Aviation equipment fault diagnosis method and system based on dynamic coupling gradient analysis

Abstract

The invention provides an aeronautical equipment fault diagnosis method and system based on dynamic coupling gradient analysis, and relates to the technical field of aeronautical equipment fault diagnosis. According to the method, a multi-stage hole coupling propagation theory framework is constructed based on a Swiss cheese model, and a quantitative diagnosis system fusing a gradient field, hole dynamics and phase space early warning is established; intelligent parameter calibration is realized through a space-time convolutional network; constructing a coupling gradient field model to quantify a multi-factor interaction effect; describing continuous state evolution by using a hole kinetic equation; and developing a phase space early warning mechanism to identify a failure critical point. Meanwhile, the invention provides a system for the fault diagnosis method, so that the system can be applied to multiple scenes. According to the method, the inherent defects of a traditional method in processing coupling, time sequence and non-linear characteristics are overcome, accurate early warning and root cause positioning of the multi-factor coupling faults of the aviation equipment are achieved, early warning is advanced by more than 50 hours, the maintenance cost is reduced by more than 92%, and the method has remarkable practical application value.

Description

Technical Field

[0001] This invention relates to the field of aviation equipment fault diagnosis technology, and specifically to an aviation equipment fault diagnosis method and system based on dynamic coupling gradient analysis. Background Technology

[0002] As cutting-edge products of modern industry, the reliability and safety of aviation equipment are of paramount importance. However, during long-term, high-load operation, aviation equipment is inevitably affected by various factors, such as mechanical wear, material fatigue, environmental corrosion, electrical faults, and improper operation. These factors often do not act independently but are coupled, evolve sequentially, and exhibit complex nonlinear characteristics, leading to the complexity of multi-factor coupling in the failure modes of aviation equipment.

[0003] Traditional methods for fault diagnosis in aviation equipment mainly include threshold-based alarms, expert system-based methods, signal processing-based methods, and traditional machine learning model-based methods. These methods have shown some effectiveness in handling single, linear, or uncoupled faults. For example, threshold-based alarm methods determine whether the system is abnormal by setting upper and lower limits for key parameters; expert system-based methods rely on manually defined rules and empirical knowledge for fault reasoning; signal processing-based methods (such as Fourier transform and wavelet analysis) aim to extract fault features from vibration, acoustic, and other signals; while traditional machine learning models (such as support vector machines and decision trees) learn from historical data to build fault classification or prediction models.

[0004] However, with the increasing integration and complexity of aviation equipment, the aforementioned traditional methods have revealed many inherent defects in handling multi-factor coupled faults: Insufficient coupling analysis capabilities: Traditional methods often treat each monitoring parameter as independent or simply linearly correlated, making it difficult to effectively capture the complex nonlinear coupling relationships and interaction mechanisms between various components and systems within aviation equipment. This results in the inability to accurately identify the inducing factors and propagation paths of complex faults caused by multiple factors, thus affecting the accuracy of fault location.

[0005] The processing of time-series features is limited. The occurrence and evolution of aircraft equipment failures is a continuous and dynamic process with obvious time-series dependencies. Traditional methods may struggle to fully extract the temporal dynamic information and long-term dependencies contained in time-series data, resulting in relatively weak early warning and trend prediction capabilities for failures. Diagnosis is often only possible after a failure occurs, making it difficult to provide sufficient advance warning.

[0006] Modeling nonlinear characteristics is difficult because the physical processes and fault mechanisms within aerospace equipment systems are typically nonlinear. Traditional linear or simple nonlinear models struggle to effectively fit and characterize these complex nonlinear mapping relationships, resulting in insufficient adaptability and robustness of diagnostic models to real-world operating conditions, and a tendency for false alarms or missed alarms.

[0007] Insufficient early warning accuracy and lead time, limited by insufficient processing capabilities for coupling, temporal, and nonlinear characteristics, traditional fault diagnosis methods often struggle to provide accurate early warnings for aircraft equipment faults, particularly in their poor ability to identify potential and hidden faults. This results in limited early warning lead time, failing to provide sufficient time windows for maintenance decisions, thereby increasing maintenance costs and potentially threatening flight safety.

[0008] Root cause localization is difficult. When a fault occurs, traditional methods can often only provide a superficial fault classification, but it is difficult to analyze and trace the root cause of the fault, let alone reveal the fault propagation mechanism within the system. This poses a great challenge to subsequent fault investigation and maintenance strategy formulation.

[0009] In summary, existing technologies have significant shortcomings in diagnosing faults involving multi-factor coupling, temporal evolution, and nonlinear characteristics in aerospace equipment, making it difficult to meet the demands of modern aerospace equipment for high-precision, high-timeliness, and high-reliability fault diagnosis. Therefore, there is an urgent need for an advanced method that can overcome these deficiencies and achieve accurate early warning and root cause localization of multi-factor coupled faults in aerospace equipment. Summary of the Invention

[0010] To address the shortcomings of the existing technologies, the present invention aims to provide a method and system for diagnosing aviation equipment faults based on dynamic coupled gradient analysis. This method can quantify the coupling and interaction of multiple factors, achieve continuous state evolution tracking, and accurately identify aviation equipment faults with nonlinear failure phase transitions, thereby significantly improving diagnostic accuracy and early warning capabilities.

[0011] Specifically, in a first aspect, the present invention provides a method for diagnosing faults in aerospace equipment based on dynamic coupled gradient analysis, which includes the following steps: S1. Utilize sensors deployed on aviation equipment to collect raw data reflecting the operational status of aviation equipment in real time; S2. The raw data collected in step S1 is preprocessed using a parameter intelligent calibration module based on spatiotemporal convolutional networks to eliminate data noise and redundant information, and feature enhancement is performed to obtain standardized feature data. S3. Based on the standardized feature data output in step S2, a coupled gradient field model is constructed to quantify the dynamic coupling strength and interaction relationship between key parameters within the aerospace equipment. The coupled gradient field model can capture the nonlinear interaction effects between key parameters and characterize the potential propagation path and core influencing factors of faults. The expression of the coupled gradient field model is as follows: ; in, The gradient field is used for coupling, and N is the number of key parameters affecting the operating status of the aviation equipment. These represent two different key parameters or features; express and The coupling coefficient between them; Represents the loss function For key parameters and The partial derivative of the coupling relationship; M is the number of internal potential failure modes or risk factors; Gk represents the... The activation weight of each failure mode or risk factor; Indicates the first Gradient fields corresponding to various failure modes or risk factors; To adjust the weighting coefficients of the contribution of the coupling terms, Weighting coefficients used to adjust the contribution of failure modes or risk factors; S4. Based on the coupled gradient field model, a hole dynamics equation integrating the Swiss cheese model theory is established to describe the continuous evolution process of the aerospace equipment system state and the dynamic mechanism of fault propagation. The hole dynamics evolution trajectory is obtained based on the hole dynamics equation, where the expression of the hole dynamics equation is: in, This indicates the current attenuation status or hole size of the aviation equipment system; The decay rate is the rate of change of the decay state of an aviation equipment system over time, and its dimensions are... ; For a dimensionless scalar field, for The spatial second derivative of the field; The diffusion coefficient is denoted as . This is a coefficient representing the system's ability to self-repair or resist degradation. This is a function representing the effect of external environmental factors or operational disturbances on system attenuation. The weighting coefficients represent the influence of external factors on pore dynamics. S5. Based on the hole dynamics evolution trajectory obtained in step S4, the phase space early warning mechanism is used to identify the failure critical point of the aviation equipment and trigger an early warning.

[0012] Preferably, the expression for the failure critical point in step S5 is: ; in, For aviation equipment systems The key state parameter vector at each moment; for Vector of rate of change over time; The system attenuation state calculated in step S4; This is the reference parameter vector for the system under normal operating conditions. for Vector of rate of change over time; This is a safety radius or distance threshold defined in phase space. When the system state deviates from the normal trajectory by more than this threshold, it is judged as abnormal. The preset system attenuation state threshold is used when the system attenuation state is... When the value is below this threshold, it indicates that the system has entered a high-risk failure zone; ∨ is a logical OR operator, indicating that one of the conditions is met, which triggers an early warning.

[0013] Preferably, the data in step S1 includes vibration signals, temperature, pressure, current, and voltage.

[0014] Preferably, the parameter intelligent calibration module based on spatiotemporal convolutional networks described in step S2 achieves feature enhancement in the following way: a. Targeted design of multi-layer spatiotemporal convolutional structure: Construct a deep network consisting of alternating temporal and spatial convolutional layers; wherein, the temporal convolutional layer is used to capture the hidden temporal dependencies and dynamic patterns in the raw data collected by different sensors; the spatial convolutional layer is used to learn the spatial correlation and cooperative change patterns between different sensors; b. Introduce an attention mechanism to focus on key information: embed an attention module in the spatiotemporal convolutional network so that the model can adaptively learn and assign higher weights to time periods and sensor areas that are of priority for fault diagnosis; c. Integrating residual connections and batch normalization to optimize training: Deep networks are constructed using residual connections; at the same time, batch normalization technology is used to accelerate the convergence speed of the model and enhance its adaptability to changes in data under different working conditions. d. Optimize the feature space using contrastive learning: During the model training phase, a contrastive learning strategy is introduced to reduce the distance between feature representations of similar states, while increasing the distance between feature representations of different states.

[0015] Preferably, step S3, constructing the coupled gradient field model, specifically includes the following sub-steps: S31. Parameter Selection and Initialization: Based on domain knowledge and historical data analysis of aviation equipment, select N key parameters closely related to the operational status and initialize the parameters. and Coupling coefficient between and the Activation weights of different failure modes ; S32. Update Coupling Coefficients: Use nonlinear regression models or deep learning models to learn and update parameters from historical data. and Coupling coefficient between ; S33. Define the loss function: Define a loss function L to quantify the deviation between the current system state and the desired healthy state; S34. Gradient Calculation and Update: Using backpropagation or adaptive optimization algorithms, iteratively calculate the loss function L with respect to parameters. and Coupling coefficient between and the Activation weights of different failure modes The partial derivatives are calculated and the parameters in the coupled gradient field model are updated until the model converges.

[0016] Preferably, solving the hole dynamics equation in step S4 specifically includes the following sub-steps: S41. Attenuation State Initialization: Based on the initial health status of the aircraft equipment or the maintenance records after the update, initialize the current attenuation state of the system. ; S42. Parameter Calibration and Optimization: Using historical fault data, the sensitivity coefficient of the coupled gradient field is adjusted. System self-repair capability coefficient and external factor weighting coefficients Perform calibration and optimization; S43. External Factor Modeling: Establish a mathematical model of external environmental factors or operational disturbances F(t); S44. Real-time iterative solution: based on the real-time acquired raw data and the calculated coupled gradient field. The system's decay state is dynamically updated by iteratively solving the hole dynamics equations in real time using numerical integration methods. and The rate of change was determined, and the dynamic evolution trajectory of the pores was obtained.

[0017] Preferably, the specific method for obtaining the dynamic evolution trajectory of the pore based on the pore dynamics equation in step S44 includes: S441. Discretize the time axis into a form with a fixed step size. Multiple points in time; S442, at each time point Obtain the real-time coupled gradient field calculated in step S3. and external disturbance function ; S443. Using numerical integration, based on the current time... System attenuation state and coupled gradient field Calculate the next time step System attenuation state ; S444, System decay state calculated at all time points The connections form the dynamic evolution trajectory of the pores.

[0018] Preferably, step S5 specifically includes the following sub-steps: S51. Construct a health status baseline: During the normal operation of aviation equipment, collect real-time data and use principal component analysis or kernel principal component analysis to construct the normal operation trajectory of the system and the reference parameter vector and its rate of change vector in the multidimensional feature space. S52. High-Risk Region Criterion Learning: Combining historical fault data and expert knowledge, this method uses machine learning classifiers or statistical methods to learn and define high-risk region criteria in phase space, i.e., determining the safe radius or distance threshold. and system decay state threshold ; S53. Real-time state projection and distance calculation: Project the standardized feature data extracted in real time onto the constructed phase space, and calculate the Euclidean distance or other metric distance between the real-time state parameter vector and the health state baseline. S54. Multi-condition fusion early warning: Real-time determination of whether the current system state meets any of the following conditions: The distance between the real-time state and the reference trajectory exceeds the safety radius or distance threshold. Or the system attenuation state calculated in real time. Exceeding the preset decay threshold If any of the conditions are met, an early warning will be triggered.

[0019] Preferably, the method further includes step S6: based on the early warning information, combined with the fault mode library and expert system, perform fault root cause localization and fault propagation path analysis, and output corresponding maintenance suggestions or early warning decisions to form a diagnostic closed loop, specifically including the following sub-steps: S61. Fault mode library matching: Match the feature patterns corresponding to the identified failure critical points with the pre-established fault mode library to preliminarily determine the fault type. S62. Root Cause Analysis: By combining key influencing factors and failure propagation paths, trace the root cause of the failure. S63. Maintenance Decision and Early Warning Strategy: Based on fault type, severity, and remaining life prediction, generate detailed maintenance recommendations and early warning strategies. S64. Knowledge Base Update: Update the system knowledge base with new fault modes, parameter calibration results and warning thresholds to continuously optimize the performance of the diagnostic model.

[0020] Secondly, this invention provides a multi-factor coupled fault diagnosis system for aerospace equipment based on dynamic coupled gradient analysis. It includes a data acquisition module, a parameter intelligent calibration module, a coupled gradient field construction module, a pore dynamics calculation module, and an early warning module; The data acquisition module uses sensors deployed on aviation equipment to collect raw data reflecting the operational status of the aviation equipment in real time; The parameter intelligent calibration module preprocesses the collected raw data, eliminates data noise and redundant information, and performs feature enhancement to obtain standardized feature data. The Coupled Gradient Field Construction Module is used to construct a Coupled Gradient Field Model to quantify the dynamic coupling strength and interaction relationship between key parameters inside aerospace equipment. The Coupled Gradient Field Model can capture the nonlinear interaction effect between key parameters and characterize the potential propagation path and core influencing factors of faults. The hole dynamics calculation module is used to establish hole dynamics equations that integrate the Swiss cheese model theory based on the coupled gradient field model, describe the continuous evolution process of the state of aerospace equipment systems and the dynamic mechanism of fault propagation, and obtain the hole dynamics evolution trajectory based on the hole dynamics equations. Based on the obtained hole dynamics evolution trajectory, the early warning module uses a phase space early warning mechanism to identify the failure critical point of the aviation equipment and trigger an early warning.

[0021] Compared with the prior art, the present invention has the following beneficial effects: (1) The method for diagnosing multi-factor coupled faults of aviation equipment based on dynamic coupled gradient analysis provided by the present invention can quantify the coupling interaction of multiple factors by constructing a coupled gradient field model, realize continuous state evolution tracking, and accurately identify aviation equipment faults with nonlinear failure phase transitions, thereby greatly improving diagnostic accuracy and early warning capability, and has great application potential.

[0022] (2) The multi-factor coupled fault diagnosis method for aviation equipment based on dynamic coupled gradient analysis provided by the present invention establishes a hole dynamics equation that integrates the Swiss cheese model theory on the basis of the coupled gradient field model, describes the continuous evolution process of the aviation equipment system state and the dynamic mechanism of fault propagation, and obtains the hole dynamics evolution trajectory based on the hole dynamics equation, thereby providing a better foundation for fault early warning.

[0023] (3) The phase space early warning mechanism of the present invention, by comprehensively considering the static deviation and dynamic evolution of the system state and combining the quantified attenuation degree, realizes accurate early warning and root cause location of multi-factor coupled faults of aviation equipment, significantly improves the timeliness and accuracy of early warning, and achieves the effect of early warning advance of up to 50 hours and maintenance cost reduction of 92%.

[0024] (4) The multi-factor coupled fault diagnosis method for aviation equipment based on dynamic coupled gradient analysis provided by the present invention greatly enhances the separability of fault features by performing feature enhancement on the original data, providing high-quality and high-discrimination input for subsequent fault diagnosis and early warning, thereby improving the accuracy and robustness of the entire diagnosis method. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a diagram illustrating the overall architecture of the method of the present invention; Figure 3 This is a schematic diagram illustrating the working principle of the phase space early warning mechanism of the present invention; Figure 4 This is an overall system block diagram of the present invention. Detailed Implementation

[0026] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0027] This invention provides a fault diagnosis method for aerospace equipment based on dynamic coupling gradient analysis, such as... Figure 1 As shown, it includes the following steps: S1. Use sensors deployed on aviation equipment to collect raw data reflecting the operating status of aviation equipment in real time.

[0028] S2. The raw data collected in step S1 is preprocessed using a parameter intelligent calibration module based on spatiotemporal convolutional networks to eliminate data noise and redundant information, and feature enhancement is performed to obtain standardized feature data.

[0029] S3. Based on the standardized feature data output in step S2, a coupled gradient field model is constructed to quantify the dynamic coupling strength and interaction relationship between key parameters within the aerospace equipment. The coupled gradient field model can capture the nonlinear interaction effects between key parameters and characterize the potential propagation path and core influencing factors of faults. The expression of the coupled gradient field model is as follows: ; in, For coupled gradient fields, The number of key parameters affecting the operational status of aviation equipment; These represent two different key parameters or features; express and The coupling coefficient between them; Represents the loss function For key parameters and Partial derivatives of the coupling relationship; The number of potential internal failure modes or risk factors; Indicates the first The activation weight of each failure mode or risk factor; Indicates the first Gradient fields corresponding to various failure modes or risk factors; To adjust the weighting coefficients of the contribution of the coupling terms, Weighting coefficients used to adjust the contribution of failure modes or risk factors.

[0030] Step S3, constructing the coupled gradient field model, specifically includes the following sub-steps: S31. Parameter Selection and Initialization: Based on domain knowledge of aviation equipment and historical data analysis, parameters are selected... Key parameters closely related to the operating status were identified and initialized. and Coupling coefficient between and the Activation weights of different failure modes .

[0031] S32. Update Coupling Coefficients: Use nonlinear regression models or deep learning models to learn and update parameters from historical data. and Coupling coefficient between .

[0032] S33. Define the loss function: Define a loss function L to quantify the deviation between the current system state and the desired healthy state.

[0033] S34. Gradient Calculation and Update: Using backpropagation or adaptive optimization algorithms, iteratively calculate the loss function L with respect to parameters. and Coupling coefficient between and the Activation weights of different failure modes The partial derivatives are calculated and the parameters in the coupled gradient field model are updated until the model converges.

[0034] S4. Based on the coupled gradient field model, a hole dynamics equation integrating the Swiss cheese model theory is established to describe the continuous evolution process of the aerospace equipment system state and the dynamic mechanism of fault propagation. The hole dynamics evolution trajectory is obtained based on the hole dynamics equation, thus yielding the real-time system decay state. The expression for the hole dynamics equation is: in, This indicates the current attenuation status or hole size of the aviation equipment system; The decay rate is the rate of change of the decay state of an aviation equipment system over time, and its dimensions are... ; For a dimensionless scalar field, for The spatial second derivative of the field; The diffusion coefficient is denoted as . This is a coefficient representing the system's ability to self-repair or resist degradation. This is a function representing the effect of external environmental factors or operational disturbances on system attenuation. The weighting coefficients represent the influence of external factors on pore dynamics.

[0035] Step S4, solving the hole dynamics equations, specifically includes the following sub-steps: S41. Attenuation State Initialization: Based on the initial health status of the aircraft equipment or the maintenance records after the update, initialize the current attenuation state of the system. .

[0036] S42. Parameter Calibration and Optimization: Using historical fault data, the sensitivity coefficient of the coupled gradient field is adjusted. System self-repair capability coefficient and external factor weighting coefficients Perform calibration and optimization.

[0037] S43. External Factor Modeling: Establishing external environmental factors or operational disturbances. The mathematical model.

[0038] S44. Real-time iterative solution: based on the real-time acquired raw data and the calculated coupled gradient field. The system's decay state is dynamically updated by iteratively solving the hole dynamics equations in real time using numerical integration methods. and The rate of change is calculated, and the dynamic evolution trajectory of the pore is obtained. Specifically, the method for obtaining the dynamic evolution trajectory of the pore based on the pore dynamic equation includes: S441. Discretize the time axis into a form with a fixed step size. Multiple time points.

[0039] S442, at each time point Obtain the real-time coupled gradient field calculated in step S3. and external disturbance function .

[0040] S443. Using numerical integration, based on the current time... System attenuation state and coupled gradient field Calculate the next time step System attenuation state .

[0041] S444, System decay state calculated at all time points The connections form the dynamic evolution trajectory of the pores.

[0042] S5. Based on the hole dynamics evolution trajectory obtained in step S4, the phase space early warning mechanism is used to identify the failure critical point of the aerospace equipment and trigger an early warning. The expression for the failure critical point is: ; in, For aviation equipment systems The key state parameter vector at each moment; for Vector of rate of change over time; The system attenuation state calculated in step S4; This is the reference parameter vector for the system under normal operating conditions. for Vector of rate of change over time; This is a safety radius or distance threshold defined in phase space. When the system state deviates from the normal trajectory by more than this threshold, it is judged as abnormal. The preset system attenuation state threshold is used when the system attenuation state is... When the value is below this threshold, it indicates that the system has entered a high-risk failure zone; ∨ is a logical OR operator, indicating that one of the conditions is met, which triggers an early warning.

[0043] Step S5 specifically includes the following sub-steps: S51. Construct a health status baseline: During the normal operation of aviation equipment, collect real-time data and use principal component analysis or kernel principal component analysis to construct the normal operation trajectory of the system and the reference parameter vector and its rate of change vector in the multidimensional feature space.

[0044] S52. High-Risk Region Criterion Learning: Combining historical fault data and expert knowledge, this method uses machine learning classifiers or statistical methods to learn and define high-risk region criteria in phase space, i.e., determining the safe radius or distance threshold. and system decay state threshold .

[0045] S53. Real-time state projection and distance calculation: Project the real-time extracted standardized feature data into the constructed phase space, and calculate the Euclidean distance or other metric distance between the real-time state parameter vector and the health state baseline.

[0046] S54. Multi-condition fusion early warning: Real-time determination of whether the current system state meets any of the following conditions: The distance between the real-time state and the reference trajectory exceeds the safety radius or distance threshold. Or the system attenuation state calculated in real time. Exceeding the preset decay threshold If any of the conditions are met, an early warning will be triggered.

[0047] The method also includes step S6: Based on the early warning information, combined with the fault mode library and expert system, perform fault root cause localization and fault propagation path analysis, and output corresponding maintenance suggestions or early warning decisions to form a diagnostic closed loop, which specifically includes the following sub-steps: S61. Fault mode library matching: Match the feature patterns corresponding to the identified failure critical points with the pre-established fault mode library to preliminarily determine the fault type. S62. Root Cause Analysis: By combining key influencing factors and failure propagation paths, trace the root cause of the failure. S63. Maintenance Decision and Early Warning Strategy: Based on fault type, severity, and remaining life prediction, generate detailed maintenance recommendations and early warning strategies. S64. Knowledge Base Update: Update the system knowledge base with new fault modes, parameter calibration results and warning thresholds to continuously optimize the performance of the diagnostic model.

[0048] Secondly, this invention provides a multi-factor coupled fault diagnosis system for aerospace equipment based on dynamic coupled gradient analysis. like Figure 4 As shown, it includes a data acquisition module 1, a parameter intelligent calibration module 2, a coupled gradient field construction module 3, a hole dynamics calculation module 4, and an early warning module 5.

[0049] Data acquisition module 1 uses sensors deployed on aviation equipment to collect raw data reflecting the operating status of aviation equipment in real time.

[0050] The parameter intelligent calibration module 2 preprocesses the collected raw data, eliminates data noise and redundant information, and performs feature enhancement to obtain standardized feature data.

[0051] The Coupled Gradient Field Construction Module 3 is used to construct a Coupled Gradient Field Model to quantify the dynamic coupling strength and interaction relationship between key parameters inside aerospace equipment. The Coupled Gradient Field Model can capture the nonlinear interaction effect between key parameters and characterize the potential propagation path and core influencing factors of faults.

[0052] The pore dynamics calculation module 4 is used to establish a pore dynamics equation that integrates the Swiss cheese model theory based on the coupled gradient field model, to describe the continuous evolution process of the state of the aerospace equipment system and the dynamic mechanism of fault propagation, and to obtain the pore dynamics evolution trajectory based on the pore dynamics equation.

[0053] Based on the obtained hole dynamics evolution trajectory, the early warning module 5 uses a phase space early warning mechanism to identify the failure critical point of the aviation equipment and trigger an early warning. Specific Implementation This invention provides a fault diagnosis method for aviation equipment based on dynamic coupled gradient analysis. Through real-time data acquisition, intelligent preprocessing, coupled gradient field construction, establishment of pore dynamics equations, and a phase space early warning mechanism, it achieves precise monitoring of the operational status of aviation equipment and early warning of faults. Figure 1 and Figure 2 As shown, it includes the following steps: S1. Raw data acquisition.

[0055] In this embodiment, high-precision sensors (such as piezoelectric accelerometers, thermocouples, pressure sensors, Hall effect sensors, etc.) deployed on key components of aviation equipment such as aero engines, landing gear, and flight control systems are used to collect raw data reflecting the operating status of the equipment in real time. This data includes, but is not limited to: Vibration signal: The sampling frequency can be set from 10kHz to 50kHz to capture fault characteristics such as gear wear, bearing damage, and structural loosening.

[0056] Temperature: Sampling frequency from 1Hz to 10Hz, monitoring engine exhaust temperature, bearing temperature, lubricating oil temperature, etc., reflecting thermal management and friction status.

[0057] Pressure: Sampling frequency from 10Hz to 100Hz, monitoring hydraulic system pressure, fuel system pressure, pneumatic pressure, etc., indicating fluid dynamic performance.

[0058] Current and voltage: Sampling frequency from 1kHz to 10kHz, used to monitor the operating status of electrical components such as motors and actuators, and to determine whether there are abnormalities such as short circuits, open circuits, and overloads. All collected raw data are timestamped for subsequent time-series analysis.

[0059] S2. Preprocessing using a parameter intelligent calibration module based on a spatiotemporal convolutional network (STCNN): The raw data collected in step S1 is preprocessed using a parameter intelligent calibration module based on a spatiotemporal convolutional network (STCNN). Specifically, the STCNN structure includes: Temporal convolutional layers: These layers use one-dimensional convolutional kernels to extract features from the time series of sensor data, capturing temporal dependencies such as periodicity and trend changes. For example, for vibration signals, high-frequency impact and low-frequency flutter features can be extracted using convolutional kernels of different scales.

[0060] Spatial convolutional layer: A two-dimensional convolutional kernel is used to process feature maps from different sensors (considered as spatially related points) to capture the spatial correlation between sensors. For example, multiple temperature sensors may jointly indicate overheating in a region, or multiple pressure sensors may work together to reflect system leaks.

[0061] Attention mechanism: Self-attention or cross-attention mechanism is introduced in STCNN to dynamically adjust the weights of different temporal and spatial features, so that the network pays more attention to key parameters and time periods that are strongly related to the fault.

[0062] Data denoising: By utilizing the filtering properties of convolutional networks and learning effective feature representations, high-frequency random noise and occasional interference can be automatically filtered out.

[0063] Redundant information elimination: STCNN can identify and remove highly relevant redundant features through automatic feature learning, retaining only the information most critical to the system state representation.

[0064] Feature Enhancement and Optimization: The features learned by STCNN are nonlinear mappings of the original data, which can more effectively distinguish between normal and abnormal states. Through training, STCNN automatically standardizes and normalizes these extracted features (e.g., Min-Max normalization or Z-score normalization), outputting standardized feature data that more accurately reflects the system state. For example, for vibration data of aero-engines, STCNN can transform the original vibration waveform into dimensionless feature values ​​reflecting the degree of bearing wear.

[0065] The parameter intelligent calibration module based on spatiotemporal convolutional networks described in step S2, in order to efficiently process multi-source and heterogeneous sensor data from aviation equipment and extract deep features highly correlated with fault evolution, innovatively integrates a variety of advanced deep learning technologies to achieve feature enhancement in the following ways: a. Targeted design of a multi-layer spatiotemporal convolutional structure: Construct a deep network consisting of alternating temporal and spatial convolutional layers. The temporal convolutional layers are specifically designed to capture the temporal dependencies and dynamic patterns hidden in high-frequency sensor data such as vibration and pressure; the spatial convolutional layers are specifically designed to learn the spatial correlations and cooperative change patterns between different sensors (such as temperature and vibration sensors distributed at different locations on the engine), thereby comprehensively depicting the overall operational status of the equipment.

[0066] b. Introducing an attention mechanism to focus on key information: An attention module is embedded in the spatiotemporal convolutional network, enabling the model to adaptively learn and assign higher weights to the time periods and sensor regions most indicative of fault diagnosis. For example, in the early stages of a fault, this mechanism can automatically focus on weak abnormal signals, overcoming the problem of key information being overwhelmed by noise in traditional methods.

[0067] c. Integrating residual connections and batch normalization to optimize training: Using residual connections to build deep networks effectively solves the gradient vanishing problem that may occur when the depth increases, making it possible to build more complex models; at the same time, using batch normalization technology significantly accelerates the convergence speed of the model and enhances its adaptability (generalization ability) to changes in data under different working conditions.

[0068] d. Optimizing the Feature Space using Contrastive Learning: During the model training phase, a contrastive learning or metric learning strategy is introduced. The core idea is to actively cluster feature representations of similar states (e.g., different levels of normal operation) while separating feature representations of different states (e.g., normal and fault states) in the feature space. This significantly enhances the separability of fault features, providing high-quality, highly discriminative input for subsequent fault diagnosis and early warning, thereby improving the accuracy and robustness of the entire diagnostic method.

[0069] S3. Construct a coupled gradient field model: Based on the standardized feature data output in step S2, a coupled gradient field model is constructed to quantify the dynamic coupling strength and interaction relationships among various factors within the aerospace equipment. Its expression is: ; in: The number of key parameters affecting the operational status of aviation equipment, for example, for turbine engines, It can include several key parameters such as rotational speed, vibration, temperature, pressure, and lubricating oil flow. These represent two different key parameters or features, and are standardized feature data output from step S2. Indicates parameters and The coupling coefficient between parameters can be calculated using various methods, such as: Pearson correlation coefficient (to measure the strength of linear coupling); mutual information (to measure the strength of nonlinear coupling and capture more complex dependencies); Granger causality (to measure time-delay coupling and determine whether a change in one parameter leads to a subsequent change in another); and deep learning-based coupling metrics (calculated by learning complex nonlinear mappings between parameters using neural networks).

[0070] Represents the loss function For parameters and Partial derivatives of the coupling relationship. Loss function. This can be defined as the error between the predicted fault and the actual fault, such as the mean squared error (MSE) or cross-entropy. The partial derivative, calculated using the backpropagation algorithm, reflects the impact of changes in coupling strength on the system state.

[0071] The number of potential failure modes or risk factors within the system, for example, It can include various specific fault modes such as bearing wear, blade breakage, hydraulic leakage, and circuit short circuit.

[0072] Indicates the first The activation weights or sensitivities of a fault mode can be obtained through historical fault data analysis, expert experience, or training of machine learning models (such as support vector machines and neural networks). The larger the value, the more significant the impact of the failure mode on system degradation.

[0073] Indicates the first The gradient field corresponding to each failure mode describes the evolution direction and intensity of that mode in the feature space. For example, for bearing wear failure, It may indicate changes in characteristics such as an increase in high-frequency components of the vibration signal and an increase in temperature.

[0074] The weighting coefficients used to adjust the contributions of coupling terms and failure mode terms are hyperparameters trained and optimized based on the characteristics and historical data of the aerospace equipment. For example, optimal parameters are determined by minimizing the prediction error on the training set using optimization methods such as grid search or genetic algorithms. and value.

[0075] S4. Establish the pore dynamics equations that incorporate the Swiss cheese model theory. : Based on the coupled gradient field model, a dynamic equation for the holes, incorporating the Swiss cheese model theory, is established. This term describes the continuous evolution of the state of an aviation equipment system and the dynamic mechanism of fault propagation. Its expression is: ; in, This indicates the current attenuation status or hole size of the aviation equipment system. The value is between 0 and 1, where 0 indicates complete health and 1 indicates complete failure. The decay rate is the rate of change of the decay state of an aviation equipment system over time, and its dimensions are: (e.g., / hour, / day); Defined as a dimensionless scalar field, representing local stress, damage accumulation, or risk potential at different locations (x, y, z) within a system. For example, on an engine turbine blade... The value may be much higher than that of the passenger seat. value; (Laplace operator): Spatial second derivative of the field It measures the field. The degree of local "unevenness" at a certain point. A sharp risk peak (such as a tiny crack) can have a significant impact. A value signifies a strong "source of diffusion"; The diffusion coefficient describes the rate at which risk or damage propagates within a system. This is a coefficient representing the system's ability to self-repair or resist degradation. This is a function representing the effect of external environmental factors or operational disturbances on system attenuation. The weighting coefficients represent the influence of external factors on pore dynamics.

[0076] Step S4, solving the hole dynamics equations, specifically includes the following sub-steps: S41. Attenuation State Initialization: Based on the initial health status of the aircraft equipment or the maintenance records after the update, initialize the current attenuation state of the system. ; S42. Parameter Calibration and Optimization: Using historical fault data, the sensitivity coefficient of the coupled gradient field is adjusted. System self-repair capability coefficient and external factor weighting coefficients Perform calibration and optimization; S43. External Factor Modeling: Establishing external environmental factors or operational disturbances. Mathematical model; S44. Real-time iterative solution: based on the real-time acquired raw data and the calculated coupled gradient field. The system's decay state is dynamically updated by iteratively solving the hole dynamics equations in real time using numerical integration methods. and The rate of change was determined, and the dynamic evolution trajectory of the pores was obtained.

[0077] Since this equation is an ordinary differential equation, and its driving term The real-time calculations from S3 and the external perturbation term f_ext(t) are dynamically changing over time, making analytical solutions impossible. Therefore, this invention employs a numerical integration method for iterative solutions to obtain the dynamic evolution trajectory of the pores. The specific implementation process is as follows: S441, Time Discretization and Parameter Initialization. The continuous runtime axis is divided into a series of discrete time points. The time step is ,Right now . The value needs to be determined based on the sensor data sampling frequency and the system's dynamic response characteristics; for example, it can be set to 0.1 seconds. According to step S41, the initial attenuation state of the system is initialized. .

[0078] S442, Real-time input acquisition. At any given calculation time... The system obtains all necessary inputs at that moment from the upstream module: it obtains the real-time calculated coupled gradient field from the coupled gradient field construction module. And calculate its Laplace value. Obtain the disturbance function values ​​under the current environmental and operating conditions from the external factors modeling module*. Get the decay state at the current moment. (The values ​​used in the first iteration are initial values; subsequent iterations use the results calculated in the previous step.) Obtain the model coefficients calibrated in step S42. .

[0079] S443. Iteratively solve for the state at the next time step. This embodiment preferably uses the fourth-order Runge-Kutta method (RK4) for solving, as it offers a good balance between accuracy and computational stability. The specific calculation process is as follows: make Then the hole dynamics equation can be simplified as follows: .

[0080] From arrive In the iterative steps, the calculation process is as follows: 1. Calculate the slope ; 2. Calculate the slope ; 3. Calculate the slope ; 4. Calculate the slope ; 5. Calculate and update the decay state at the next time step: ; In another embodiment, to simplify the calculation, the forward Euler method can also be used for the solution, and its iterative formula is as follows: ; S444. Construct the evolutionary trajectory. Repeat steps two and three; the system will calculate the evolutionary trajectory sequentially. A series of discrete decay state values. Connecting these values ​​in chronological order constitutes the complete dynamic evolution trajectory of the pore. This trajectory visually demonstrates the complete process of the evolution of the health status of aviation equipment over time, and is key input data for the subsequent step S5 to conduct phase space early warning.

[0081] Through the detailed numerical solution steps described above, this invention transforms a theoretical differential equation model into an algorithm that can be implemented in a computer system and run in real time, ensuring the technical feasibility and operability of the entire fault diagnosis method.

[0082] S5, Based on phase space early warning mechanism Identify failure critical points and trigger early warnings: Based on the pore dynamics evolution trajectory obtained in step S4, utilize the phase space early warning mechanism. Identify the critical failure points of aviation equipment and trigger early warnings. The mechanism of the phase-space early warning mechanism is as follows: Figure 3 As shown. The expression for the failure critical point is: in: Xt represents the aviation equipment system in The key state parameter vector at each moment contains features extracted and calibrated from S2, such as the normalized vibration features and temperature features output by STCNN.

[0083] for The rate of change vector over time reflects the dynamic trend of the system state and can be obtained by analyzing... This can be obtained by methods such as differential or Kalman filtering.

[0084] This is a reference parameter vector for the system under normal operating conditions, which is usually obtained by modeling health data, such as by establishing a multivariate Gaussian model.

[0085] This is the reference vector for the rate of change of the system under normal operating conditions.

[0086] ; This represents the Euclidean distance between the current state of the system and the normal reference state in phase space. This distance measures the static deviation and dynamic evolution of the system state.

[0087] This is a safety radius or distance threshold defined in phase space. When the system state deviates from the normal trajectory by more than this threshold, it is judged as abnormal. The preset system attenuation state threshold is used when the system attenuation state is... When the value falls below this threshold, it indicates that the system has entered a high-risk failure zone; ∨ represents the logical "OR" operator, meaning that an alert is triggered as long as one of the conditions is met.

[0088] This phase space early warning mechanism, by comprehensively considering the static deviation and dynamic evolution of the system state and combining the quantified attenuation degree, achieves accurate early warning and root cause location of multi-factor coupled failures of aviation equipment, significantly improving the timeliness and accuracy of early warning, achieving an early warning lead time of up to 50 hours and a 92% reduction in maintenance costs.

[0089] when When the conditions are met, the system immediately triggers an early warning and notifies maintenance personnel via human-machine interface, SMS, email, and other means. The warning information includes the current system status, predicted failure mode, remaining life estimate, and recommended maintenance measures. This mechanism enables precise early warning and root cause localization of multi-factor coupled failures in aviation equipment, significantly improving the timeliness and accuracy of warnings.

[0090] In a specific embodiment, the method further includes step S6: based on the early warning information, combined with the fault mode library and expert system, perform fault root cause localization and fault propagation path analysis, and output corresponding maintenance suggestions or early warning decisions to form a diagnostic closed loop.

[0091] S61. Fault Mode Library Matching: Match the feature patterns corresponding to the identified failure critical points with the pre-established fault mode library to preliminarily determine the possible fault types.

[0092] S62. Root Cause Tracing: By combining the key influencing factors and fault propagation paths revealed by the coupled gradient field model, the root cause of the failure can be traced.

[0093] S63. Maintenance Decisions and Early Warning Strategies: Based on the fault type, severity, and remaining life prediction, generate detailed maintenance recommendations (such as replacing parts and adjusting parameters) and early warning strategies (such as reducing operating load and scheduling maintenance), and provide feedback to aviation equipment managers to guide them in carrying out preventive maintenance.

[0094] S64. Knowledge Base Update: Update the system knowledge base with the experience gained from this fault diagnosis and handling, including new fault modes, parameter calibration results, and early warning thresholds, to continuously optimize the performance of the diagnostic model.

[0095] In this embodiment, the present invention achieves a 50-hour advance warning and a 92% reduction in maintenance costs through the above method. For example, in a field verification of a certain type of aero-engine, this method successfully predicted vibration anomalies caused by early bearing wear, issuing an early warning 48 hours earlier than traditional vibration monitoring methods. This allows the maintenance team to conduct planned maintenance before the fault spreads, avoiding expensive engine replacement and reducing maintenance costs from millions of dollars to hundreds of thousands of dollars.

[0096] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A fault diagnosis method for aviation equipment based on dynamic coupled gradient analysis, characterized in that: It includes: S1. Utilize sensors deployed on aviation equipment to collect raw data reflecting the operational status of aviation equipment in real time; S2. The raw data collected in step S1 is preprocessed using a parameter intelligent calibration module based on spatiotemporal convolutional networks to eliminate data noise and redundant information, and feature enhancement is performed to obtain standardized feature data. S3. Based on the standardized feature data output in step S2, a coupled gradient field model is constructed to quantify the dynamic coupling strength and interaction relationship between key parameters within the aerospace equipment. The coupled gradient field model can capture the nonlinear interaction effects between key parameters and characterize the potential propagation path and core influencing factors of faults. The expression of the coupled gradient field model is as follows: ; in, For coupled gradient fields, The number of key parameters affecting the operational status of aviation equipment; These represent two different key parameters or features; express and Coupling coefficient between Represents the loss function For key parameters and Partial derivatives of the coupling relationship; The number of potential internal failure modes or risk factors; Indicates the first The activation weight of each failure mode or risk factor; Indicates the first Gradient fields corresponding to various failure modes or risk factors; To adjust the weighting coefficients of the contribution of the coupling terms, Weighting coefficients used to adjust the contribution of failure modes or risk factors; S4. Based on the coupled gradient field model, a hole dynamics equation integrating the Swiss cheese model theory is established to describe the continuous evolution process of the aerospace equipment system state and the dynamic mechanism of fault propagation. The hole dynamics evolution trajectory is obtained based on the hole dynamics equation, where the expression of the hole dynamics equation is: ; in, This indicates the current attenuation status or hole size of the aviation equipment system; The decay rate represents the rate of change of the decay state of an aviation equipment system over time, and its dimensions are... ; For a dimensionless scalar field, for The spatial second derivative of the field; The diffusion coefficient is denoted as . This is a coefficient representing the system's ability to self-repair or resist degradation. This is a function representing the effect of external environmental factors or operational disturbances on system attenuation. The weighting coefficients represent the influence of external factors on pore dynamics. S5. Based on the hole dynamics evolution trajectory obtained in step S4, the phase space early warning mechanism is used to identify the failure critical point of the aviation equipment and trigger an early warning.

2. The method for fault diagnosis of aerospace equipment based on dynamic coupling gradient analysis according to claim 1, characterized in that: The expression for the failure critical point in step S5 is: ; in, This is the vector of key state parameters of the aerospace equipment system at time t; for Vector of rate of change over time; The system attenuation state calculated in step S4; This is the reference parameter vector for the system under normal operating conditions. for Vector of rate of change over time; This is a safety radius or distance threshold defined in phase space. When the system state deviates from the normal trajectory by more than this threshold, it is judged as abnormal. The preset system attenuation state threshold is used when the system attenuation state is... When the value is below this threshold, it indicates that the system has entered a high-risk failure zone; ∨ is a logical OR operator, indicating that one of the conditions is met, which triggers an early warning.

3. The method for fault diagnosis of aerospace equipment based on dynamic coupling gradient analysis according to claim 1, characterized in that: The data mentioned in step S1 includes vibration signals, temperature, pressure, current, and voltage.

4. The method for fault diagnosis of aerospace equipment based on dynamic coupling gradient analysis according to claim 1, characterized in that: The parameter intelligent calibration module based on spatiotemporal convolutional networks described in step S2 achieves feature enhancement in the following ways: a. Targeted design of multi-layer spatiotemporal convolutional structure: Construct a deep network consisting of alternating temporal and spatial convolutional layers; wherein, the temporal convolutional layer is used to capture the hidden temporal dependencies and dynamic patterns in the raw data collected by different sensors; the spatial convolutional layer is used to learn the spatial correlation and cooperative change patterns between different sensors; b. Introduce an attention mechanism to focus on key information: embed an attention module in the spatiotemporal convolutional network so that the model can adaptively learn and assign higher weights to time periods and sensor areas that are of priority for fault diagnosis; c. Integrating residual connections and batch normalization to optimize training: Deep networks are constructed using residual connections; at the same time, batch normalization technology is used to accelerate the convergence speed of the model and enhance its adaptability to changes in data under different working conditions. d. Optimize the feature space using contrastive learning: During the model training phase, a contrastive learning strategy is introduced to reduce the distance between feature representations of similar states, while increasing the distance between feature representations of different states.

5. The method for fault diagnosis of aerospace equipment based on dynamic coupling gradient analysis according to claim 1, characterized in that: Step S3, constructing the coupled gradient field model, specifically includes the following sub-steps: S31. Parameter Selection and Initialization: Based on domain knowledge and historical data analysis of aviation equipment, select N key parameters closely related to the operational status and initialize the parameters. and Coupling coefficient between The activation weight Gk for the k-th fault mode; S32. Update the coupling coefficient: Use a nonlinear regression model or a deep learning model to learn and update the coupling coefficient C(xi,xj) between parameters xi and xj from historical data; S33. Define the loss function: Define a loss function L to quantify the deviation between the current system state and the desired healthy state; S34. Gradient Calculation and Update: Using backpropagation or adaptive optimization algorithms, iteratively calculate the partial derivatives of the loss function L with respect to the coupling coefficient C(xi,xj) between parameters xi and xj and the activation weight Gk of the k-th fault mode, and update the parameters in the coupled gradient field model until the model converges.

6. The method for fault diagnosis of aerospace equipment based on dynamic coupling gradient analysis according to claim 1, characterized in that: Step S4, solving the hole dynamics equations, specifically includes the following sub-steps: S41. Attenuation State Initialization: Based on the initial health status of the aircraft equipment or the maintenance records after the update, initialize the current attenuation state of the system. ; S42. Parameter Calibration and Optimization: Using historical fault data, the sensitivity coefficient κ of the coupled gradient field and the system self-repair capability coefficient are calibrated and optimized. and external factor weighting coefficients Perform calibration and optimization; S43. External Factor Modeling: Establish a mathematical model of external environmental factors or operational disturbances F(t); S44. Real-time iterative solution: based on the real-time acquired raw data and the calculated coupled gradient field. The system's decay state is dynamically updated by iteratively solving the hole dynamics equations in real time using numerical integration methods. and The rate of change was determined, and the dynamic evolution trajectory of the pores was obtained.

7. The method for fault diagnosis of aerospace equipment based on dynamic coupling gradient analysis according to claim 5, characterized in that: The specific method for obtaining the dynamic evolution trajectory of a hole based on the hole dynamics equation in step S44 includes: S441. Discretize the time axis into a form with a fixed step size. Multiple points in time; S442, at each time point Obtain the real-time coupled gradient field calculated in step S3. and external disturbance function ; S443. Using numerical integration, based on the current time... System attenuation state and coupled gradient field Calculate the next time step System attenuation state ; S444, System decay state calculated at all time points The connections form the dynamic evolution trajectory of the pores.

8. The method for fault diagnosis of aerospace equipment based on dynamic coupling gradient analysis according to claim 1, characterized in that: Step S5 specifically includes the following sub-steps: S51. Construct a health status baseline: During the normal operation of aviation equipment, collect real-time data and use principal component analysis to construct the normal operation trajectory of the system and the reference parameter vector and its rate of change vector in a multi-dimensional feature space. S52. High-Risk Region Criterion Learning: Combining historical fault data, using machine learning classifiers or statistical methods, learn and define high-risk region criteria in phase space, i.e., determine the safe radius or distance threshold. and system decay state threshold ; S53. Real-time state projection and distance calculation: Project the standardized feature data extracted in real time onto the constructed phase space, and calculate the Euclidean distance or other metric distance between the real-time state parameter vector and the health state baseline. S54. Multi-condition fusion early warning: Real-time determination of whether the current system state meets any of the following conditions: The distance between the real-time state and the reference trajectory exceeds the safety radius or distance threshold. ; Or the system attenuation status calculated in real time Exceeding the preset decay threshold If any of the conditions are met, an early warning will be triggered.

9. The method for fault diagnosis of aerospace equipment based on dynamic coupling gradient analysis according to claim 1, characterized in that: The method also includes step S6: Based on the early warning information, combined with the fault mode library and expert system, perform fault root cause localization and fault propagation path analysis, and output corresponding maintenance suggestions or early warning decisions to form a diagnostic closed loop, which specifically includes the following sub-steps: S61. Fault mode library matching: Match the feature patterns corresponding to the identified failure critical points with the pre-established fault mode library to preliminarily determine the fault type. S62. Root Cause Analysis: By combining key influencing factors and failure propagation paths, trace the root cause of the failure. S63. Maintenance Decision and Early Warning Strategy: Based on the fault type, severity, and remaining life prediction, generate detailed maintenance recommendations and early warning strategies; S64. Knowledge Base Update: Update the system knowledge base with new fault modes, parameter calibration results and warning thresholds to optimize the performance of the diagnostic model in real time.

10. A multi-parameter coupled fault diagnosis system for aviation equipment, used in the aviation equipment fault diagnosis method based on dynamic coupled gradient analysis as described in any one of claims 1-9, characterized in that: It includes a data acquisition module, a parameter intelligent calibration module, a coupled gradient field construction module, a pore dynamics calculation module, and an early warning module; The data acquisition module uses sensors deployed on aviation equipment to collect raw data reflecting the operational status of the aviation equipment in real time; The parameter intelligent calibration module preprocesses the collected raw data, eliminates data noise and redundant information, and performs feature enhancement to obtain standardized feature data. The Coupled Gradient Field Construction Module is used to construct a Coupled Gradient Field Model to quantify the dynamic coupling strength and interaction relationship between key parameters inside aerospace equipment. The Coupled Gradient Field Model can capture the nonlinear interaction effect between key parameters and characterize the potential propagation path and core influencing factors of faults. The hole dynamics calculation module is used to establish hole dynamics equations that integrate the Swiss cheese model theory based on the coupled gradient field model, describe the continuous evolution process of the state of aerospace equipment systems and the dynamic mechanism of fault propagation, and obtain the hole dynamics evolution trajectory based on the hole dynamics equations. Based on the obtained hole dynamics evolution trajectory, the early warning module uses a phase space early warning mechanism to identify the failure critical point of the aviation equipment and trigger an early warning.

Images