Spatial-temporal feature mining and risk assessment method, system and device for flight operation data, and storage medium

Abstract

The invention relates to the technical field of operation state monitoring and risk assessment, provides a spatial-temporal feature mining and risk assessment method, system and device for flight operation data, and a storage medium, and solves the problem of low early recognition accuracy of a composite fault of a mobile device. The method comprises the following steps: acquiring flight operation data and optical signal data of the mobile equipment, wherein the flight operation data comprises component state data and environmental parameter data; according to the optical signal data, a structural deformation field of the mobile equipment is combined to form a strain thermodynamic diagram of the structural surface; performing space-time alignment fusion on the strain thermodynamic diagram, the component state data and the environmental parameter data, and extracting space-time features from a fusion result through an attention enhancement mechanism; and performing kernel fuzzy C-means clustering operation on the spatial-temporal characteristics to identify a potential fault mode combination, and determining a flight risk assessment result according to the potential fault mode combination. According to the invention, the early recognition accuracy of the composite fault of the mobile equipment is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mobile device operation state monitoring and risk assessment, and particularly relates to a flight operation data spatiotemporal feature mining and risk assessment method, system, device and storage medium. BACKGROUND

[0002] Under complex working conditions, the operation safety of a mobile device depends on real-time monitoring and risk assessment of multi-source data, and it is necessary to accurately capture the coupling effect of structural deformation, component state and environmental factors to identify potential composite failure modes and assess risk levels.

[0003] At present, a monitoring method based on multi-sensor data fusion adopts a fixed-weight data superposition method, combines a traditional clustering algorithm to classify operation states, and determines a risk level through a preset threshold.

[0004] This scheme relies on static weight allocation and is difficult to adapt to dynamically changing working condition characteristics, resulting in insufficient sensitivity to composite failures. At the same time, the processing capability of traditional clustering methods for nonlinear separable data is limited, and the accuracy of risk assessment is limited by the rationality of the preset threshold. SUMMARY

[0005] The present application provides a flight operation data spatiotemporal feature mining and risk assessment method, system, device and storage medium to solve the problem of low early identification accuracy of mobile device composite failures in the prior art.

[0006] In a first aspect, the present application provides a flight operation data spatiotemporal feature mining and risk assessment method, comprising: obtaining flight operation data and optical signal data of a mobile device, the flight operation data comprising component state data and environmental parameter data; forming a strain thermodynamic map of a structural surface according to the optical signal data in combination with a structural deformation field of the mobile device; spatiotemporally aligning and fusing the strain thermodynamic map, the component state data and the environmental parameter data, and extracting spatiotemporal features from the fusion result through an enhanced attention mechanism; performing a kernel fuzzy C-means clustering operation on the spatiotemporal features to identify a potential failure mode combination, and determining a flight risk assessment result according to the potential failure mode combination.

[0007] Optionally, the kernel fuzzy C-means clustering operation on the spatiotemporal features to identify a potential failure mode combination, and determining a flight risk assessment result according to the potential failure mode combination, comprises: mapping the spatio-temporal features to a kernel space to generate a feature vector set, the kernel space being an implicit feature space generated by a kernel function transformation in a kernel fuzzy C-means clustering; initializing cluster center points of the feature vector set, and forming a cluster group set based on the cluster center points and the feature vector set; extracting core feature combinations of each cluster group in the cluster group set as potential failure mode combinations; mapping the potential failure mode combinations to a flight risk assessment result based on a preset risk level reference table.

[0008] Optionally, the forming of the cluster group set based on the cluster center points and the feature vector set comprises: adopting a genetic algorithm to iteratively optimize positions of the cluster center points; generating a correlation strength matrix based on the optimized positions through a preset membership function; determining final cluster attribution results of each feature vector in the feature vector set according to maximum value indexes of row vectors of the correlation strength matrix; grouping all feature vectors according to the final cluster attribution results to form cluster groups corresponding to each cluster center point, and integrating all the cluster groups to form the cluster group set.

[0009] Optionally, the generating of the correlation strength matrix based on the optimized positions through the preset membership function comprises: calculating similarity measure values of each feature vector in the feature vector set to each cluster center point in the kernel space based on the optimized positions; composing similarity vectors of each feature vector from all the similarity measure values corresponding to the feature vector; inputting the similarity vectors into the membership function to output correlation strength values of the feature vector to each cluster center point; aggregating the correlation strength values of all feature vectors to construct a correlation strength matrix between each feature vector and each cluster center point.

[0010] Optionally, the spatio-temporal alignment and fusion of the strain thermodynamic map, the component state data and the environmental parameter data, and the extraction of spatio-temporal features from the fusion result through an enhanced attention mechanism comprise: independently time-slicing the strain thermodynamic map, the component state data and the environmental parameter data according to a unified timestamp to obtain three types of time slice groups at the same time; unifying mapping of the three types of time slice groups to a preset three-dimensional space grid coordinate system according to a structure space model of the mobile device; In the three-dimensional space grid coordinate system, the three types of time slice groups in the same space grid cell are dynamically weighted and fused to generate a fused grid tensor; A spatio-temporal attention weighting operation is performed on the fused grid tensor to generate a spatio-temporal feature.

[0011] Optionally, the spatio-temporal attention weighting operation performed on the fused grid tensor to generate a spatio-temporal feature comprises: The fused grid tensor is divided into multiple time windows along the time dimension, and a time attention weight is generated based on the change trend of the grid data in each time window; In the spatial dimension, a spatial attention weight is generated according to the data change amplitude between the grid cells in the fused grid tensor; The fused grid tensor is decomposed into multiple feature channels according to the data source type, and a channel attention weight is generated based on the covariance between the feature channels; The time attention weight, the spatial attention weight, and the channel attention weight are subjected to tensor product operation to construct a three-dimensional attention weight tensor; The three-dimensional attention weight tensor and the fused grid tensor are multiplied element by element to output a spatio-temporal feature.

[0012] Optionally, the strain thermal map of the structure surface is formed according to the optical signal data in combination with the structural deformation field of the mobile device, comprising: The phase offset distribution is parsed from the optical signal data; According to the preset optical fiber sensing network topology relationship in the structural deformation field of the mobile device, the phase offset is bound to the corresponding spatial position point; Based on each spatial position point and the corresponding phase offset, a local deformation variable is calculated through a preset phase and deformation conversion model; The difference information between the local deformation variables of adjacent spatial position points is analyzed to determine the strain intensity distribution of the structure surface; The strain intensity distribution is subjected to data visualization conversion according to a preset color gradient rule to generate a strain thermal map of the structure surface.

[0013] In a second aspect, the present application provides a system for spatio-temporal feature mining and risk assessment of flight operation data, comprising: An acquisition module is configured to acquire flight operation data and optical signal data of a mobile device, wherein the flight operation data comprises component state data and environmental parameter data; A formation module is configured to form a strain thermal map of a structure surface according to the optical signal data in combination with a structural deformation field of the mobile device; an extraction module configured to perform spatio-temporal alignment and fusion on the strain thermodynamic diagram, the component state data, and the environmental parameter data, extract spatio-temporal features from the fusion result by enhancing an attention mechanism, and perform kernel fuzzy C-means clustering on the spatio-temporal features to identify a potential fault mode combination, and determine a flight risk assessment result according to the potential fault mode combination. an identification module configured to perform kernel fuzzy C-means clustering on the spatio-temporal features to identify a potential fault mode combination, and determine a flight risk assessment result according to the potential fault mode combination.

[0014] In a third aspect, the present application provides a computing device including a processor and a memory, wherein the memory stores a computer program, and the processor is configured to execute the computer program to implement the flight operation data spatio-temporal feature mining and risk assessment method according to any one of the first aspect.

[0015] In a fourth aspect, the present application provides a computer storage medium storing computer program instructions, wherein the computer program instructions are executed by a processor to implement the flight operation data spatio-temporal feature mining and risk assessment method according to any one of the first aspect.

[0016] In the present application, a flight operation data spatio-temporal feature mining and risk assessment method is provided, which includes: obtaining flight operation data and light signal data of a mobile device, wherein the flight operation data includes component state data and environmental parameter data; forming a strain thermodynamic diagram of a structure surface according to the light signal data in combination with a structural deformation field of the mobile device; performing spatio-temporal alignment and fusion on the strain thermodynamic diagram, the component state data, and the environmental parameter data, extracting spatio-temporal features from the fusion result by enhancing an attention mechanism; performing kernel fuzzy C-means clustering on the spatio-temporal features to identify a potential fault mode combination, and determining a flight risk assessment result according to the potential fault mode combination.

[0017] The technical scheme provided by the present application has the following beneficial effects: The present application realizes the synchronous collection of multi-source heterogeneous data, and provides a complete data basis for comprehensive risk assessment. The light signal is converted into a high-precision deformation distribution, and the local stress concentration area of the structure is intuitively presented. Through dynamic weighted fusion and an attention mechanism, the key features of the environmental-structure-device coupling are highlighted. The potential feature combination of the composite fault is identified, and the precise matching of the risk level is realized.

[0018] Further, the present application maps the spatio-temporal features to a high-dimensional kernel space to generate a feature vector set, forms a grouping set by optimizing the clustering center points, extracts the core feature combination of each group as a fault mode, and finally outputs a risk assessment result based on a preset rule.

[0019] And, the separation ability of the nonlinear feature is improved by the kernel space transformation, and the boundary uncertainty of the fault mode is accurately captured by combining the fuzzy clustering, so that the reliable identification and risk quantification of the complex fault under complex working conditions are realized.

[0020] These aspects or other aspects of the present application will be more apparent in the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0022] Figure 1 A flow chart of a flight operation data spatio-temporal feature mining and risk assessment method provided by the embodiments of the present application; Figure 2 A structural schematic diagram of a flight operation data spatio-temporal feature mining and risk assessment system provided by the embodiments of the present application; Figure 3 A structural schematic diagram of a computing device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0023] In order to make the person skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application.

[0024] In some of the descriptions in the specification and claims of the present application and the above-mentioned drawings, a plurality of operations appearing in a specific order are included, but it should be clearly understood that these operations can be executed or in parallel without the order appearing in the text, and the serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes can include more or less operations, and the operations can be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second" and the like in this paper are used to distinguish different messages, devices, modules, etc., and do not represent the order of sequence, nor do "first" and "second" represent different types.

[0025] In the existing mobile device operation risk assessment method, fixed weight fusion of multi-source data is combined with traditional clustering analysis, which has obvious deficiencies: on the one hand, fixed weight is difficult to adapt to the dynamic coupling relationship between structural deformation, component vibration and environmental factors, resulting in insufficient sensitivity of sudden composite failure; on the other hand, the traditional clustering has limited ability to distinguish nonlinear characteristics, and it is easy to forcibly classify features with weak correlation, so that the risk determination result deviates greatly from the actual working condition. These problems are caused by the insufficient adaptability of data processing method to complex operation scene.

[0026] In view of the above limitations, the present application provides a kind of flight operation data spatio-temporal feature mining and risk assessment method, its innovation lies in: by real-time analysis of structural deformation distribution of light signal, combined with the spatio-temporal alignment fusion of component state and environmental parameters, key coupling features are extracted using adaptive weight distribution and attention mechanism;Further, an improved fuzzy clustering algorithm is used to accurately identify fault mode combination in high-dimensional feature space, and finally match the preset rule to output risk level. This method breaks through the limitations of fixed weight and linear analysis, and through the cooperation of dynamic fusion and nonlinear clustering, it not only improves the ability to capture composite failure, but also ensures the accuracy of risk determination, effectively solving the problem of poor adaptability of existing technology to complex working conditions.

[0027] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0028] Figure 1 A flowchart of a flight operation data spatio-temporal feature mining and risk assessment method provided by the embodiments of the present application is shown in Figure 1 As shown in the figure, the method comprises: Step 101: obtaining flight operation data and light signal data of a mobile device, wherein the flight operation data comprises component state data and environmental parameter data.

[0029] In step 101, the flight operation data includes component state data (such as engine vibration amplitude, rudder deflection angle) and environmental parameter data (such as wind speed, temperature, air pressure). The component refers to the key functional components in the mobile device that are directly related to flight safety, including but not limited to engine rotors, wing rudders, landing gear hydraulic devices, and other core mechanical components that generate vibration / displacement signals. Wing deformation refers to the shape change of the wing structure under the action of aerodynamic load during flight, which is an important part of the component state data in the flight operation data. It is monitored and recorded in real time by the optical fiber sensor as deformation data, reflecting the structural safety state of the aircraft. Engine abnormality refers to the state where the engine operating parameters deviate from the normal range, which is the core monitoring content of the component state data in the flight operation data. The abnormal data collected by the vibration sensor can directly trigger risk assessment and warning. Sudden wind refers to the sudden strong airflow change in the atmosphere, which is a key meteorological element in the environmental parameter data. It is collected and recorded in real time by the wind speed sensor, directly affecting the flight safety assessment and fault mode judgment. The optical signal data refers to the laser signal collected by the optical fiber sensor network arranged on the surface of the mobile device structure. The phase, intensity and other optical characteristics of the laser signal will change measurably with the structural deformation, which is used to invert the micro-deformation of the device surface.

[0030] In the embodiments of the present application, first, the optical signal data is collected in real time by the optical fiber sensor network installed at the key parts of the mobile device, and the component state data (such as vibration sensor readings) and environmental parameter data (such as meteorological sensor data) are obtained from the onboard monitoring system. All data are recorded with a unified time stamp to ensure the time consistency of subsequent analysis.

[0031] For example, during the flight of a certain mobile device, the optical fiber sensor monitors the wing deformation in real time and collects the optical phase change data; at the same time, the engine vibration sensor records the vibration amplitude data, and the meteorological sensor records the current wind speed and temperature data. All data are stored synchronously according to the millisecond-level time stamp to form the original data set.

[0032] Step 102: According to the optical signal data, combined with the structural deformation field of the mobile device, a strain thermal map of the structure surface is formed.

[0033] In step 102, the structural deformation field represents the deformation distribution of the mobile device structure under stress, which is determined by the pre-set optical fiber sensor network topology. The strain thermal map represents the strain intensity distribution of each region on the structure surface, which is visually displayed by color gradient.

[0034] In the embodiments of the present application, the phase offset of the optical signal data is analyzed, and the spatial position information of the optical fiber sensor is combined to calculate the local deformation of each sensing point. According to the deformation difference between adjacent sensing points, the strain intensity is calculated, and finally the thermal map covering the entire structure surface is generated, in which the color depth represents the strain size.

[0035] For example, according to the phase shift data collected by the optical fiber sensor, the deformation values of three adjacent sensing points in a certain area of the wing are calculated as 0.12 mm, 0.15 mm and 0.10 mm, respectively. The local strain is calculated by the strain formula ε = ΔL / L (where ΔL is the deformation difference, and L is the spacing between sensing points), and finally a thermal map of the strain concentration at the root of the left wing is generated.

[0036] Step 103: spatiotemporal alignment and fusion of the strain thermal map, the component state data and the environmental parameter data, and extraction of spatiotemporal features from the fusion result by enhancing the attention mechanism.

[0037] In step 103, spatiotemporal alignment and fusion means to uniformly calibrate data from different sources in time and space, ensuring data correlation.

[0038] In the embodiments of the present application, the thermal map, the component state data and the environmental parameter data are uniformly sliced in time and matched in position information in a three-dimensional space grid. The multi-source data in the same grid cell is assigned a dynamic weight (for example, the strain data weight is higher, and the environmental data weight is lower), and a fusion tensor is generated by weighted fusion. Finally, the key spatiotemporal features are strengthened and redundant information is suppressed by using the attention mechanism.

[0039] For example, at 30 seconds, the strain value in a certain grid cell of the left wing is 0.15 rad, the vibration value is 0.17 mm, and the wind speed is 25 m / s. After weighted fusion with weights of 0.6, 0.3 and 0.1, the feature value of the cell is 0.6 x 0.15 + 0.3 x 0.17 + 0.1 x 25 = 3.091. After attention weighting of the grid data of the whole machine, the feature value of the high strain area of the left wing is improved.

[0040] Step 104: performing kernel fuzzy C-means clustering operation on the spatiotemporal features to identify potential fault mode combinations, and determining the flight risk assessment result according to the potential fault mode combinations.

[0041] In step 104, the kernel fuzzy C-means clustering representation maps data to a high-dimensional space for fuzzy grouping, solving the problem of non-linear separability. The potential fault mode combination represents the typical fault feature set extracted after clustering. The flight risk assessment result is the quantitative safety level output after matching the potential fault mode combination identified by kernel fuzzy C-means clustering with the preset risk level table. Its specific meaning includes: 1) risk level division (such as level one to five, the larger the value, the higher the risk), reflecting the matching degree of the current operating state and the typical fault mode; 2) fault type positioning (such as structural deformation overrun and strong turbulence coupling), clearly identifying the risk source; 3) early warning suggestion (such as immediate repair or continuous monitoring), providing a basis for operation and maintenance decision-making. This result is generated by mapping and comparing the clustered feature combination with historical accident data, which not only includes the comprehensive evaluation of the current state, but also reflects the fault development trend prediction.

[0042] In the embodiments of the present application, the spatio-temporal features are mapped to a high-dimensional kernel space, a number of cluster centers are initialized, and the center positions are adjusted through iterative optimization. The membership degrees of each feature vector and the center are calculated to generate a correlation strength matrix, and finally the cluster groups are divided according to the maximum membership degree. The core feature combination of each group is extracted, and the evaluation result is output by matching the preset risk level table.

[0043] For example, a certain clustering identifies "vibration 0.2 mm + strain 0.16 rad + wind speed 28 m / s" as a class of fault mode, and determines it as a level three risk according to the risk table, triggering an early warning.

[0044] The method realizes the accurate identification and risk assessment of complex faults under complex working conditions through dynamic fusion and intelligent clustering of multi-source data, solves the problems of insufficient sensitivity to sudden abnormalities and high misjudgment rate of traditional methods, and improves the reliability of safety monitoring.

[0045] In order to solve the problem of insufficient accuracy of complex fault mode recognition in flight operation data, in some embodiments, step 104: performing a kernel fuzzy C-means clustering operation on the spatio-temporal features to identify a potential fault mode combination, and determining a flight risk assessment result according to the potential fault mode combination, including: Step 201: mapping the spatio-temporal features to a kernel space to generate a feature vector set, the kernel space being an implicit feature space generated by a kernel function transformation in kernel fuzzy C-means clustering.

[0046] In step 201, the kernel space refers to a high-dimensional feature space to which the original feature data is mapped through mathematical transformation, and this transformation can better distinguish complex data patterns. The feature vector set is a high-dimensional data set formed after the kernel function conversion of the spatio-temporal features.

[0047] In the embodiments of the present application, first, a suitable kernel function is selected to perform nonlinear transformation on the space-time features, so as to map the originally difficult-to-linearly separate feature data to a higher-dimensional space, and in the new space, similar features are automatically clustered, thereby creating favorable conditions for subsequent clustering analysis.

[0048] Step 202: initializing cluster center points of the feature vector set, and forming a cluster group set based on the cluster center points and the feature vector set.

[0049] In step 202, the cluster center points refer to the center positions representing various feature groups in the high-dimensional kernel space. The cluster group set is a set of feature vectors grouped according to similarity through an optimization algorithm.

[0050] In the embodiments of the present application, a plurality of initial center positions are randomly set, and then the positions of the center points are continuously adjusted through iterative calculation, so that each feature vector can find the most similar center point, and finally a plurality of feature groups are formed, and the features in each group have high similarity.

[0051] Step 203: extracting core feature combinations of each cluster group in the cluster group set as potential fault mode combinations.

[0052] In step 203, the core feature combination is composed of the coupling states of three types of key features: 1) structural strain features (such as a deformation interval of 0.15-0.18 radians of a wing surface); 2) component state features (such as a vibration amplitude of 0.20-0.22 millimeters of an engine); and 3) environmental features (such as a gust wind speed of 28-30 meters per second). A specific example: when the clustering analysis identifies the feature combination of "wing deformation 0.16 radians + engine vibration 0.21 millimeters + encountering a 29-meter-per-second gust", it constitutes a typical fault mode combination reflecting the coupling of aerodynamic load and mechanical vibration.

[0053] In the embodiments of the present application, statistical analysis is performed on all feature vectors in each cluster group, and the central distribution intervals of the feature values of each group are found out, and these intervals are combined to form representative fault feature modes.

[0054] Step 204: mapping the potential fault mode combinations to flight risk assessment results based on a preset risk level reference table.

[0055] In step 204, the preset risk level reference table is a correspondence table of fault feature combinations and risk levels established in advance.

[0056] In the embodiments of the present application, the identified feature combinations are compared with a pre-established fault case library, and the closest historical fault mode is found, so as to determine the risk level of the current equipment operating state.

[0057] Here's a specific example: Following the scenario mentioned above where the mobile device detected strain concentration at the root of the left wing and abnormal engine vibration during flight, after the system acquired the characteristic data at 30 seconds (left wing strain 0.15 radians, vibration 0.17 mm, wind speed 25 m / s), it first used the Gaussian kernel function The features are mapped to high-dimensional space, where x represents the feature vector, y represents the kernel center point, and σ is set to 0.5 to control the mapping range. The kernel space coordinates of the feature vector at this moment are calculated to be [0.82, 0.15, 0.03]; the coordinates of the three cluster center points are initialized to [0.80, 0.20, 0.00], [0.10, 0.90, 0.00], and [0.30, 0.30, 0.40], and are updated to [0.85, 0.12, 0.03], [0.15, 0.80, 0.05], and [0.25, 0.25, 0.50] after 5 iterative optimizations; the degree of membership of the feature vector to each center point is calculated using the formula , where d is the kernel space distance. The correlation strengths between the vector and the three centers are 0.92, 0.08, and 0.00, respectively, so it is classified into the first cluster group. Finally, the core feature range extracted from this group is a combination of strain 0.14-0.18 radians, vibration 0.16-0.20 mm, and wind speed 24-26 m / s. After matching the risk level table, it is confirmed to be a secondary risk state that requires key monitoring. The system immediately issues a graded warning signal to prompt operators to strengthen monitoring of the area.

[0058] In the embodiment of the present application, through the synergistic effect of kernel space transformation and fuzzy clustering, accurate capture of complex fault characteristics and intelligent judgment of risk levels are achieved, which effectively improves the ability to identify safety hazards in complex operating environments and provides a reliable decision-making basis for equipment maintenance.

[0059] In order to solve the problem of insufficient center point position optimization and feature grouping accuracy in flight data cluster analysis, in some embodiments, step 202: forming a cluster group set based on the cluster center point and the feature vector set includes: Step 301: Adopting a genetic algorithm to iteratively optimize the position of the cluster center point.

[0060] In step 301, the genetic algorithm operation is an optimization method that simulates the natural evolution process and gradually improves the position of the cluster center point through operations such as selection, crossover and mutation.

[0061] In an embodiment of the present application, several groups of center point position schemes are first randomly generated, the clustering effect index corresponding to each group of schemes is calculated, the scheme with the best effect is retained as the parent generation, and new child generation schemes are generated through position crossover and slight perturbations. After multiple iterations, the optimal center point distribution is obtained.

[0062] Step 302: Based on the optimized positions, generate a correlation strength matrix through a preset membership function.

[0063] In step 302, the membership function is used to calculate the correlation degree of the feature vector and each center point. The correlation strength matrix is a two-dimensional table recording the correlation strength of all feature vectors and each center point.

[0064] In the embodiments of the present application, according to the optimized center point positions, the distance of each feature vector to each center point is calculated, the distance is converted into a correlation strength value between 0 and 1 through a preset conversion function, and finally the strength matrix is generated, in which the row represents the feature vector and the column represents the center point.

[0065] Step 303: According to the row vector maximum value index of the correlation strength matrix, determine the final clustering attribution result of each feature vector in the feature vector set.

[0066] In step 303, the row vector maximum value index refers to the column number of the maximum value of each row in the correlation strength matrix. The final clustering attribution result is the center point number to which each feature vector is assigned.

[0067] In the embodiments of the present application, each row of the correlation strength matrix is scanned to find the cell with the maximum value in the row, and the column number corresponding to the cell is taken as the final attribution class of the feature vector in the row.

[0068] Step 304: According to the final clustering attribution result, group all feature vectors according to the corresponding clustering center points to form a clustering group corresponding to each clustering center point, integrate all clustering groups to form a clustering group set.

[0069] In step 304, the clustering group refers to the set of all feature vectors belonging to the same center point.

[0070] In the embodiments of the present application, all feature vectors are classified into corresponding center points according to the attribution result, and the feature vectors under each center point form an independent group, which together constitute a complete clustering group set.

[0071] The following is a specific example: Based on the core space feature vector [0.82, 0.15, 0.03] obtained by the foregoing mobile device at the 30-second moment, the system first optimizes the initial center point position using a genetic algorithm, and the three initial center points [0.80, 0.20, 0.00], [0.10, 0.90, 0.00] and [0.30, 0.30, 0.40] are updated to [0.85, 0.12, 0.03], [0.15, 0.80, 0.05] and [0.25, 0.25, 0.50] after iterative optimization, wherein each iteration optimizes the clustering tightness index J = Σμ²d² corresponding to each center point scheme, wherein μ represents the membership degree and d represents the distance between the feature vector and the center point; then the correlation strength of the feature vector and the optimized center point is calculated, and the formula The correlation values with the three centers are calculated as 0.92, 0.08 and 0.00 respectively, wherein d is calculated by ||x-y||, x represents the feature vector coordinates, and y represents the center point coordinates; according to the maximum value 0.92, the feature vector is divided into the clustering group corresponding to the first center point; finally, the division results of all 200 feature vectors are integrated to form three clustering groups.

[0072] In the embodiments of the present application, through intelligent optimization of the center point position and accurate calculation of the feature correlation degree, efficient grouping of flight data is realized, ensuring the close gathering of similar features and the clear separation of different features, and providing a reliable data basis for subsequent fault mode identification.

[0073] In order to solve the problem of insufficient quantization precision of the correlation between the feature vector and the clustering center in the clustering analysis of flight data, in some embodiments, step 302: based on the optimized position, an association strength matrix is generated by a preset membership function, comprising: Step 401: based on the optimized position, the similarity measure value of each feature vector in the feature vector set to each clustering center point in the core space is calculated respectively.

[0074] In step 401, the similarity measure value refers to the closeness value of the feature vector and the clustering center point in the core space, and the larger the value, the more similar the features.

[0075] In the embodiments of the present application, the optimized clustering center point coordinates and the core space coordinates of the feature vector set are first obtained, then the distance of each feature vector to each center point is calculated, and finally the distance is converted into a similarity value through a preset similarity calculation formula.

[0076] Step 402: all similarity measure values corresponding to each feature vector form a similarity vector.

[0077] In step 402, the similarity vector refers to a set of similarity values of a single feature vector to all cluster center points, and the dimension is equal to the number of cluster center points.

[0078] In the embodiment of the present application, for each feature vector, the similarity values of the feature vector to each center are arranged in a fixed center point order to form an ordered numerical sequence, and the sequence is the similarity vector of the feature vector.

[0079] In step 403, the similarity vector is input into the membership function, and the association strength value of the feature vector to each cluster center point is output.

[0080] In step 403, the association strength value refers to a probability value of the feature vector belonging to a certain cluster center point after the membership function is standardized, and the value ranges from 0 to 1.

[0081] In the embodiment of the present application, the similarity vector is input into a preset conversion function to normalize the original similarity, so that the sum of the association strength of each feature vector to all center points is 1, thereby obtaining the probability distribution of the association strength.

[0082] In step 404, the association strength values of all feature vectors are aggregated to construct an association strength matrix between each feature vector and each cluster center point.

[0083] In the embodiment of the present application, the association strength values of each vector are sequentially filled into the corresponding row of the matrix according to the fixed order of the feature vectors, and finally a complete matrix with the number of rows equal to the number of feature vectors and the number of columns equal to the number of center points is formed.

[0084] The following is a specific example: Based on the core space feature vector [0.82, 0.15, 0.03] obtained by the mobile device at the 30th second and the three optimized center point positions [0.85, 0.12, 0.03], [0.15, 0.80, 0.05] and [0.25, 0.25, 0.50], the Euclidean distance of the feature vector to each center point is first calculated, wherein the distance , , ; then the distance values are converted into similarity measure values by the similarity calculation formula S = 1 / (1+d2), obtaining 0.983, 0.527 and 0.740; then the three similarity values form a similarity vector [0.983, 0.527, 0.740], which is input into the membership function μ = S / ∑S for normalization processing, where ∑S = 0.983 + 0.527 + 0.740 = 2.250, and finally the association strength values of the feature vector with the three center points are 0.437, 0.234 and 0.329 respectively; after the system performs the same calculation process on all 200 feature vectors, a 200-row 3-column association strength matrix is constructed, where each row of data represents the association degree of the corresponding feature vector with the three center points, for example, the first row [0.437, 0.234, 0.329] represents the membership probability distribution of the first feature vector with the three center points, and this matrix provides a quantitative basis for subsequent accurate division of feature vectors into corresponding cluster groups.

[0085] In the embodiments of the present application, by accurately calculating the feature similarity and standardizing conversion, a scientific association strength quantification system is established, ensuring the objectivity and accuracy of feature vector clustering grouping, and providing reliable data support for flight safety risk assessment.

[0086] In order to solve the problem of insufficient accuracy of spatio-temporal feature extraction in multi-source flight data fusion, in some embodiments, step 103: the strain thermodynamic map, the component state data and the environmental parameter data are spatio-temporally aligned and fused, and spatio-temporal features are extracted from the fusion results by enhancing the attention mechanism, including: Step 501: independently time-slice the strain thermodynamic map, the component state data and the environmental parameter data according to a unified timestamp, and obtain three types of time-slice groups at the same time.

[0087] In step 501, the time-slice group refers to a momentary data segment extracted from continuous data stream at the same time point, and the three types of time-slice groups correspond to the momentary states of strain, component and environment data respectively.

[0088] In the embodiments of the present application, a unified time reference axis is first established, and then data segments at the same time are synchronously extracted from the three types of original data according to the set time interval, ensuring that all data are strictly aligned in the time dimension.

[0089] Step 502: according to the structure space model of the mobile device, the three types of time-slice groups are uniformly mapped to a preset three-dimensional space grid coordinate system.

[0090] In step 502, the structural space model of the mobile device is established by a pre-established three-dimensional digital model of the device, which contains the geometric dimensions, material properties and sensor layout position information of all key components, for uniformly mapping data from different sources to the same physical space coordinate system, ensuring that strain, vibration and other data can be accurately corresponded to specific locations of the structure. The three-dimensional space grid coordinate system is a three-dimensional grid system formed by dividing the mobile device structure according to a certain resolution, and each grid cell corresponds to a specific location on the structure.

[0091] In the embodiments of the present application, the grid coordinate system is pre-established according to the structural drawing of the device, and then each type of data is mapped to the corresponding grid cell according to its collection position information, so that the data from different sources are accurately matched in space position.

[0092] Step 503: In the three-dimensional space grid coordinate system, the three types of time slice groups in the same space grid cell are dynamically weighted and fused to generate a fused grid tensor.

[0093] In step 503, the fused grid tensor refers to a three-dimensional data block formed by fusing multi-source data according to weights in a unified space-time grid.

[0094] In the embodiments of the present application, the three types of data in each grid cell are analyzed, the fusion weights are dynamically allocated according to the importance and reliability of the data, and then the weighted data is superimposed and integrated to form a tensor data containing complete feature information.

[0095] Step 504: Perform a spatio-temporal attention weighting operation on the fused grid tensor to generate a spatio-temporal feature.

[0096] In step 504, spatio-temporal attention weighting refers to a process of dynamically adjusting the importance of data according to its variation characteristics in time and space. The spatio-temporal feature is a multi-dimensional vector sequence after attention weighting, and each vector contains a risk feature quantization value (such as a risk coefficient of 0.82 at the left wing leading edge 30 seconds) at a specific time and space position. The fused grid tensor is an unweighted original fused data cube; the former highlights key risk signals, and the latter retains all original information. The difference between the two is that the former is a refined and risk-focused expression of the latter.

[0097] In the embodiments of the present application, the variation law of the fused tensor in each dimension is analyzed, a higher weight is allocated to the fluctuating region, and a lower weight is allocated to the stable region, and finally the spatio-temporal feature set reflecting the key risk characteristics is extracted.

[0098] The following is a specific example: Based on the flight data collected by the aforementioned mobile device at the 30-second time point, the system first synchronously slices the strain thermal map of the left wing root area, engine vibration data, and environmental wind speed data according to a 30.0-30.9 second time window, obtaining 10 groups of time-aligned data slices, wherein the slice at the 30.5 second time point contains a strain value of 0.15 radian, a vibration value of 0.17 mm, and a wind speed of 25 m / s; according to the pre-stored three-dimensional structural model of the wing, these data are mapped to grid element No. G205, which corresponds to a specific position on the leading edge of the left wing; during fusion processing, the system automatically assigns weights according to the importance of the data, with a strain data weight of 0.6, a vibration data weight of 0.3, and a wind speed data weight of 0.1. The initial fusion value of the element is obtained by weighted calculation: 0.6 x 0.15 + 0.3 x 0.17 + 0.1 x 25 = 3.091; subsequent attention weighting analysis finds that the strain fluctuation in this area during the 30.0-30.9 second period reaches 0.12-0.18 radian, with a change amplitude, so an attention weight of 0.9 is given, and the final feature value of the element is 3.091 x 0.9 = 2.782; at the same time, the corresponding area of the right wing only obtains a weight of 0.3 due to the gentle change in data, and the feature value is 1.953; after processing all the grid elements of the entire aircraft, the system forms a complete spatiotemporal feature dataset.

[0099] In the embodiments of the present application, through precise spatiotemporal alignment and intelligent weight allocation, efficient fusion and key feature extraction of multi-source flight data are realized, laying a data foundation for accurate identification of composite failure modes and improving the reliability of flight safety monitoring.

[0100] To solve the problem of inaccurate key feature extraction after multi-source data fusion, in some embodiments, step 504: performing spatiotemporal attention weighting operation on the fusion grid tensor to generate spatiotemporal features, comprising: Step 601: dividing the fusion grid tensor into multiple time windows along the time dimension, and generating time attention weights based on the change trend of grid data in each time window.

[0101] In step 601, the change trend of grid data in the time window refers to the degree of fluctuation of data in each time period, which is obtained by calculating the standard deviation of data of each grid element in the time period. The larger the standard deviation, the more intense the fluctuation. Specifically, the statistical dispersion degree of the data sequence of consecutive time points in the time window is calculated. The time attention weight refers to the weight value allocated according to the degree of change of data over time. The higher the weight of the time period, the higher the change.

[0102] In the embodiments of the present application, the fusion grid tensor is first divided into several time segments according to a fixed time length, then the fluctuation amplitude of data in each segment is calculated, and finally the corresponding time attention weight is allocated according to the fluctuation size.

[0103] Step 602: In the spatial dimension, a spatial attention weight is generated according to the data variation amplitude between grid cells in the fusion grid tensor.

[0104] In step 602, the data variation amplitude between grid cells refers to the difference degree of data in adjacent spatial positions, which is obtained by calculating the gradient value of data in adjacent grid cells. The greater the gradient value, the greater the spatial variation. Specifically, the spatial attention weight is obtained by averaging the absolute value of the difference between the data in the central grid cell and the data in the surrounding cells.

[0105] In the embodiments of the present application, the data variation gradient between adjacent grid cells is analyzed, and higher spatial attention weights are given to regions with sharp changes, and lower weights are given to smooth regions.

[0106] Step 603: The fusion grid tensor is decomposed into multiple feature channels according to the data source type, and a channel attention weight is generated based on the covariance between the feature channels.

[0107] In step 603, the covariance between the feature channels refers to the strength of the linkage between different data sources, which is obtained by calculating the covariance matrix of the time series data in each channel. The greater the absolute value of the covariance, the stronger the correlation. Specifically, the covariance is calculated according to the formula cov(X, Y) = E[(X-μx)(Y-μy)]. The channel attention weight is a weight value allocated according to the strength of the correlation between different data sources. The stronger the correlation, the higher the weight of the channel.

[0108] In the embodiments of the present application, the covariance matrix between each data channel is calculated, and the corresponding channel attention weight is allocated according to the correlation between the channels.

[0109] Step 604: The time attention weight, the spatial attention weight, and the channel attention weight are subjected to tensor product operation to construct a three-dimensional attention weight tensor.

[0110] In step 604, the three-dimensional attention weight tensor is a weight distribution cube formed by the comprehensive operation of the weights in the time, space, and channel dimensions.

[0111] In the embodiments of the present application, the weight matrices in the three dimensions are subjected to tensor product operation to generate a three-dimensional weight tensor consistent with the dimensions of the fusion grid tensor.

[0112] Step 605: The three-dimensional attention weight tensor is multiplied element by element with the fusion grid tensor to output the spatio-temporal features.

[0113] In step 605, element-by-element multiplication refers to the operation of multiplying the values of the weight tensor and the fusion tensor at the corresponding positions two by two.

[0114] In the embodiments of the present application, the three-dimensional attention weight tensor is multiplied by the position corresponding value of the fusion grid tensor, and finally the spatio-temporal feature data of the strengthened key features and the weakened secondary features are obtained.

[0115] The following is a specific example: Based on the fusion grid data generated by the mobile device during 30.0-30.9 seconds, the system first divides this time into three time windows (30.0-30.3 seconds, 30.3-30.6 seconds, 30.6-30.9 seconds), calculates the standard deviation of the strain value of the left wing G205 grid unit in each window as 0.02, 0.03 and 0.01 radian respectively, and allocates the time attention weight 0.4, 0.6, 0.2 according to the standard deviation; in the spatial dimension, the average strain difference of G205 unit and the adjacent three units is 0.04 radian, and the vibration difference is 0.03 mm, and the spatial attention weight 0.7 is obtained by comprehensive consideration; according to the covariance formula , the strain-vibration covariance is 0.85, the strain-wind speed is 0.62, and the vibration-wind speed is 0.45, and the channel weight is 0.5, 0.3 and 0.2 according to the allocation; the weight of three dimensions 0.6 (time), 0.7 (space), 0.5 (channel) is multiplied to obtain the composite weight of G205 unit in 30.3-30.6 seconds window 0.6×0.7×0.5=0.21, and the final feature value 0.65 is obtained by multiplying the weight with the original fusion value 3.091; at the same time, the corresponding area of the right wing has low final feature value 0.19 due to low weight in each dimension (0.3, 0.2, 0.3), and the system differentiates the weight to enhance the high-risk features of the left wing leading edge.

[0116] In the embodiments of the present application, through the three-dimensional attention weighting mechanism, the intelligent focusing of the key features of flight data is realized, the recognition sensitivity of the composite fault features is effectively improved, and more accurate data support is provided for flight safety warning.

[0117] In order to solve the problem of insufficient precision of strain distribution visualization in structure deformation monitoring, in some embodiments, step 102: forming a strain thermodynamic map of the structure surface according to the optical signal data and combining the structure deformation field of the mobile device, comprises: Step 701: analyzing the phase shift distribution from the optical signal data.

[0118] In step 701, the phase shift distribution refers to the distribution of the phase change of the optical signal collected by the optical fiber sensor in space, reflecting the deformation degree of each part of the structure.

[0119] In an embodiment of the present application, the original optical signal is first demodulated and processed, and the phase change data of each sensing point is extracted to form a phase change distribution map covering the entire monitoring area. The specific implementation process is: after the original optical signal is collected by each sensing point in the optical fiber sensing network, the original optical signal is input into the phase demodulation device, and the interference optical signal component carrying deformation information is separated in the phase demodulation device, and the phase of the interference optical signal component is phase resolved, and the real-time phase offset value corresponding to each sensing point is output, and the phase offset value is mapped to the corresponding area on the wing surface according to the spatial layout coordinates of the sensing point. Example: Taking aircraft wing health monitoring as an example, an optical fiber sensor array is arranged on the wing surface in a 20cm×20cm grid. When the aircraft encounters turbulence: each optical fiber sensing point collects the original optical signal in real time and transmits it to the onboard demodulation unit; the demodulation unit performs interference fringe analysis on the optical signal of the third grid point on the left wing and extracts a phase offset of 0.15rad.

[0120] Step 702: Bind the phase offset to the corresponding spatial position point according to the optical fiber sensing network topology relationship preset in the structural deformation field of the mobile device.

[0121] In step 702, the topological relationship of the optical fiber sensor network refers to the spatial description of the layout positions and connection methods of the optical fiber sensors on the surface of the structure.

[0122] In an embodiment of the present application, a correspondence is established between each phase change data point and the specific position coordinates of the structure surface according to a pre-stored optical fiber layout scheme, ensuring that the data accurately matches the spatial position.

[0123] Step 703: Based on each spatial position point and the corresponding phase offset, calculate the local deformation through a preset phase and deformation conversion model.

[0124] In step 703, the phase and deformation conversion model is essentially a physical relationship model established based on the fiber Bragg grating sensing principle. For example, the standard photoelastic effect formula is adopted: deformation amount = phase offset × fiber wavelength / (2π × photoelastic coefficient), where the photoelastic coefficient is determined by the characteristics of the fiber material, realizing the quantitative conversion of the optical signal phase change to the physical deformation amount.

[0125] In the embodiment of the present application, for each located phase data point, the actual deformation amount at that position is calculated using a preset formula based on the physical properties of the material and the optical fiber parameters.

[0126] Step 704: Analyze the difference information between the local deformation amounts of adjacent spatial positions to determine the strain intensity distribution of the structure surface.

[0127] In step 704, adjacent points in the optical fiber sensing network have a physical spacing not exceeding a preset threshold, and usually at least three adjacent points (front / back / left or right) are needed to calculate a reliable strain gradient to ensure spatial continuity. The difference information refers to the change rate of the deformation variable between adjacent sensing points, which is obtained by calculating the ratio of the deformation variable difference to the spacing between the sensing points, and reflects the severity of local stretching / compression of the structure. For example, the difference information of point A deformation 0.1 mm and adjacent point B deformation 0.15 mm is 0.05 mm / 10 cm=500 micro-strain. The strain intensity distribution refers to the quantitative description of the severity of deformation in each region on the surface of the structure.

[0128] In the embodiments of the present application, the deformation variable difference between adjacent measurement points is calculated, combined with the spacing data between the points, to obtain the strain intensity value of each local region, and form a complete strain distribution field.

[0129] Step 705: data visualization conversion of the strain intensity distribution according to a preset color gradient rule to generate a strain thermodynamic map of the surface of the structure.

[0130] In step 705, the color gradient rule refers to a visualization scheme that maps strain values to different colors.

[0131] In the embodiments of the present application, according to a preset color mapping table, the strain intensity value of each point is converted into a corresponding color code, and finally a thermodynamic map is generated to intuitively display the strain distribution.

[0132] The following is a specific example: Based on the aforementioned mobile device left wing monitoring scenario in flight, the optical fiber sensor network collects phase shift data of optical signals of three adjacent sensing points (numbered P1, P2, P3, with a spacing of 0.1 meters) as 0.12 radians, 0.15 radians and 0.10 radians, respectively. According to the preset optical fiber layout topology, the coordinates of the three points corresponding to the left wing leading edge are determined as (X12, Y05), (X13, Y05) and (X14, Y05); the deformation variable is calculated through the phase-deformation conversion model ΔL=λ·Δφ / 2πn, where λ is the wavelength of light 1550 nanometers, n is the refractive index of the optical fiber 1.45, and Δφ is the phase shift, and the deformation variables of the three points are 0.12 mm, 0.15 mm and 0.10 mm, respectively; the strain between P1-P2 is calculated as ε=(0.15-0.12) / 0.1=0.3, and the strain between P2-P3 is calculated as ε=(0.15-0.10) / 0.1=0.5; according to the color mapping rule, the strain value 0.3 corresponds to yellow and the strain value 0.5 corresponds to red, and finally the generated thermodynamic map displays the P2-P3 region as a red warning and the P1-P2 region as a yellow prompt.

[0133] In the embodiment of the present application, the visual conversion of the optical fiber sensing data to the strain distribution is realized, the intuitive presentation of the structural deformation state is realized, and a reliable deformation evaluation basis is provided for flight safety monitoring.

[0134] Figure 2 A structure diagram of a flight operation data spatiotemporal feature mining and risk assessment system provided in the embodiment of the present application is shown in Figure 2 The system comprises: An acquisition module 21 is configured to acquire flight operation data and optical signal data of a mobile device, wherein the flight operation data comprises component state data and environmental parameter data.

[0135] A formation module 22 is configured to form a strain thermodynamic map of a structural surface according to the optical signal data in combination with a structural deformation field of the mobile device.

[0136] An extraction module 23 is configured to perform spatiotemporal alignment and fusion on the strain thermodynamic map, the component state data and the environmental parameter data, and extract spatiotemporal features from the fusion result by enhancing an attention mechanism.

[0137] An identification module 24 is configured to perform kernel fuzzy C-means clustering operation on the spatiotemporal features to identify a potential fault mode combination, and determine a flight risk assessment result according to the potential fault mode combination.

[0138] Figure 2 The flight operation data spatiotemporal feature mining and risk assessment system can perform Figure 1 The implementation principle and technical effects of the flight operation data spatiotemporal feature mining and risk assessment method in the embodiment shown in

[0139] In one possible design, Figure 2 The flight operation data spatiotemporal feature mining and risk assessment system in the embodiment shown in Figure 3 may be implemented as a computing device, which may comprise a storage component 31 and a processing component 32. The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0140] The processing component 32 is configured to perform the flight operation data spatiotemporal feature mining and risk assessment method in the embodiment described above. Figure 1

[0141] ​The processing component 32 can include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component can also be one or more Application-Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Process Device (DSPD), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components, for executing the above method.

[0142] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage devices or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0143] Of course, the computing device can also include other components, such as an input / output interface, a display component, a communication component, etc.

[0144] The input / output interface provides an interface between the processing component and peripheral interface modules, which can be output devices, input devices, etc.

[0145] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0146] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform, and the computing device can refer to a cloud server, and the processing component, the storage component, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0147] The embodiment of the application further provides a computer storage medium, which stores a computer program, and the computer program can realize the above-mentioned method when being executed by a computer. Figure 1 The embodiment of the application further provides a method for spatiotemporal feature mining and risk assessment of flight operation data.

[0148] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-mentioned system, device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0149] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme. Those skilled in the art can understand and implement without creative labor.

[0150] From the foregoing description of the embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and the necessary general hardware platform, and of course, it can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the method described in each embodiment or some part of the embodiment.

[0151] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the application, and not to limit them; although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application.

Claims

1. A method for spatiotemporal feature mining and risk assessment of flight operation data, characterized in that, The method comprises the following steps: acquiring flight operation data and light signal data of a mobile device, the flight operation data comprising component state data and environmental parameter data; forming a strain thermal map of a structural surface according to the light signal data in combination with a structural deformation field of the mobile device; spatiotemporally aligning and fusing the strain thermal map, the component state data, and the environmental parameter data, and extracting spatiotemporal features from the fusion result through an enhanced attention mechanism; performing kernel fuzzy C-means clustering on the spatiotemporal features to identify a potential fault mode combination, and determining a flight risk assessment result according to the potential fault mode combination.

2. The method of claim 1, wherein, The method of performing kernel fuzzy C-means clustering on the spatiotemporal features to identify a potential fault mode combination, and determining a flight risk assessment result according to the potential fault mode combination comprises the following steps: mapping the spatiotemporal features to a kernel space to generate a feature vector set, the kernel space being an implicit feature space generated by a kernel function transformation in kernel fuzzy C-means clustering; initializing cluster center points of the feature vector set, and forming a cluster group set based on the cluster center points and the feature vector set; extracting core feature combinations of each cluster group in the cluster group set as potential fault mode combinations; mapping the potential fault mode combinations to a flight risk assessment result based on a preset risk level reference table.

3. The method of claim 2, wherein, The method of forming a cluster group set based on the cluster center points and the feature vector set comprises the following steps: iteratively optimizing the positions of the cluster center points using a genetic algorithm operation; generating a correlation strength matrix based on the optimized positions through a preset membership function; determining the final clustering attribution results of each feature vector in the feature vector set according to the maximum value indexes of the row vectors of the correlation strength matrix; grouping all feature vectors according to the cluster center points to which they belong based on the final clustering attribution results, forming cluster groups corresponding to each cluster center point, and integrating all cluster groups to form a cluster group set.

4. The method of claim 3, wherein, The method of generating a correlation strength matrix based on the optimized positions through a preset membership function comprises the following steps: calculating the similarity measurement values of each feature vector in the feature vector set to each cluster center point in the kernel space based on the optimized positions; composing a similarity vector from all similarity measurement values corresponding to each feature vector; inputting the similarity vector into the membership function to output the correlation strength values of the feature vector to each cluster center point; aggregating the correlation strength values of all feature vectors to construct a correlation strength matrix between each feature vector and each cluster center point.

5. The method of claim 1, wherein, The method of spatiotemporally aligning and fusing the strain thermal map, the component state data, and the environmental parameter data, and extracting spatiotemporal features from the fusion result through an enhanced attention mechanism comprises the following steps: independently time-slicing the strain thermal map, the component state data, and the environmental parameter data according to a unified timestamp to obtain three types of time slice groups at the same time; unifying and mapping the three types of time slice groups to a preset three-dimensional space grid coordinate system according to a structural space model of the mobile device; In the three-dimensional space grid coordinate system, the three types of time slice groups in the same space grid cell are dynamically weighted and fused to generate a fused grid tensor; Performing a spatio-temporal attention weighting operation on the fused grid tensor to generate a spatio-temporal feature.

6. The method of claim 5, wherein, The spatio-temporal attention weighting operation on the fused grid tensor to generate a spatio-temporal feature includes: Dividing the fused grid tensor into multiple time windows along the time dimension, and generating a time attention weight based on the change trend of the grid data in each time window; In the spatial dimension, a spatial attention weight is generated according to the data change amplitude between the grid cells in the fused grid tensor; The fused grid tensor is decomposed into multiple feature channels according to the data source type, and a channel attention weight is generated based on the covariance between the feature channels; Performing a tensor product operation on the time attention weight, the spatial attention weight, and the channel attention weight to construct a three-dimensional attention weight tensor; Element-wise multiplying the three-dimensional attention weight tensor and the fused grid tensor to output a spatio-temporal feature.

7. The method of claim 1, wherein, The strain thermal map of the structure surface is formed according to the optical signal data in combination with the structural deformation field of the mobile device, including: Parsing the phase offset distribution from the optical signal data; Binding the phase offset to the corresponding spatial position point according to the preset fiber sensing network topology relationship in the structural deformation field of the mobile device; Calculating the local deformation variable through the preset phase and deformation conversion model based on each spatial position point and the corresponding phase offset; Determining the strain intensity distribution of the structure surface by analyzing the difference information between the local deformation variables of adjacent spatial position points; Converting the strain intensity distribution to data visualization according to the preset color gradient rule to generate a strain thermal map of the structure surface.

8. A system for spatiotemporal feature mining and risk assessment of flight operation data, characterized in that, It includes: An acquisition module is configured to acquire flight operation data and optical signal data of a mobile device, wherein the flight operation data includes component state data and environmental parameter data; A forming module is configured to form a strain thermal map of a structure surface according to the optical signal data in combination with a structural deformation field of the mobile device; An extraction module is configured to perform spatio-temporal alignment fusion on the strain thermal map, the component state data, and the environmental parameter data, and extract a spatio-temporal feature from the fusion result through an enhanced attention mechanism; An identification module is configured to perform kernel fuzzy C-means clustering on the spatio-temporal feature to identify a potential fault mode combination, and determine a flight risk assessment result according to the potential fault mode combination.

9. A computing device, comprising: A processing component and a storage component are included; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the flight operation data spatio-temporal feature mining and risk assessment method of any one of claims 1-7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, the flight operation data spatio-temporal feature mining and risk assessment method of any one of claims 1-7 is implemented.

Images