such as voltage, current, and contact resistance) and mechanical parameters (such as pull-in time and release time). Many potential faults manifest in their early stages as abnormalities in the synergistic relationship between parameters, rather than a single parameter exceeding the standard. For example, "the contact resistance value is still within the acceptable range, but mechanical jamming of the electromagnetic mechanism causes a significant increase in the pull-in time," or "the coil power supply voltage is normal, but the contact return process is abnormal." Such abnormal parameter synergy is a precursor to performance degradation, but traditional single-parameter judgment methods are completely ineffective in this regard. These latent faults that are not detected in time are very likely to develop into sudden failures during flight, posing a serious threat to flight safety.

[0004] Secondly, poor fault location accuracy leads to low maintenance efficiency. Even when existing methods detect anomalies, the diagnostic conclusion often remains at the general level of "a fault exists in the electrical box," failing to pinpoint the specific faulty relay number within the box, let alone identify the specific fault type (such as contact oxidation or short circuits between coil turns). This forces maintenance work into a "blind disassembly and blind inspection" dilemma, requiring maintenance personnel to completely disassemble the electrical box and individually measure and replace dozens of relays. Statistics show that the average time for locating and troubleshooting a single fault exceeds 60 minutes, significantly reducing maintenance efficiency, extending aircraft downtime, and resulting in high labor and spare parts costs. Summary of the Invention

[0005] To address the aforementioned shortcomings of existing technologies, this invention provides a method, system, and storage medium for diagnosing aircraft electrical boxes based on multidimensional feature and rule matching.

[0006] In a first aspect, the present invention provides a method for diagnosing aircraft electrical box faults based on multidimensional feature and rule matching, comprising:

[0007] S1. Through a global clock synchronization mechanism, multi-dimensional operating parameters of one or more relays in the aircraft electrical box are collected synchronously.

[0008] S2. Based on historical normal state data of multidimensional operating parameters, a feature space is constructed through feature extraction algorithm, and this feature space is defined as the normal parameter space model.

[0009] S3. Based on fault tree theory, establish the mapping relationship between typical fault modes and multidimensional parameter anomaly combinations composed of original measurement parameters, and form a "fault-multidimensional feature" mapping rule library.

[0010] S4. The real-time collected synchronous multidimensional operating parameter sequence is transformed by the feature extraction algorithm, projected into the normal parameter space model in step S2, and its statistics are calculated. If the statistics exceed the preset statistical control limit, it is determined that there is an anomaly, and the original measurement parameters corresponding to this moment are recorded.

[0011] S5. Match the original measurement parameters recorded in step S4 at the abnormal moment with the "fault-multidimensional feature" mapping rule base in step S3, and output the fault location information and diagnostic confidence of the aircraft electrical box. The fault location information includes the fault relay number and fault type.

[0012] Further improvements to this technical solution include that the multi-dimensional operating parameters in step S1 include at least coil voltage, coil current, contact resistance, pull-in time, and release time.

[0013] Further improvements to this technical solution include step S2, which includes:

[0014] S21. Collect historical normal data of multidimensional operating parameters of relays in electrical boxes of the same type of aircraft, remove isolated outliers that exceed the 3σ range, and classify and organize them according to multidimensional operating parameters to form a multidimensional operating parameter sample set.

[0015] S22. Standardize the parameters in the multidimensional operating parameter sample set by standard deviation normalization.

[0016] S23. Calculate the covariance matrix between each parameter based on the standardized data, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and sort them in descending order of eigenvalues.

[0017] S24. Calculate the cumulative variance contribution rate of the eigenvalues, select the first m principal components with a cumulative variance contribution rate of not less than 85%, and concatenate the corresponding eigenvectors column by column to form the loading matrix. The subspace spanned by the matrix is ​​the normal parameter space model.

[0018] Further improvements to this technical solution include step S3, which includes:

[0019] S31. Based on fault tree theory, decompose the typical fault modes of relays in the aircraft electrical box layer by layer, including contact oxidation, coil short circuit and abnormal return.

[0020] S32. Collect historical fault data of multi-dimensional operating parameters corresponding to each typical fault mode, and analyze the abnormal performance and synergistic correlation characteristics of coil voltage, coil current, contact resistance, pull-in time and release time under each fault.

[0021] S33. Establish a logical AND mapping relationship between each typical fault mode and the corresponding abnormal combination of multidimensional operating parameters, and associate it with a unique fault type identifier and the information of the relay channel to which it belongs. After classifying and organizing the fault types, a structured "fault-multidimensional feature" mapping rule library is formed.

[0022] Further improvements to this technical solution include the following: the "fault-multidimensional feature" mapping rule base in step S3 includes:

[0023] The mapping rule corresponding to the "contact oxidation" fault mode is: the contact resistance is greater than the preset contact resistance and the pull-in time is greater than the preset pull-in time threshold.

[0024] The mapping rule corresponding to the "coil short circuit" fault mode is: the coil current is greater than the preset coil current and the coil resistance is less than the preset coil resistance;

[0025] The mapping rule for the "bounceback anomaly" fault mode is: the release time is greater than the preset release time and the peak value of the bounceback voltage is greater than the preset peak value of the bounceback voltage.

[0026] Further improvements to this technical solution include step S4, which includes:

[0027] S41. Perform low-pass filtering on the real-time acquired synchronous multidimensional operating parameters, and then standardize the filtered multidimensional operating parameters.

[0028] S42. The standardized real-time multidimensional operating parameters are transformed by the feature extraction algorithm in step S2, projected onto the normal parameter space model, and the T² statistic and Q statistic are calculated.

[0029] S43. Retrieve the statistical control limits preset based on historical normal data in step S2, and compare the real-time calculated statistics with the statistical control limits one by one.

[0030] S44. If any statistic exceeds the corresponding control limit, it is determined that there is an abnormality in the operation of the aircraft electrical box, and the original multidimensional operating parameters and associated collection timestamps at that moment are recorded.

[0031] Further improvements to this technical solution include the following: the formula for calculating the T² statistic is:

[0032] ;

[0033] in, t Let be the score vector of the sample to be detected in the principal component space. ; z This is the standardized vector of the sample to be detected; V The load matrix is ​​formed by the first m eigenvectors; Let be a diagonal matrix formed by the first m eigenvalues. , This represents the first m eigenvalues ​​obtained after eigenvalue decomposition of the covariance matrix; for The inverse matrix;

[0034] The formula for calculating the Q statistic is:

[0035] ;

[0036] in, The reconstructed vector is based on the first m principal components. ; This is the transpose of the load matrix; Let represent the squared 2-norm of a vector.

[0037] Further improvements to this technical solution include step S5, which includes:

[0038] S51. Extract the corresponding original measurement parameters from the abnormal moment data cache recorded in step S4, including coil voltage value, coil current value, contact resistance value, pull-in time value and release time value, and verify the integrity of the parameters.

[0039] S52. Call the "fault-multidimensional feature" mapping rule library constructed in step S3, and filter out all typical fault modes and associated multidimensional parameter anomaly combinations corresponding to the relay number based on the relay number.

[0040] S53. Verify one by one whether the abnormal parameters after classification meet the logical "AND" condition of the abnormal combination of multi-dimensional parameters in each filtering rule, and record all candidate rules that are completely matched.

[0041] S54. Select the rule with the highest preset reliability from the candidate rules and determine the fault type corresponding to the rule;

[0042] S55. Integrate the fault relay number, the determined fault type and the corresponding confidence level, and output the complete fault location information and diagnostic confidence level in a standardized text format.

[0043] Secondly, the present invention provides an aircraft electrical box fault diagnosis system based on multi-dimensional feature and rule matching, comprising:

[0044] The data synchronization acquisition module is used to synchronously acquire multi-dimensional operating parameters of one or more relays in the aircraft electrical box through a global clock synchronization mechanism. The multi-dimensional operating parameters include at least coil voltage, coil current, contact resistance, pull-in time, and release time.

[0045] The feature space modeling module is used to construct a feature space based on historical normal state data of multidimensional operating parameters through feature extraction algorithms, and this feature space is defined as the normal parameter space model.

[0046] The rule base construction module is used to establish a mapping relationship between typical fault modes and multi-dimensional parameter anomaly combinations composed of original measurement parameters based on fault tree theory, forming a "fault-multi-dimensional feature" mapping rule base.

[0047] The fault diagnosis and analysis module is used to transform the real-time collected synchronous multidimensional operating parameter sequence through a feature extraction algorithm, project it into the normal parameter space model, and calculate its statistics. If the statistics exceed the preset statistical control limit, it is determined that there is an anomaly and the original measurement parameters corresponding to this moment are recorded. The recorded original measurement parameters at the abnormal moment are matched with the "fault-multidimensional feature" mapping rule library.

[0048] The fault result output module is used to output the fault location information and diagnostic confidence level of the aircraft electrical box. The fault location information includes the fault relay number and fault type.

[0049] Thirdly, the present invention provides a computer storage medium storing instructions that, when executed on a computer, cause the computer to perform the methods described in the above aspects.

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

[0051] This invention utilizes a global clock synchronization mechanism to accurately acquire multi-dimensional operating parameters such as coil voltage, current, contact resistance, and pull-in / release time, overcoming the limitations of traditional single-parameter judgment. It combines principal component analysis to construct a normal parameter space model and uses T² and Q statistics to comprehensively capture the coordinated anomalies between parameters. Even if a single parameter does not exceed the standard, deviations in the multi-dimensional parameter linkage relationship from the normal model can be identified, fundamentally solving the problem of parameter coordination anomaly failure in traditional methods, significantly reducing the risk of missed latent faults, and ensuring flight safety.

[0052] This invention constructs a "fault-multidimensional feature" mapping rule base based on fault tree theory. It establishes a precise association between typical fault modes and abnormal parameter combinations through logical "AND" relationships. Combined with a process of filtering and matching by relay number, it can directly output the faulty relay number, specific fault type (e.g., contact oxidation, coil short circuit), and diagnostic confidence level. This overcomes the limitations of traditional "blind disassembly and inspection," reducing the time for single fault location and troubleshooting from over 60 minutes to minutes, significantly reducing aircraft downtime, lowering labor and spare parts costs, and improving the accuracy and efficiency of aircraft maintenance.

[0053] This invention integrates the advantages of data-driven and knowledge-driven approaches. Feature space modeling ensures the sensitivity of anomaly detection, while rule-based matching relies on fault tree theory and historical fault data, providing diagnostic conclusions with clear physical mechanism support. Simultaneously, standardized data processing procedures, low-pass filtering for noise reduction, and parameter integrity verification ensure data quality and diagnostic stability under complex flight environments. Quantitative confidence metrics facilitate rapid decision-making by maintenance personnel, perfectly meeting the stringent requirements of the aviation industry for high reliability and interpretability in diagnostic technology. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic flowchart illustrating a method according to an embodiment of the present invention.

[0056] Figure 2 A schematic diagram of the electrical box circuit for heating the angle of attack and sideslip angle of a certain type of aircraft.

[0057] Figure 3 This is a schematic diagram showing the changes in relay coil voltage and contact resistance data collected at 1-second intervals in a certain type of electrical box.

[0058] Figure 4 This is a logical diagram of the "fault-multidimensional feature" mapping rule base corresponding to a certain type of electrical box constructed in this invention.

[0059] Figure 5 This is a schematic block diagram of a system according to an embodiment of the present invention. Detailed Implementation

[0060] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the specific embodiments. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0062] Figure 1 This is a schematic flowchart illustrating an aircraft electrical box fault diagnosis method based on multi-dimensional feature and rule matching provided by the present invention. Figure 1 The implementing entity can be an aircraft electrical box fault diagnosis system based on multi-dimensional feature and rule matching. Depending on different requirements, the order of the steps in this flowchart can be changed, and some can be omitted.

[0063] like Figure 1 As shown, the method includes:

[0064] S1. Through a global clock synchronization mechanism, multi-dimensional operating parameters of one or more relays in the aircraft electrical box are collected synchronously.

[0065] S2. Based on historical normal state data of multidimensional operating parameters, a feature space is constructed through feature extraction algorithm, and this feature space is defined as the normal parameter space model.

[0066] S3. Based on fault tree theory, establish the mapping relationship between typical fault modes and multidimensional parameter anomaly combinations composed of original measurement parameters, and form a "fault-multidimensional feature" mapping rule library.

[0067] S4. The real-time collected synchronous multidimensional operating parameter sequence is transformed by the feature extraction algorithm, projected into the normal parameter space model in step S2, and its statistics are calculated. If the statistics exceed the preset statistical control limit, it is determined that there is an anomaly, and the original measurement parameters corresponding to this moment are recorded.

[0068] S5. Match the original measurement parameters recorded in step S4 at the abnormal moment with the "fault-multidimensional feature" mapping rule base in step S3, and output the fault location information and diagnostic confidence of the aircraft electrical box. The fault location information includes the fault relay number and fault type.

[0069] To facilitate understanding of the present invention, the following description further illustrates the aircraft electrical box fault diagnosis method based on multidimensional feature and rule matching, using the principle of the present invention and the process of fault diagnosis of aircraft electrical boxes based on multidimensional feature and rule matching in the embodiments.

[0070] Specifically, the multi-dimensional operating parameters in step S1 include at least coil voltage, coil current, contact resistance, pull-in time, and release time.

[0071] Secondly, step S2 includes:

[0072] S21. Collect historical normal data of multidimensional operating parameters of relays in electrical boxes of the same type of aircraft, remove isolated outliers that exceed the 3σ range, and classify and organize them according to multidimensional operating parameters to form a multidimensional operating parameter sample set.

[0073] S22. Standardize the parameters in the multidimensional operating parameter sample set by standard deviation normalization.

[0074] S23. Calculate the covariance matrix between each parameter based on the standardized data, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and sort them in descending order of eigenvalues.

[0075] S24. Calculate the cumulative variance contribution rate of the eigenvalues, select the first m principal components with a cumulative variance contribution rate of not less than 85%, and concatenate the corresponding eigenvectors column by column to form the loading matrix. The subspace spanned by the matrix is ​​the normal parameter space model.

[0076] Next, step S3 includes:

[0077] S31. Based on fault tree theory, decompose the typical fault modes of relays in the aircraft electrical box layer by layer, including contact oxidation, coil short circuit and abnormal return.

[0078] S32. Collect historical fault data of multi-dimensional operating parameters corresponding to each typical fault mode, and analyze the abnormal performance and synergistic correlation characteristics of coil voltage, coil current, contact resistance, pull-in time and release time under each fault.

[0079] S33. Establish a logical AND mapping relationship between each typical fault mode and the corresponding abnormal combination of multidimensional operating parameters, and associate it with a unique fault type identifier and the information of the relay channel to which it belongs. After classifying and organizing the fault types, a structured "fault-multidimensional feature" mapping rule library is formed.

[0080] The "fault-multidimensional feature" mapping rule base in step S3 includes:

[0081] The mapping rule corresponding to the "contact oxidation" fault mode is: the contact resistance is greater than the preset contact resistance and the pull-in time is greater than the preset pull-in time threshold.

[0082] The mapping rule corresponding to the "coil short circuit" fault mode is: the coil current is greater than the preset coil current and the coil resistance is less than the preset coil resistance;

[0083] The mapping rule for the "bounceback anomaly" fault mode is: the release time is greater than the preset release time and the peak value of the bounceback voltage is greater than the preset peak value of the bounceback voltage.

[0084] Specifically, the mapping rule for the "contact oxidation" fault mode is: contact resistance > 50mΩ and pull-in time > 100ms;

[0085] The mapping rule corresponding to the "coil short circuit" fault mode is: coil current > 5A and coil resistance < 1Ω;

[0086] The mapping rule for the "bounceback anomaly" fault mode is: release time > 60ms and bounce voltage peak > 15V.

[0087] Additionally, step S4 includes:

[0088] S41. Perform low-pass filtering on the real-time acquired synchronous multidimensional operating parameters, and then standardize the filtered multidimensional operating parameters.

[0089] S42. The standardized real-time multidimensional operating parameters are transformed by the feature extraction algorithm in step S2, projected onto the normal parameter space model, and the T² statistic and Q statistic are calculated.

[0090] S43. Retrieve the statistical control limits preset based on historical normal data in step S2, and compare the real-time calculated statistics with the statistical control limits one by one.

[0091] S44. If any statistic exceeds the corresponding control limit, it is determined that there is an abnormality in the operation of the aircraft electrical box, and the original multidimensional operating parameters and associated collection timestamps at that moment are recorded.

[0092] Furthermore, the formula for calculating the T² statistic is:

[0093] ;

[0094] in, t Let be the score vector of the sample to be detected in the principal component space. ; z This is the standardized vector of the sample to be detected; V The load matrix is ​​formed by the first m eigenvectors; Let be a diagonal matrix formed by the first m eigenvalues. , This represents the first m eigenvalues ​​obtained after eigenvalue decomposition of the covariance matrix. The largest eigenvalue indicates that the variance of the data is the largest in the direction of the first principal component. The larger the eigenvalue, the more data variation information is contained in that principal component direction. for The inverse matrix;

[0095] The formula for calculating the Q statistic is:

[0096] ;

[0097] in, The reconstructed vector is based on the first m principal components. ; This is the transpose of the load matrix; Let represent the squared 2-norm of a vector.

[0098] Finally, step S5 includes:

[0099] S51. Extract the corresponding original measurement parameters from the abnormal moment data cache recorded in step S4, including coil voltage value, coil current value, contact resistance value, pull-in time value and release time value, and verify the integrity of the parameters.

[0100] S52. Call the "fault-multidimensional feature" mapping rule library constructed in step S3, and filter out all typical fault modes and associated multidimensional parameter anomaly combinations corresponding to the relay number based on the relay number.

[0101] S53. Verify one by one whether the abnormal parameters after classification meet the logical "AND" condition of the abnormal combination of multi-dimensional parameters in each filtering rule, and record all candidate rules that are completely matched.

[0102] S54. Select the rule with the highest preset reliability from the candidate rules and determine the fault type corresponding to the rule;

[0103] S55. Integrate the fault relay number, the determined fault type and the corresponding confidence level, and output the complete fault location information and diagnostic confidence level in a standardized text format.

[0104] This invention constructs a normal parameter space model based on massive amounts of normal data and incorporates principal component analysis. Simultaneously, based on fault tree theory, it establishes a mapping rule base between typical fault modes and abnormal combinations of original parameters. During the diagnostic process, the T² and Q statistics are used to comprehensively capture the collaborative anomaly characteristics between parameters. Even if a single parameter in the original data does not exceed the standard, it can be identified as long as the multi-dimensional parameter linkage deviates from the normal parameter space model, thus overcoming the limitations of traditional single-parameter judgment. This invention fundamentally solves the problem of parameter collaborative anomaly failure in traditional methods, significantly reduces the risk of missing latent faults, and provides reliable assurance for flight safety.

[0105] Preferably, in step S1, the global clock synchronization mechanism ensures that the timestamp error of all multidimensional operating parameters is no greater than 1 microsecond (μs).

[0106] Preferably, in step S2, the feature extraction algorithm uses principal component analysis (PCA), and its specific steps are as follows:

[0107] 1) Standardize the historical normal dataset to eliminate the interference of differences in the units or magnitudes of different monitoring parameters on the accuracy of subsequent feature extraction. Let the original data matrix be... ,in N This represents the number of samples during normal historical operation. p This represents the dimensions of monitoring parameters used to characterize the operating status of the equipment; the standardized matrix is... Standardization can be achieved through existing standardization methods (such as Z-score standardization) to ensure that the mean of each parameter approaches 0 and the standard deviation approaches 1 after standardization.

[0108] 2) Based on the above standardized data matrix Z Calculate the covariance matrix Σ among the monitoring parameters to characterize the degree of linear correlation among different parameters after standardization. Its mathematical expression is:

[0109] ;

[0110] in, Representing a standardized data matrix Z The transpose of the matrix; These are the unbiased estimates of the sample covariance.

[0111] 3) Perform eigenvalue decomposition on the covariance matrix Σ to extract the main variation information in the data. The eigenvalue decomposition satisfies the following relationship:

[0112] ;

[0113] in, ( k =1,2,……,p Let be the eigenvalues ​​of the covariance matrix Σ, and let all eigenvalues ​​be arranged in descending order (i.e., ...). ≥ ≥……≥ ); Eigenvalues The corresponding unit eigenvector is used to characterize the first... k The direction of each principal component.

[0114] The present invention preferably selects the previous m Principal components ( m < p Construct a spatial model of the normal parameters of the equipment (i.e., the feature space). m The value of is determined by the "cumulative variance contribution rate": before definition m The cumulative variance contribution rate is the proportion of the cumulative sum of the eigenvalues ​​to the total sum of all eigenvalues. This cumulative variance contribution rate is required to be no less than 85% to ensure that the selected principal components can retain the vast majority of information related to the normal state of the equipment in the original data.

[0115] Select the first m The eigenvectors corresponding to each eigenvalue are concatenated column-wise to form the loading matrix. Then the former m The subspace spanned by the principal components is the "normal parameter space model" used for subsequent fault diagnosis.

[0116] 4) To accurately define the boundaries of the above-mentioned normal parameter space model and distinguish between the "normal operating state" and the "abnormal fault state" of the equipment, this invention further preferably introduces two core statistics—Hotelling's T² statistic and the squared prediction error Q statistic, specifically defined as follows:

[0117] The statistic measures the degree of deviation of the sample to be tested from the center of the spatial model within the principal component space (i.e., the normal parameter space). Its mathematical expression is:

[0118] ;

[0119] ;

[0120] ;

[0121] in, Let be the score vector of the sample to be tested in the principal component space after standardization. Let be the standardized vector of the sample to be detected. For the front m A diagonal matrix composed of eigenvalues. for The inverse matrix.

[0122] The Q statistic is used to measure the number of samples that have not been previously detected. m The magnitude of the residuals explained by each principal component, i.e., the "deviation residuals" of the sample from the normal parameter space, is mathematically expressed as:

[0123] ;

[0124] in, Standardized vector of the sample to be detected Based on the previous m The reconstructed vectors of the principal components, Let represent the squared 2-norm of a vector.

[0125] Based on the aforementioned historical normal dataset, calculate the corresponding values ​​for all historical normal samples. The values ​​of the statistical test and the Q-statistic are determined through statistical analysis (such as existing statistical methods like chi-square distribution fitting and normal distribution fitting). Control limits of statistics Control limits for Q statistic .

[0126] This embodiment uses the heating electrical box for the angle of attack and sideslip angle of a certain type of aircraft as the application object, such as... Figure 2 As shown, the electrical box has two built-in core relays (denoted as K1: angle of attack heating relay and K2: side-slip angle heating relay). The specific implementation process of the fault diagnosis method and system of the present invention is described in detail for common faults (contact oxidation, coil short circuit, abnormal contact return).

[0127] Specifically, this embodiment describes an aircraft electrical box fault diagnosis method based on multi-dimensional parameter space modeling and rule base matching. It includes the following steps:

[0128] S1 collects five core operating parameters for relays K1 to K2, including coil voltage, coil current, contact resistance, pull-in time, and release time. Coil voltage U The measurement involves connecting a voltage sensor probe in parallel across the relay coil to acquire the real-time voltage across the coil; the coil current... I The measurement involves connecting a current sensor in series with the coil power supply circuit to collect the coil circuit current; the contact resistance (typically in the mΩ range) (e.g.) Figure 3 The measurement (as shown) is susceptible to systematic errors caused by lead resistance and contact resistance. Its theoretical basis is the four-wire method (Kelvin measurement), and its core is "separating the current loop and the voltage measurement loop". The current loop applies a constant current to the resistor (contact) being measured through a constant current source, while the voltage measurement loop directly measures the voltage across the contact through a high-impedance voltmeter. Uc Calculated using Ohm's law R c Relay pull-in time T p Release time T r The timing parameters need to be extracted from the synchronously acquired coil voltage and contact voltage waveforms using waveform feature point recognition theory. The core is to "establish a mapping relationship between signal transitions and physical actions." Based on multi-channel timing consistency theory, the coil voltage waveform (excitation signal) and the contact voltage waveform (response signal) are aligned using timestamps to determine the timing relationship of "excitation-response." The "rising edge of the coil voltage" is used as the trigger signal, and the "falling edge of the contact voltage" is used as the stop signal; the time difference between the two is the trigger signal. T p The trigger signal is the falling edge of the coil voltage, and the stop signal is the rising edge of the contact voltage; the time difference between the two is... T r .

[0129] All sensor data is connected to the 8-channel 18-bit synchronous sampling channel of the MCC USB-1808X data acquisition card. A 10MHz global clock provided by a temperature-controlled crystal oscillator is connected to the "External Clock Input" interface of the MCC USB-1808X data acquisition card via a BNC cable. This ensures that all acquisition channels (voltage, current, resistance, time) are based on the clock of the temperature-controlled crystal oscillator, ensuring that multiple parameters are aligned in the time dimension. Sampling time deviation is verified through oscilloscope testing: the coil voltage and contact voltage signals of the K1 relay are selected, and the timestamps of both signals are simultaneously acquired. This is tested continuously 1000 times, and the maximum timestamp error is statistically analyzed to verify whether the requirement of "synchronization error ≤ 1μs" is met.

[0130] Continuously collect parameter data from more than 10 electrical boxes of the same model with angle of attack and sideslip angle heating under normal operating conditions. Collect 100 sets of samples per electrical box per day (each set of samples contains 1-second time-series data of 5 parameters from K1 to K2, for a total of 1000 data points / parameters). Collect data continuously for more than 10 days to obtain more than 1000 sets of historical normal samples, which are stored in the database. The data format is shown in Table 1 (taking a K1 relay of an electrical box with angle of attack and sideslip angle heating as an example):

[0131] Table 1: Stored data of K1 relay

[0132]

[0133] S2, taking the historical normal data of the heating electrical box K1 relay at a certain angle of attack and sideslip angle as an example (N=1000 samples, p=5 parameters: U , I , Rc , T p , T r ), construct the original data matrix Z-score normalization method is used to... X After processing, a standardized matrix is ​​obtained. The standardized formula is:

[0134] ;

[0135] in, X ij For the first i The first group of samples j Parameter values, μ j For the first j The mean of each parameter, σ j For the first j The standard deviation of each parameter.

[0136] With K1 R c Taking the parameters as an example, the calculation is as follows: μ Rc =32.5mΩ, σ Rc =1.8mΩ, a certain sample X i,Rc =35.1mΩ, then its standardized value Z i,Rc =(35.1-32.5) / 1.8≈1.44.

[0137] Based on the normalized matrix Z Calculate the covariance matrix according to the formula. :

[0138] ;

[0139] Substituting N=1000, the specific value of Σ is calculated as follows (rounded to 3 decimal places):

[0140] ;

[0141] This matrix reflects the degree of linear correlation among the standardized parameters (e.g., T p and T r The covariance is 0.456, indicating a moderate positive correlation between the two.

[0142] Perform eigenvalue decomposition on Σ and solve for the eigenvalues. Five eigenvalues ​​and their corresponding eigenvectors were obtained, and they are arranged in descending order of eigenvalues ​​as shown in Table 2.

[0143] Table 2: Eigenvalues ​​of the covariance matrix

[0144]

[0145] This invention requires a cumulative variance contribution rate ≥ 85%. As shown in the table above, the cumulative variance contribution rate of the first three principal components is 92.14%, which meets the requirement. Therefore, m=3 is selected, and the first three eigenvectors are concatenated column-wise to form the loading matrix. :

[0146] ;

[0147] The subspace formed by the first three principal components is the "normal parameter space model" of K1.

[0148] The T² statistic follows a Hotelling T² distribution, and its control limit formula is:

[0149] ;

[0150] in, α =0.05 (significance level) F α ( m , Nm ) is the upper part of the F distribution α Quantiles. Substituting m=3 and N=1000, we can find the F-distribution table. F 0.05 (3,997)≈2.61, calculated as follows:

[0151] ;

[0152] The Q statistic follows a normal distribution. The Q value is calculated based on 1000 normal samples. ), to obtain the mean of the Q values. μ Q =0.12, σ Q =0.03. Taking the 3σ criterion (covering 99.73% of the normal sample), then:

[0153] ;

[0154] Will =7.84、 =0.21 is stored in the configuration file of the feature space modeling module for subsequent anomaly detection.

[0155] Preferably, in step S3, the mapping relationship in the rule base is a logical "AND" relationship (e.g., ...). Figure 4 (As shown). For example:

[0156] The fault mode "contact oxidation" is mapped to the rule: "contact resistance > R_contact_threshold" and "pick-up time > T_pick-up_threshold";

[0157] The fault mode "coil short circuit" is mapped to the rule: "coil current > I_threshold" and "coil resistance". <R_threshold”;

[0158] The thresholds R_contact_threshold, T_pick-up_threshold, and I_threshold are determined through analysis of historical fault data. The electrical box relay referenced in this invention has an R_contact_threshold of 50mΩ, a T_pick-up_threshold of 100ms, an I_threshold of 5A, and an R_threshold of 1Ω.

[0159] S3, based on fault tree theory (FTA), analyzes three typical fault modes of the angle-of-attack and sideslip angle heating electrical box relays (taking contact oxidation, coil short circuit, and abnormal contact return faults as examples). Combining the parameter characteristics of 100 historical fault samples (50 groups of contact oxidation, 25 groups of coil short circuit, and 25 groups of abnormal contact return), the mapping relationship between "fault mode - multi-dimensional parameter anomaly combination" is determined and stored in the fault rule table of the database in the form of "IF-THEN" rules. The specific rules are shown in Table 3.

[0160] Table 3: Fault Mode Mapping Rule Table

[0161]

[0162] Among them, the peak value of the rebound voltage U bounce by T r Using the corresponding "moment when the coil voltage begins to drop" as a reference, a ±2ms waveform segment is taken, and the extreme values ​​of the coil voltage are extracted after low-pass filtering.

[0163] Preferably, in step S4, if the sample to be tested... Value exceeds or Q value exceeds If so, the device corresponding to the sample is determined to be in an abnormal operating state.

[0164] S4, during the normal operation of the angle-of-attack and sideslip angle heating electrical boxes, initiates the real-time diagnostic process as follows:

[0165] 1) The data synchronization acquisition module collects the parameters of relay K1 in real time to obtain real-time data at a certain moment: U =27.9V I =3.2A、 R c =58mΩ T p =105ms T r =44.2ms;

[0166] 2) Use the Z-score standardization formula from step S2 to calculate the standardized vector. z =[0.056, 0.526, 1.403, 2.211, 0.292];

[0167] 3) Calculate the score vector: t = zV =[0.997,0.533,0.099];

[0168] calculate Statistic: ,in, =diag( λ 1, λ 2, λ 3) = diag(2.812, 1.536, 0.628), =diag(0.355,0.651,1.592), therefore T 2 =0.554;

[0169] Calculate the Q statistic: =[0.256,0.564,-0.582,0.694,0.890], therefore, =2.57;

[0170] 4) Real-time calculation T 2 =0.554< =7.84, Q=2.57> =0.21, satisfying the condition that "Q value exceeds..." If the abnormal condition is detected, it is determined that relay K1 is malfunctioning, and the original parameters at this time are recorded. U =27.9V I =3.2A、 R c =58mΩ T p =105ms T r =44.2ms.

[0171] S5, the fault diagnosis and analysis module extracts the stored abnormal parameters of the K1 relay ( U =27.9V I =3.2A、 R c =58mΩ T p =105ms T r =44.2ms), and match it with the rules in the "Fault Mode Mapping Rule Table":

[0172] 1) Abnormal parameters: R c =58mΩ>50mΩ, T p =105ms>100ms, satisfying the "IF ( )" condition for contact oxidation. R c >50mΩ) AND ( T p >100ms) condition;

[0173] 2) Other rules: I =3.2A<5A (short circuit of coil not satisfied). T r =44.2ms < 60ms (the contact bounce error is not met), therefore the unique matching rule is contact oxidation.

[0174] Diagnostic result calculation: The diagnostic confidence level for contact oxidation is 0.95. Therefore, the fault information is determined as follows: the fault relay number is K1 (angle of attack heating relay), the fault type is contact oxidation, and the diagnostic confidence level is 0.95.

[0175] The results are displayed visually, with K1 relay highlighted in the "Topology Diagram" of the graphical interface, and a diagnostic list pops up (including fault number, type, confidence level, and abnormal parameter values).

[0176] This embodiment fully demonstrates the application process of the present invention in aircraft electrical box fault diagnosis through specific parameter calculations and step derivations. Its technical solution can be directly transferred to electrical box diagnosis scenarios for other aircraft models; adaptation can be achieved simply by adjusting the threshold and model parameters according to the relay parameters of the target electrical box. If it is necessary to expand the fault types (e.g., coil open circuit, contact adhesion), this can be achieved by supplementing fault samples and adding rule base entries, demonstrating good scalability and engineering practicality.

[0177] In some embodiments, the aircraft electrical box fault diagnosis system 200 based on multidimensional feature and rule matching may include multiple functional modules composed of computer program segments. The computer programs for each program segment in the aircraft electrical box fault diagnosis system 200 based on multidimensional feature and rule matching may be stored in the memory of a computer device and executed by at least one processor to perform (see details). Figure 1 (Description) Functionality of aircraft electrical box fault diagnosis based on multidimensional feature and rule matching.

[0178] In this embodiment, the aircraft electrical box fault diagnosis system 200 based on multi-dimensional feature and rule matching can be divided into multiple functional modules according to its functions, such as... Figure 5 As shown. The functional modules may include: a data synchronization acquisition module 210, a feature space modeling module 220, a rule base construction module 230, a fault diagnosis and analysis module 240, and a fault result output module 250. The module referred to in this invention is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and are stored in memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0179] The data synchronization acquisition module 210 is used to synchronously acquire multi-dimensional operating parameters of one or more relays in the aircraft electrical box through a global clock synchronization mechanism. The multi-dimensional operating parameters include at least coil voltage, coil current, contact resistance, pull-in time, and release time.

[0180] The feature space modeling module 220 is used to construct a feature space based on historical normal state data of multidimensional operating parameters through feature extraction algorithms, and the feature space is defined as a normal parameter space model.

[0181] The rule base construction module 230 is used to establish a mapping relationship between typical fault modes and multidimensional parameter anomaly combinations composed of original measurement parameters based on fault tree theory, forming a "fault-multidimensional feature" mapping rule base.

[0182] The fault diagnosis and analysis module 240 is used to transform the real-time acquired synchronous multidimensional operating parameter sequence through a feature extraction algorithm, project it into the normal parameter space model, and calculate its statistics. If the statistics exceed the preset statistical control limit, it is determined that there is an anomaly and the original measurement parameters corresponding to this moment are recorded. The recorded original measurement parameters at the abnormal moment are matched with the "fault-multidimensional feature" mapping rule library.

[0183] The fault result output module 250 is used to output the fault location information and diagnostic confidence of the aircraft electrical box. The fault location information includes the fault relay number and fault type.

[0184] Preferably, the data synchronization acquisition module is responsible for high-precision synchronous acquisition of the multi-dimensional operating parameters of each relay in the aircraft electrical box.

[0185] Preferably, the data synchronization acquisition module includes a multi-channel synchronization acquisition card (e.g., an MCC USB-1808X acquisition card with 8 18-bit synchronous sampling channels) and / or a sensor network with a built-in high-precision clock. This module uses a hardware-level global clock signal to uniformly schedule the sampling timing of all acquisition channels, ensuring that the acquisition timestamps of heterogeneous parameters such as coil voltage, coil current, contact resistance, pull-in time, and release time remain synchronized. The synchronization error is preferably no greater than 1 microsecond (μs), fundamentally solving the core problem of inconsistent time bases for multi-source signals and providing a high-quality data foundation for subsequent time-series correlation analysis.

[0186] Preferably, the feature space modeling module is the core of intelligent analysis and modeling in the aircraft electrical box fault diagnosis system.

[0187] Preferably, the feature space modeling module integrates multivariate statistical analysis algorithms such as Principal Component Analysis (PCA). Its workflow is as follows: It receives historical normal data from the data synchronization acquisition module, first performs standardization preprocessing to eliminate the influence of dimensions; then, preferably, it uses the PCA algorithm to perform eigenvalue decomposition on the preprocessed data covariance matrix, extracting the eigenvalues... m Principal components with a cumulative variance contribution rate of no less than 85% are used to construct a low-dimensional "normal parameter space" model (i.e., feature space). Finally, this module calculates and stores statistical control limits for anomaly detection based on the normal distribution assumption or empirical distribution, including those reflecting the degree of variability within the principal space. Statistical control limits and the control limits of the Q statistic reflecting the model residuals .

[0188] Preferably, the rule base construction module is responsible for storing and managing diagnostic knowledge based on physical mechanisms.

[0189] Preferably, the rule base construction module allows users to associate typical fault modes (e.g., contact oxidation, coil short circuit, bounce anomaly) with abnormal combinations of conditions composed of original measurement parameters, based on fault tree analysis (FTA) theory. The mapping relationships are preferably stored and managed using logical AND relationships (e.g., production rules in the form of "IF-THEN" stored in a relational database or knowledge graph), such as: "IF (contact resistance > 50mΩ) AND (pull-in time > 100ms)THEN (fault type = contact oxidation, confidence weight = 0.95)". This module enables the digitization and solidification of experts' domain knowledge.

[0190] Preferably, the fault diagnosis and analysis module is the online diagnosis and decision-making center of the aircraft electrical box fault diagnosis system.

[0191] Preferably, the working logic of the fault diagnosis and analysis module has a clear temporal sequence: First, it calls the feature space modeling module to project the real-time acquired synchronous data stream onto the feature space and calculates its value in real time. The module compares the Q-statistic with preset control limits to achieve preliminary, highly sensitive anomaly detection. Subsequently, once an anomaly is detected, the module immediately retrieves the original measurement parameter sequence that triggered the anomaly from the cache and calls rules from the rule base construction module for matching, achieving fault location and identification based on physical meaning. This two-level serial "detection-location" architecture balances the sensitivity of feature space methods with the interpretability of rule-based methods.

[0192] Preferably, the fault result output module is used to integrate and display the final diagnostic results of the aircraft electrical box fault diagnosis system. This module supports multiple output formats, including but not limited to: intuitively displaying the fault relay number, fault type, and diagnostic confidence level in the form of highlights, lists, or topology diagrams on a graphical user interface; generating structured diagnostic report files for archiving or uploading to the maintenance management system; and pushing key diagnostic information to the airborne maintenance system or ground station through standard interfaces (such as RS232 serial communication protocol, Ethernet) to provide immediate and clear basis for maintenance decisions.

[0193] Compared with the prior art, the present invention has the following advantages:

[0194] 1) A qualitative leap in latent fault detection capabilities, significantly enhancing safety assurance: Traditional single-parameter threshold methods are completely ineffective against latent faults where "parameter values ​​are normal but the coordination relationship is abnormal." This invention upgrades the diagnostic criteria from isolated parameter values ​​to the evaluation of dynamic coupling relationships between parameters by constructing a high-precision, time-aligned multi-dimensional feature space. Even if all individual parameter values ​​are qualified, if their inherent linkage relationship deviates from the "normal parameter space" model, the system can sensitively detect it, reducing the false negative rate and fundamentally eliminating the sudden risks caused by latent faults.

[0195] 2) Significantly improved fault location accuracy, greatly optimizing maintenance efficiency: Traditional methods can only locate faults down to the "electrical box level," leading to time-consuming and labor-intensive "blind disassembly and inspection." This invention, through a two-level diagnostic architecture of "feature space anomaly perception → original parameter rule matching," improves fault location accuracy to specific relay numbers and fault types. Maintenance personnel can directly replace faulty components based on the diagnostic conclusions, reducing the average troubleshooting time from over 60 minutes to minutes, significantly reducing aircraft downtime and maintenance costs.

[0196] 3) Diagnostic reliability balances sensitivity and interpretability: This invention innovatively integrates a data-driven model (feature space) and a knowledge-driven model (rule base). The feature space provides correlation analysis sensitivity that surpasses that of the human brain, while the rule base based on logical "AND" relationships fully replicates expert diagnostic thinking. The diagnostic chain is clear and transparent, the conclusions are supported by explicit physical mechanisms, the confidence level is calculable, and the decision-making process is traceable, perfectly meeting the stringent reliability requirements of the aviation industry.

[0197] 4) Low data dependency and high engineering applicability: Unlike deep learning methods that require massive amounts of fault data, the core model of this invention can be constructed using only system health status data. The rule base relies on mature fault tree theory and limited engineering experience, and has no strict requirements for rare fault samples. This characteristic enables this invention to be quickly deployed in real-world engineering scenarios, possessing strong practicality and promotional value.

[0198] The present invention also provides a computer storage medium, wherein the computer storage medium may store a program, which, when executed, may include some or all of the steps provided in the embodiments of the present invention. The storage medium may be a magnetic disk, an optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0199] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, or other media capable of storing program code. It includes several instructions to cause a computer terminal (which may be a personal computer, server, or a second terminal, network terminal, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0200] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the present invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the present invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should also be covered within the protection scope of the present invention.

Claims

1. A method for diagnosing aircraft electrical box faults based on multidimensional feature and rule matching, characterized in that, include: S1. Through a global clock synchronization mechanism, multi-dimensional operating parameters of one or more relays in the aircraft electrical box are collected synchronously. S2. Based on historical normal state data of multidimensional operating parameters, a feature space is constructed through feature extraction algorithm, and this feature space is defined as the normal parameter space model. S3. Based on fault tree theory, establish the mapping relationship between typical fault modes and multidimensional parameter anomaly combinations composed of original measurement parameters, and form a "fault-multidimensional feature" mapping rule library. S4. After transforming the real-time collected synchronous multidimensional operating parameter sequence through a feature extraction algorithm, project it into the normal parameter space model of step S2 and calculate its statistics. If the statistic exceeds the preset statistical control limit, an anomaly is determined, and the original measurement parameters at this moment are recorded. S5. Match the original measurement parameters recorded in step S4 at the abnormal moment with the "fault-multidimensional feature" mapping rule base in step S3, and output the fault location information and diagnostic confidence of the aircraft electrical box. The fault location information includes the fault relay number and fault type. The multi-dimensional operating parameters in step S1 include at least coil voltage, coil current, contact resistance, pull-in time, and release time.

2. The aircraft electrical box fault diagnosis method based on multi-dimensional feature and rule matching according to claim 1, characterized in that, Step S2 includes: S21. Collect historical normal data of multidimensional operating parameters of relays in electrical boxes of the same type of aircraft, remove isolated outliers that exceed the 3σ range, and classify and organize them according to multidimensional operating parameters to form a multidimensional operating parameter sample set. S22. Standardize the parameters in the multidimensional operating parameter sample set by standard deviation normalization. S23. Calculate the covariance matrix between each parameter based on the standardized data, and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and sort them in descending order of eigenvalues. S24. Calculate the cumulative variance contribution rate of the eigenvalues, select the first m principal components with a cumulative variance contribution rate of not less than 85%, and concatenate the corresponding eigenvectors column by column to form the loading matrix. The subspace spanned by the matrix is ​​the normal parameter space model.

3. The aircraft electrical box fault diagnosis method based on multi-dimensional feature and rule matching according to claim 1, characterized in that, Step S3 includes: S31. Based on fault tree theory, decompose the typical fault modes of relays in the aircraft electrical box layer by layer, including contact oxidation, coil short circuit and abnormal return. S32. Collect historical fault data of multi-dimensional operating parameters corresponding to each typical fault mode, and analyze the abnormal performance and synergistic correlation characteristics of coil voltage, coil current, contact resistance, pull-in time and release time under each fault. S33. Establish a logical "AND" mapping relationship between each typical fault mode and the corresponding abnormal combination of multidimensional operating parameters, and associate it with a unique fault type identifier and the information of the relay channel to which it belongs. After classifying and organizing the fault types, a structured "fault-multidimensional feature" mapping rule library is formed.

4. The aircraft electrical box fault diagnosis method based on multi-dimensional feature and rule matching according to claim 3, characterized in that, The "fault-multidimensional feature" mapping rule base in step S3 includes: The mapping rule corresponding to the "contact oxidation" fault mode is: the contact resistance is greater than the preset contact resistance and the pull-in time is greater than the preset pull-in time threshold. The mapping rule for the "coil short circuit" fault mode is: the coil current is greater than the preset coil current and the coil resistance is less than the preset coil resistance. The mapping rule for the "bounceback anomaly" fault mode is: the release time is greater than the preset release time and the peak value of the bounceback voltage is greater than the preset peak value of the bounceback voltage.

5. The aircraft electrical box fault diagnosis method based on multi-dimensional feature and rule matching according to claim 2, characterized in that, Step S4 includes: S41. Perform low-pass filtering on the real-time acquired synchronous multidimensional operating parameters, and then standardize the filtered multidimensional operating parameters. S42. The standardized real-time multidimensional operating parameters are transformed by the feature extraction algorithm in step S2, projected onto the normal parameter space model, and the T² statistic and Q statistic are calculated. S43. Retrieve the statistical control limits preset based on historical normal data in step S2, and compare the real-time calculated statistics with the statistical control limits one by one. S44. If any statistic exceeds the corresponding control limit, it is determined that there is an abnormality in the operation of the aircraft electrical box, and the original multidimensional operating parameters and associated collection timestamps at that moment are recorded.

6. The aircraft electrical box fault diagnosis method based on multi-dimensional feature and rule matching according to claim 5, characterized in that, The formula for calculating the T² statistic is: ； in, t Let be the score vector of the sample to be detected in the principal component space. ; z This is the standardized vector of the sample to be detected; V The load matrix is ​​formed by the first m eigenvectors; Let be a diagonal matrix formed by the first m eigenvalues. , This represents the first m eigenvalues ​​obtained after eigenvalue decomposition of the covariance matrix; for The inverse matrix; The formula for calculating the Q statistic is: ； in, The reconstructed vector is based on the first m principal components. ; This is the transpose of the load matrix; Let represent the squared 2-norm of a vector.

7. The aircraft electrical box fault diagnosis method based on multi-dimensional feature and rule matching according to claim 1, characterized in that, Step S5 includes: S51. Extract the corresponding original measurement parameters from the abnormal moment data cache recorded in step S4, including coil voltage value, coil current value, contact resistance value, pull-in time value and release time value, and verify the integrity of the parameters. S52. Call the "fault-multidimensional feature" mapping rule library constructed in step S3, and filter out all typical fault modes and associated multidimensional parameter anomaly combinations corresponding to the relay number based on the relay number. S53. Verify one by one whether the abnormal parameters after classification meet the logical "AND" condition of the abnormal combination of multi-dimensional parameters in each filtering rule, and record all candidate rules that are completely matched. S54. Select the rule with the highest preset reliability from the candidate rules and determine the fault type corresponding to the rule; S55. Integrate the fault relay number, the determined fault type and the corresponding confidence level, and output the complete fault location information and diagnostic confidence level in a standardized text format.

8. A fault diagnosis system for aircraft electrical boxes based on multi-dimensional feature and rule matching, characterized in that, include: The data synchronization acquisition module is used to synchronously acquire multi-dimensional operating parameters of one or more relays in the aircraft electrical box through a global clock synchronization mechanism. The multi-dimensional operating parameters include at least coil voltage, coil current, contact resistance, pull-in time, and release time. The feature space modeling module is used to construct a feature space based on historical normal state data of multidimensional operating parameters through feature extraction algorithms, and this feature space is defined as the normal parameter space model. The rule base construction module is used to establish a mapping relationship between typical fault modes and multi-dimensional parameter anomaly combinations composed of original measurement parameters based on fault tree theory, forming a "fault-multi-dimensional feature" mapping rule base; The fault diagnosis and analysis module is used to transform the real-time collected synchronous multidimensional operating parameter sequence through a feature extraction algorithm, project it into the normal parameter space model, and calculate its statistics. If the statistics exceed the preset statistical control limit, it is determined that there is an anomaly and the original measurement parameters corresponding to this moment are recorded. The recorded original measurement parameters at the time of the anomaly are matched with the "fault-multidimensional feature" mapping rule library. The fault result output module is used to output the fault location information and diagnostic confidence level of the aircraft electrical box. The fault location information includes the fault relay number and fault type.

9. A computer-readable storage medium storing a computer program, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images