An aircraft electrical performance evaluation method based on trend characteristics and feature weighting

Abstract

The application provides an aircraft electrical performance evaluation method based on trend characteristics and feature weighting. The method comprises the following steps: step 1, normalization and variable grouping of electrical data; step 2, multi-feature algorithm score design; and step 3, multi-numbered aircraft score weighted fusion. The method adopts corresponding algorithms to independently score different feature types, and then realizes multi-level weighting according to variable clustering grouping and number layer weight, so as to obtain an interpretable, stable and adaptive multi-working-condition-environment aircraft electrical system comprehensive performance score.

Description

Technical Field

[0001] This invention relates to the field of aircraft electrical performance evaluation technology, and in particular to an aircraft electrical performance evaluation method based on trend characteristics and feature weighting. Background Technology

[0002] As a crucial component of modern aerospace systems, aircraft electrical systems undertake core tasks such as energy distribution, signal transmission, and critical control. Their performance and stability directly impact flight safety and mission reliability. With the increasing complexity of aircraft platforms and the diversification of missions, electrical systems exhibit significant nonlinear dynamic characteristics at different stages. Changes in system state are often accompanied by fluctuations and coupling of multi-source data such as voltage, current, power, and temperature. Traditional threshold detection or single-index methods are no longer sufficient to comprehensively reflect the system's operating status.

[0003] Currently, the following methods are commonly used for evaluating the electrical performance of aircraft: (1) Threshold-based or empirical statistical methods: By setting upper and lower limit thresholds for various electrical parameters, alarms are triggered or abnormalities are identified when limits are exceeded. This method is simple to calculate and suitable for real-time monitoring, but it relies on historical experience values, has poor adaptability to different operating conditions and models, and cannot capture trend changes and correlations between multiple parameters.

[0004] (2) Statistical or signal analysis methods based on a single feature: For example, indicators such as mean, variance, RMS (root mean square), or spectral analysis can be used to determine the health status of electrical signals. These methods typically only consider amplitude variations or energy distribution characteristics, lacking descriptions of long-term trends, consistency, and multidimensional correlation characteristics, and are prone to misjudgment under dynamic operating conditions.

[0005] (3) Methods based on machine learning or prediction models: Models such as regression, clustering, and neural networks are used to estimate system health. These methods can achieve nonlinear modeling to some extent, but they rely on a large amount of labeled data, have limited generalization ability, and suffer from high model complexity and poor interpretability, making them unsuitable for deployment in systems.

[0006] In summary, existing aircraft electrical performance evaluation technologies have the following main shortcomings: (1) It has a single characteristic and is difficult to reflect multidimensional trend changes: Existing methods often focus on a single statistical feature (such as mean or variance), failing to simultaneously characterize trend direction, fluctuation amplitude, and collaborative changes between parameters, resulting in insufficient dimensions for state assessment.

[0007] (2) Insufficient model generality and adaptability: The electrical characteristics of different models and serial numbers of aircraft vary significantly, and fixed thresholds or uniform models are often unable to adapt to the diverse environments across devices and missions.

[0008] (3) Lack of a trend consistency measurement mechanism: Traditional assessment methods focus more on instantaneous deviations and fail to utilize the trend consistency among multiple numbered devices to reflect the overall health level of the system, which can easily lead to missed detection of slow-changing anomalies.

[0009] Therefore, there is an urgent need for an electrical performance evaluation method that can combine multi-feature algorithms, possess trend consistency analysis capabilities, and perform weighted fusion for different numbering and variable groupings, so as to achieve a high-precision, interpretable, and adaptive comprehensive state assessment of aircraft electrical systems. Summary of the Invention

[0010] The purpose of this invention is to address the problems in existing technologies by proposing a method for evaluating the electrical performance of aircraft based on trend characteristics and feature weighting. This method employs corresponding algorithms to independently score different feature types, and then uses variable clustering and numbered layer weights to achieve multi-level weighting, thereby obtaining an interpretable, stable, and adaptable comprehensive performance score for the aircraft's electrical system under various operating conditions.

[0011] This invention is achieved through the following technical solution: This invention proposes a method for evaluating the electrical performance of aircraft based on trend characteristics and feature weighting, the method comprising: Step 1: Normalization and variable grouping of electrical data; Step 1 specifically includes: Step 1.1: Normalize the overall aircraft test data; Step 1.2: Perform correlation clustering grouping for different variables of different aircraft with different serial numbers; Step 2: Design of multi-feature algorithm scoring; Step 2 specifically includes: after completing data normalization and correlation clustering, designing five independent variable scoring methods for different data characteristics of aircraft electrical parameters, and quantifying each variable from the dimensions of long-term trend deviation, inter-sample correlation, first derivative change, consistency angle, and non-aligned sequence shape similarity; the five variable scoring methods include: trend scoring method, correlation scoring method, trend derivative scoring method, consistency scoring method, and DTW similarity scoring method; Step 3: Weighted fusion of scores for multiple numbered aircraft; Step 3 specifically includes: Step 3.1: After completing the variable correlation clustering and obtaining the group number to which each variable belongs, calculate the weight of each group based on the group size; Step 3.2: Calculate the weighted scores of each group under each numbered device, group by group. Step 3.3: Electrical performance assessment and health level determination.

[0012] Furthermore, the correlation clustering grouping performed in step 1.2 specifically includes: (1) Correlation calculation Calculate the Pearson correlation coefficient for the normalized time series of each pair of variables:

[0013] in, These are the means of the variables; (2) Threshold determination and clustering Set correlation threshold ,when hour, It indicates the correlation strength between variables, assuming that two variables belong to the same group; through connectivity search or hierarchical clustering methods, it automatically divides all variables that meet the conditions into several groups and outputs the group number index of each variable.

[0014] Furthermore, the trend scoring method is as follows: (1) Constructing baseline trends for variables: For each group In all numbers The average trend baseline of the variable is obtained by averaging over time points. (2) Calculate the trend deviation of the number from the baseline: for each number Group The mean square error between the time series and the baseline is calculated. The smaller the mean square error, the closer the trend of the variable is to the normal baseline; the larger the value, the more significant the long-term trend is.

[0015] Furthermore, the relevance scoring method is as follows: (1) Constructing the leave-one-out baseline curve: For each group First, calculate the sum of all numbers at each time point. Then, for a given number... Construct a baseline sequence that includes all numbers except itself; (2) Calculate the correlation coefficient between the number and the baseline: For each number ,Group Calculate the correlation coefficient between its time series and the corresponding leave-one-out baseline. The closer the correlation coefficient is to 1, the more consistent the trend of the number is with the other numbers. A value close to 0 indicates a weak correlation. A value close to -1 indicates a reverse trend.

[0016] Furthermore, the trend derivative scoring method is as follows: (1) Calculate the sequence of first derivatives for each number: For each number ,Group Calculate the first derivative of the difference approximation between adjacent time points; (2) Constructing the derivative baseline: For each group Accumulate the derivatives of all numbers at each time point, and for a given number... Construct derivative baselines for all numbers except itself; (3) Calculate the mean square error of the difference between derivatives: use the mean square error of the difference between the derivative and the baseline as the trend derivative score.

[0017] Furthermore, the consistency scoring method is as follows: (1) Constructing the leave-one-out baseline trend: For each group First, calculate the sum of all numbers at each time point. For a given number... Construct a baseline sequence that includes all numbers except itself; (2) Calculate cosine similarity: Treat the current number and the baseline sequence as two vectors and calculate the cosine of their angle.

[0018] Furthermore, the DTW similarity scoring method is as follows: (1) Selection of reference baseline curve: for each group The average trend curve, a numbered curve considered as a healthy sample, or a template curve obtained through clustering are selected as the reference sequence for DTW. ; (2) Dynamic time warping distance calculation: For each number ,Group Calculate its time series Compared with the reference baseline The dynamic time warping (DTW) distance.

[0019] Furthermore, the weight calculation method used in step 3.1 is as follows: (1) Count the number of variables within a group; (2) Calculate the weights of the unnormalized groups.

[0020] Furthermore, the weighted score calculation method used in step 3.2 is as follows: After obtaining the final group weights and group-level scores for each group Then, the electrical performance of each numbered device is weighted and integrated to obtain a comprehensive score at the number level under the scoring method:

[0021] in: To obtain the comprehensive electrical performance score for a given number under a specific dimension, the input scoring matrix is ​​modified. The above formula can calculate the score for each of the five scoring methods, enabling independent evaluation.

[0022] The present invention also proposes an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method for evaluating the electrical performance of an aircraft based on trend characteristics and feature weighting.

[0023] The beneficial effects of this invention are: (1) In view of the problem that existing methods have only one feature, this invention adopts five algorithms: trend, correlation, derivative, consistency and DTW, to comprehensively evaluate the electrical performance change characteristics from multiple dimensions such as trend direction, fluctuation amplitude and variable synergy. (2) To address the problem of significant differences in models and serial numbers, this invention achieves automatic variable grouping through correlation clustering and adaptively calculates weights based on group size, eliminating the need for manual threshold setting and significantly improving the algorithm's adaptability to different devices and tasks. (3) In view of the lack of consistency measurement in traditional methods, this invention uses Pearson correlation and cosine similarity to measure the trend consistency among multiple numbers, which can effectively detect slow-changing anomalies and group degradation, and improve the sensitivity of system health monitoring. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0025] Figure 1 This is a graph showing the simulation data of the aircraft.

[0026] Figure 2 This is a schematic diagram of the normalization results.

[0027] Figure 3 This is a schematic diagram of the correlation clustering results.

[0028] Figure 4 This is a diagram illustrating the trend scoring results.

[0029] Figure 5 This is a diagram illustrating the correlation score results.

[0030] Figure 6 This is a schematic diagram of the trend derivative scoring results.

[0031] Figure 7 This is a diagram illustrating the consistency scoring results.

[0032] Figure 8 This is a schematic diagram of the DTW similarity score results.

[0033] Figure 9 This is a schematic diagram of the weight calculation results.

[0034] Figure 10 This is a schematic diagram of the electrical performance evaluation results.

[0035] Figure 11 This is a flowchart of a method for evaluating the electrical performance of aircraft based on trend characteristics and feature weighting. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Specifically, in combination Figures 1-11 This invention proposes a method for evaluating the electrical performance of aircraft based on trend characteristics and feature weighting, the method comprising: Step 1: Normalization and variable grouping of electrical data; Step 1 specifically includes: Step 1.1: Normalize the overall aircraft test data; Step 1.2: Perform correlation clustering grouping for different variables of different aircraft with different serial numbers; Step 2: Design of multi-feature algorithm scoring; Step 2 specifically includes: after completing data normalization and correlation clustering, designing five independent variable scoring methods for different data characteristics of aircraft electrical parameters, and quantifying each variable from the dimensions of long-term trend deviation, inter-sample correlation, first derivative change, consistency angle, and non-aligned sequence shape similarity, so as to ensure the effectiveness and robustness of the method under different operating conditions and different waveform morphologies; the five variable scoring methods include: trend scoring method, correlation scoring method, trend derivative scoring method, consistency scoring method, and DTW similarity scoring method; (1) For slow-changing parameters such as voltage, current, and temperature that show a clear upward / downward / stable trend over time and require evaluation of whether the current number deviates from the overall normal trend, a trend scoring method is adopted. (2) To address the question of whether the overall trend of multiple numbered devices is consistent on the same variable dimension, especially suitable for scenarios that require comparing "whether the number is consistent with the overall behavior of the same batch of devices", a correlation scoring method is adopted; (3) For electrical quantities that change drastically in a short period of time and are sensitive to control response and dynamic performance (such as inrush current, switching transients, voltage drop / recovery process), and for scenarios where it is necessary to measure whether the rate of change is abnormal, the trend derivative scoring method is adopted. (4) For signals with similar waveforms but different amplitudes, focus on whether the trend direction is consistent, such as rising / falling, slow climbing, etc. For scenarios that are not sensitive to amplitude, use the consistency scoring method.

[0038] (5) For electrical process signals with time axis stretching, compression or local alignment deviation, such as different numbered devices with slightly different response times under the same control command, or with delay / advance, but with similar overall process shape, the DTW similarity scoring method is adopted.

[0039] Step 3: Weighted fusion of scores for multiple numbered aircraft; Step 3 specifically includes: Step 3.1: After completing the variable correlation clustering and obtaining the group number to which each variable belongs, calculate the weight of each group based on the group size; Step 3.2: Calculate the weighted scores of each group under each numbered device, group by group. Step 3.3: Electrical performance assessment and health level determination.

[0040] Furthermore, the normalization performed in step 1.1 specifically involves: (1) Global feature merging The test data for each variable across all devices and all samples are combined to form a global feature set:

[0041] in Indicates the variable number. This represents the total number of data points for this variable across all samples.

[0042] (2) Calculation of statistics Calculate the global mean for each variable. and standard deviation :

[0043]

[0044] (3) Standardization transformation Perform normalization for each aircraft sample, each test variable, and each test point in the original data structure:

[0045] The transformed data has a distribution center of 0 and a scale of 1. Global mean and variance are used to avoid normalization distortion caused by single-sample bias. This normalization step ensures that features of different numbers and variables are compared within the same dimensional space, providing a unified data foundation for subsequent trend analysis, correlation clustering, and multi-algorithm scoring.

[0046] Furthermore, the correlation clustering grouping performed in step 1.2 specifically includes: (1) Correlation calculation To reduce electrical feature redundancy and achieve grouped weighting, the Pearson correlation coefficient is calculated for the normalized time series of each pair of variables:

[0047] in, These are the means of the variables; (2) Threshold determination and clustering Set correlation threshold ,when hour, This indicates the strength of the correlation between variables, assuming that two variables belong to the same group; through connectivity search or hierarchical clustering methods, all variables that meet the conditions are automatically divided into several groups, and the group number index of each variable is output. .

[0048] Furthermore, the data structure used in step 2 is as follows:

[0049] in For number / sample index, For the grouped variable index, Using time-point indexing, this invention executes five types of scoring algorithms on demand.

[0050] Furthermore, the trend scoring method is as follows: (1) Constructing the baseline trend of variables For each group In all numbers The average over time points is used to obtain the average trend curve (baseline) of the variable:

[0051] in This is the set of numbers for which the variable is valid across all samples.

[0052] (2) Calculate the trend deviation of the number from the baseline. For each number Group Calculate the mean square error between its time series and the baseline:

[0053] The smaller the value, the closer the trend of the number on the variable is to the normal baseline; the larger the value, the more significant the long-term trend deviation.

[0054] Furthermore, the relevance scoring method is as follows: (1) Constructing the baseline curve using the leave-one-out method For each group First, calculate the sum of all the numbers at each time point:

[0055] For a certain number Construct baseline sequences with all numbers except itself:

[0056] (2) Calculate the correlation coefficient between the number and the baseline. For each number ,Group Calculate the correlation coefficient between its time series and the corresponding leave-one-out baseline:

[0057] in, The closer the value is to 1, the more consistent the trend of that number is with the other numbers; closer to 0, the weaker the correlation; and closer to -1, the opposite trend.

[0058] Furthermore, the trend derivative scoring method is as follows: (1) Calculate the first derivative sequence for each number. For each number ,Group Calculate the first derivative of the approximate difference between adjacent time points:

[0059] (2) Constructing the derivative baseline (leave-one method) For each group Sum the derivatives of all the numbers at each time point:

[0060] For a certain number Construct derivative baselines for all numbers except itself:

[0061] (3) Calculate the mean square error of the difference in derivatives The mean square error of the difference between the derivative and the baseline is used as the trend derivative score:

[0062] The smaller the value, the closer the rate of change is to normal dynamics; the larger the value, the more likely there are problems such as abnormal overshoot, hysteresis, or jitter.

[0063] Furthermore, the consistency scoring method is as follows: (1) Constructing the baseline trend of leave-one-out method For each group First, calculate the sum of all the numbers at each time point:

[0064] For a certain number Construct baseline sequences with all numbers except itself:

[0065] (2) Calculate cosine similarity Treating the current number and the baseline sequence as two vectors, calculate the cosine of their included angle:

[0066] The range of this value is: The closer to 1, the more consistent the trend direction. Even if the amplitude is different, it can be regarded as similar in shape. The closer to 0, the less consistent there is. The closer to -1, the opposite the trend direction.

[0067] Furthermore, the DTW similarity scoring method is as follows: (1) Selection of reference baseline curve For each group The average trend curve, a numbered curve considered as a healthy sample, or a template curve obtained through clustering can be used as the reference sequence for DTW. .

[0068] (2) Calculation of dynamic time warping distance For each number ,Group Calculate its time series Compared with the reference baseline Dynamic Time Warping (DTW) distance:

[0069] in To satisfy both time sequence constraints and window constraints, this invention sets alignment window parameters in its implementation to limit the alignment range and reduce computational complexity. The smaller the value, the closer the process morphology of the number is to the reference baseline under the condition of allowable time stretching / compression. If the distance increases significantly, it indicates that the process response pattern of the number on the variable is significantly different from that of the normal sample, and even traditional point-to-point comparison may not have exceeded the limit.

[0070] Furthermore, the weight calculation method used in step 3.1 is as follows: (1) Number of variables within a statistical group For each group Count the number of variables it contains:

[0071] (2) Calculate the unnormalized group weights Unnormalized weights are defined based on group size:

[0072] Where parameters Controlling preferences for large or small groups: This indicates that the weight of groups with more variables should be increased; This indicates that the weighting differences between the large group and the small group have been leveled. This indicates that each group has equal weight.

[0073] Furthermore, the weighted score calculation method used in step 3.2 is as follows: After obtaining the final group weights and group-level scores for each group Then, the electrical performance of each numbered device is weighted and integrated to obtain a comprehensive score at the number level under the scoring method:

[0074] in: To obtain a comprehensive electrical performance score for the assigned number under a specific dimension (such as trend, correlation, consistency, etc.), the input scoring matrix is ​​modified. The above formula can calculate the score for each of the five scoring methods, enabling independent evaluation.

[0075] Furthermore, step 3.3 specifically includes: After performing the above-described in-group aggregation and group weighting steps on each of the five scoring methods, a set of comprehensive scores for each number on different feature dimensions can be obtained, for example: (1) Trend scoring based on long-term trend deviation; (2) Correlation score based on trend consistency; (3) Trend derivative score based on rate of change; (4) Consistency scoring based on trend direction; (5) DTW similarity scoring based on response latency.

[0076] Depending on the project requirements, independent health thresholds can be set for each type of score to achieve multi-dimensional electrical performance evaluation.

[0077] Example This invention proposes a method for evaluating the electrical performance of aircraft based on trend characteristics and feature weighting, the method comprising the following steps: ①Preparation of aircraft simulation data; ② Data normalization and correlation clustering; ③ Application of electrical performance evaluation methods; ④ Weight generation and weighted fusion scoring.

[0078] The test data used in the implementation of this invention is a simulation dataset constructed based on the characteristics of actual aircraft test data. Due to the sensitivity of the aircraft model, real aircraft data was not used directly. The simulation dataset is constructed using power-on data obtained from four different tests on three different aircraft model numbers, as shown in the curves below. Figure 1 As shown, each curve represents all data obtained in one test (including three types of numbers). The required parameter configurations for the algorithm in the example are as follows: The threshold for correlation clustering, N, is 0.05. DTW similarity score bandwidth ratio: 0.10; Weight generation group size exponentiation :1.0; Minimum weight limit for weight generation: 0.0.

[0079] The normalized curve is as follows Figure 2 As shown, the overall data mean is 0 and the standard deviation is 1 at this point. Standardizing the units facilitates subsequent processing. The grouping information for the completed correlation clustering is as follows: Figure 3 As shown, based on the correlation between the data, the four test results were divided into two groups for subsequent evaluation and analysis. Voltage check 2 and current check were in one group, and voltage check 3 and voltage check 4 were in another group.

[0080] The scoring methods using five different analytical dimensions yielded the following results: Figures 4-8The top part displays the method name information. The obtained scoring results are arranged in the order of device number and relevance group. Different methods focus on different data characteristics. For example, the trend scoring method focuses on the trend deviation characteristics of the curve, the relevance scoring focuses on the consistency of the overall trend, the trend derivative scoring focuses on the curve change rate characteristics, the consistency scoring focuses on the curve change direction characteristics, and the DTW similarity scoring focuses on whether the response position of the curve change is consistent.

[0081] After obtaining the scores, weights are calculated based on the number of variables within each group, such as... Figure 9 As shown, based on the weights and scores of each group, the total score for each device is calculated and sorted, as follows. Figure 10 As shown, a smaller score indicates a smaller deviation from the curve and better electrical performance.

[0082] The present invention also proposes an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method for evaluating the electrical performance of an aircraft based on trend characteristics and feature weighting.

[0083] The memory in this application embodiment can be volatile memory or non-volatile memory, or it can include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the methods described in this invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0084] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0085] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0086] It should be noted that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiments can be completed by the integrated logic circuitry in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied as execution by a hardware decoding processor, or as a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0087] The above provides a detailed description of the aircraft electrical performance evaluation method based on trend characteristics and feature weighting proposed in this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for evaluating the electrical performance of aircraft based on trend characteristics and feature weighting, characterized in that, The method includes: Step 1: Normalization and variable grouping of electrical data; Step 1 specifically includes: Step 1.1: Normalize the overall aircraft test data; Step 1.2: Perform correlation clustering grouping for different variables of different aircraft with different serial numbers; Step 2: Design of multi-feature algorithm scoring; Step 2 specifically includes: after completing data normalization and correlation clustering, designing five independent variable scoring methods for different data characteristics of aircraft electrical parameters, and quantifying each variable from the dimensions of long-term trend deviation, inter-sample correlation, first derivative change, consistency angle, and non-aligned sequence shape similarity; the five variable scoring methods include: trend scoring method, correlation scoring method, trend derivative scoring method, consistency scoring method, and DTW similarity scoring method; Step 3: Weighted fusion of scores for multiple numbered aircraft; Step 3 specifically includes: Step 3.1: After completing the variable correlation clustering and obtaining the group number to which each variable belongs, calculate the weight of each group based on the group size; Step 3.2: Calculate the weighted scores of each group under each numbered device, group by group. Step 3.3: Electrical performance assessment and health level determination.

2. The method according to claim 1, characterized in that, The correlation clustering grouping performed in step 1.2 is specifically as follows: (1) Correlation calculation Calculate the Pearson correlation coefficient for the normalized time series of each pair of variables: in, These are the means of the variables; (2) Threshold determination and clustering Set correlation threshold ,when hour, It indicates the correlation strength between variables, assuming that two variables belong to the same group; through connectivity search or hierarchical clustering methods, it automatically divides all variables that meet the conditions into several groups and outputs the group number index of each variable.

3. The method according to claim 1, characterized in that, The trend scoring method is as follows: (1) Constructing baseline trends for variables: For each group In all numbers The average trend baseline of the variable is obtained by averaging over time points. (2) Calculate the trend deviation of the number from the baseline: for each number Group Calculate the mean square error between its time series and the baseline. The smaller the mean square error, the closer the trend of the number on the variable is to the normal baseline. The larger the value, the more significant the deviation in the long-term trend.

4. The method according to claim 1, characterized in that, The specific correlation scoring method is as follows: (1) Constructing the leave-one-out baseline curve: For each group First, calculate the sum of all numbers at each time point. Then, for a given number... Construct a baseline sequence that includes all numbers except itself; (2) Calculate the correlation coefficient between the number and the baseline: For each number ,Group Calculate the correlation coefficient between its time series and the corresponding leave-one-out baseline. The closer the correlation coefficient is to 1, the more consistent the trend of the number is with the other numbers. The closer it is to 0, the weaker the correlation. A value close to -1 indicates a reversal of the trend.

5. The method according to claim 1, characterized in that, The trend derivative scoring method is as follows: (1) Calculate the sequence of first derivatives for each number: For each number ,Group Calculate the first derivative of the difference approximation between adjacent time points; (2) Constructing the derivative baseline: For each group Accumulate the derivatives of all numbers at each time point, and for a given number... Construct derivative baselines for all numbers except itself; (3) Calculate the mean square error of the difference between derivatives: use the mean square error of the difference between the derivative and the baseline as the trend derivative score.

6. The method according to claim 1, characterized in that, The consistency scoring method is as follows: (1) Constructing the leave-one-out baseline trend: For each group First, calculate the sum of all numbers at each time point. For a given number... Construct a baseline sequence that includes all numbers except itself; (2) Calculate cosine similarity: Treat the current number and the baseline sequence as two vectors and calculate the cosine of their angle.

7. The method according to claim 1, characterized in that, The DTW similarity scoring method is as follows: (1) Selection of reference baseline curve: for each group The average trend curve, a numbered curve considered as a healthy sample, or a template curve obtained through clustering are selected as the reference sequence for DTW. ; (2) Dynamic time warping distance calculation: For each number ,Group Calculate its time series Compared with the reference baseline The dynamic time warping (DTW) distance.

8. The method according to claim 1, characterized in that, The weight calculation method used in step 3.1 is as follows: (1) Count the number of variables within a group; (2) Calculate the weights of the unnormalized groups.

9. The method according to claim 1, characterized in that, The weighted score calculation method used in step 3.2 is as follows: After obtaining the final group weights and group-level scores for each group Then, the electrical performance of each numbered device is weighted and integrated to obtain a comprehensive score at the number level under the scoring method: in: To obtain the comprehensive electrical performance score for a given number under a specific dimension, the input scoring matrix is ​​modified. The above formula can calculate the score for each of the five scoring methods, enabling independent evaluation.

10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-9.

Images