Method and system for analyzing and early warning emerging faults of aviation system
By constructing the SCM-DBN fusion model and introducing CAF, the analysis challenge of emergent failures in aviation systems was solved, enabling accurate diagnosis and proactive prediction of aviation system failures, thereby improving the safety and reliability of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-22
- Publication Date
- 2026-04-03
AI Technical Summary
Existing technologies are insufficient to effectively capture and handle emergent failures in aviation systems. They lack a unified framework that integrates structured defense mechanisms with dynamic causal reasoning, have inadequate dynamic evolution modeling, cannot accurately capture nonlinear coupling effects, and lack prediction of risk transition critical points, resulting in limited prediction accuracy and passive responses.
A structured emergent model fused with SCM-DBN is constructed, and a non-additive coupling amplification factor (CAF) is introduced. Through bidirectional reasoning, deep diagnosis and risk transition critical point prediction are achieved. Combined with dynamic temporal evolution modeling of key performance parameters and non-additive conditional probability tables, accurate analysis of emergent failures in aviation systems is realized.
It enables comprehensive and in-depth identification of emergent failures in aviation systems, improves the accuracy and timeliness of failure prediction, and can proactively provide early warnings and optimize intervention measures to enhance system safety and reliability.
Smart Images
Figure CN121786685A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aviation system fault diagnosis and prediction, and specifically to a method and system for analyzing and warning of emergent faults in aviation systems. Background Technology
[0002] With the rapid development of aviation technology, modern aviation systems are becoming increasingly complex, and the interactions between their internal components, subsystems, and external environment are becoming increasingly close. Traditional fault analysis methods are usually based on single fault modes or pre-defined fault trees, making it difficult to effectively capture and handle the "emergent faults" that are prevalent in aviation systems. Emergent faults refer to the phenomenon where normal or minor anomalies in a single component or subsystem, under specific conditions, lead to a sharp decline in overall system performance or even catastrophic failure through complex nonlinear interactions. These faults often exhibit unpredictability, suddenness, and high destructiveness, posing a severe challenge to aviation safety.
[0003] Specifically, existing technologies have the following limitations when dealing with emergent faults: Insufficient structured models: There is a lack of a unified framework that can effectively integrate system structure defense mechanisms and dynamic causal reasoning, making it difficult to fully describe the propagation path of faults and the interactions between levels.
[0004] Insufficient consideration of dynamic evolution: There is a lack of detailed probabilistic modeling of the dynamic time-series evolution of key performance parameters, especially insufficient consideration of the comprehensive impact of factors such as aging, maintenance, and environmental disturbances, which leads to limited prediction accuracy.
[0005] Nonlinear coupling is difficult to handle: When constructing the conditional probability table (CPT) in traditional Bayesian networks, additive or multiplicative assumptions are usually used, which makes it difficult to accurately capture the nonlinear coupling effects under the synergistic effect of multiple factors, such as superlinear enhancement or sublinear suppression effects.
[0006] Lack of risk transition critical point prediction: Existing methods often focus on fault diagnosis or simple risk prediction, but lack a clear definition and effective prediction mechanism for the risk transition critical point from slow accumulation to rapid outbreak of system risk, resulting in the inability to carry out forward-looking and proactive intervention.
[0007] Therefore, there is an urgent need for an innovative method and system that can overcome the above-mentioned shortcomings, effectively analyze emergent failures in aviation systems, and achieve accurate diagnosis and forward-looking prediction. Summary of the Invention
[0008] To address the shortcomings of the existing technologies, the present invention aims to provide a method and system for analyzing and warning of emergent failures in aviation systems. This method overcomes the limitations of existing emergent failure analysis by constructing a unified framework that integrates a structured defense model and a dynamic causal reasoning mechanism. It introduces a non-additive coupling amplification factor (CAF) to accurately capture the synergistic effects of multiple factors and innovatively proposes a risk transition critical point prediction mechanism. This enables in-depth diagnosis and proactive risk management of emergent failures in aviation systems, allowing for better analysis of these failures.
[0009] To achieve the above objectives, the present invention provides a method for analyzing and warning of emergent failures in aviation systems, comprising the following steps: S1. Constructing a structured emergent model fused with SCM-DBN: The multi-layer defense failure theory of the Swiss cheese model (SCM) of aviation systems is fused with the causal reasoning mechanism of dynamic Bayesian network (DBN) to obtain a structured emergent model fused with SCM-DBN. S2. Probabilistic and Dynamic Temporal Evolution Modeling of Key Performance Parameters: Key performance parameters (KPIs) related to system failures in aviation systems are probabilistically represented and dynamically evolved over time. The constructed state transition matrix reflects the dynamic changes in system state. The performance margin Pt of the key performance parameters decays or deteriorates with the influence of time t, operating environment Et, load Lt, maintenance history Mt, and uncertainty factors Ut. The dynamic evolution modeling of key performance parameters is expressed as follows: ; Where f(·) is the performance degradation / deterioration function; Δt is the time step; These are the key performance parameters at time t. These are the key performance parameters at time t+1; S3. Generate a non-additive conditional probability table (CPT) for the fusion coupling amplification factor (CAF): To address the nonlinear coupling effect under the synergistic effect of multiple factors in aerospace systems, a non-additive conditional probability table (CPT) for the fusion coupling amplification factor (CAF) is constructed. The CAF is used to quantitatively characterize the parent nodes of multiple fault factors at different levels. When both are in an abnormal state, the superlinear or sublinear collaborative effect on the failure probability of downstream child node S is expressed as follows: ; in, This represents the failure probability of the downstream child node S. Indicates the parent node; F i This represents a child node under independent action; S is the failure probability, and CAF is the coupling amplification factor. S4. Deep Fault Diagnosis and Risk Transition Critical Point Prediction Based on Bidirectional Reasoning: Based on the constructed SCM-DBN fusion structured emergent model, combined with key performance parameters and fusion coupling amplification factor, bidirectional reasoning is used to diagnose and predict faults in aviation systems. Specifically, this includes: using backward reasoning to trace the causes of system anomalies at multiple levels and locate the fault causes; and using forward reasoning to dynamically predict the future risk evolution of the system, predict risk transition critical points, and provide fault warnings.
[0010] Preferably, step S4 specifically includes the following sub-steps: S41. Backward diagnosis: When a system malfunctions or an anomaly is detected, the malfunction or anomaly data is input as evidence nodes into the constructed SCM-DBN fusion structured emergent model. Using the backward inference algorithm of dynamic Bayesian network, the posterior probability of the potential cause is calculated in reverse under the given evidence. S42. Forward prediction: Based on the current state of the system, combined with the current values of each key performance parameter (KPIs) obtained from step S2 and the predicted evolution trajectory, the forward inference algorithm of the dynamic Bayesian network is used to infer the probability distribution of the system state step by step.
[0011] Preferably, in step S42, at each future time step t+k, the structured emergent model fused with SCM-DBN performs the following operations: S421. Predict the performance margin Pt at time t+k based on the performance decay function f(·). +k ; S422. Based on the predicted parent node state, the non-additive conditional probability table CPT is dynamically calculated using the coupling amplification factor CAF. S423. Calculate the failure probability of a top-level failure event in an aviation system at a future time t+k. The evolution curve of future failure probability is obtained through continuous calculation. By analyzing the changing trend of the future failure probability evolution curve, the risk transition critical point is predicted, and a fault warning is triggered based on the risk transition critical point.
[0012] Preferably, the specific process of predicting the risk transition critical point in step S423 is as follows: S4231. Using the dynamic evolution model of key performance parameters established in step S2, based on the current state of the aviation system and the preset future operating conditions, iteratively calculate forward to predict the performance evolution trajectory P(t) of key performance parameters KPIs in the future period. S4232. The predicted performance evolution trajectory P(t) is used as the key dynamic input for the forward inference in step S42, and substituted into the structured emergent model fused with SCM-DBN to obtain the future failure probability evolution curve of the top-level failure event of the aviation system. ; S4233, for the two predicted curves P(t) and By performing synchronous analysis, the first time point in the future that simultaneously meets the following two conditions is defined as the risk transition critical point T_critical: Find the smallest Make: , where t satisfies and In the formula, At the current point in time, As a safety margin threshold, for Threshold.
[0013] Preferably, in step S41, the process of reverse calculation of the posterior probability of potential causes is carried out layer by layer upward along the defense hierarchy defined by the SCM until the probability distribution of all potential root causes is calculated, the node path with the highest posterior probability is identified as the fault propagation chain, and the node located at the highest level of the fault propagation chain is the fault cause.
[0014] Preferably, in step S1, the different levels of defense in SCM and their potential vulnerabilities are mapped to different levels of nodes in DBN. The defense failure mechanism of each defense level of SCM is transformed into the parent node in DBN, the system vulnerability constitutes the child node, and the concept of vulnerabilities is visualized as the fault precursor node or intermediate node.
[0015] Preferably, the state transition matrix in step S2 is dynamically updated based on actual operating data, aging patterns, maintenance strategies, and environmental disturbance models.
[0016] Preferably, the value of CAF in step S3 is greater than 1, less than 1, or equal to 1.
[0017] Secondly, the present invention also provides an emergent failure analysis system for aviation systems, including: a structured emergent model construction module, a performance parameter dynamic evolution modeling module, a non-additive CPT construction and application module, and a bidirectional reasoning and early warning module; The structured emergent model building module is used to build a structured emergent model that integrates SCM-DBN; the performance parameter dynamic evolution modeling module is used to complete the probabilistic and dynamic temporal evolution modeling of key performance parameters. The non-additive CPT construction and application module is used to generate a non-additive conditional probability table (CPT) for application-coupled amplification factor (CAF): The bidirectional reasoning and early warning module is used for in-depth fault diagnosis and risk transition critical point prediction based on bidirectional reasoning.
[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention innovatively integrates the structural causal model (SCM) with the dynamic Bayesian network (DBN) to construct a dual-dimensional analysis framework of structure-dynamic causality, which comprehensively captures the essential characteristics of emergent failures in aviation systems and completely breaks through the technical limitation that a single model can only cover structural or dynamic characteristics in a one-sided manner, thus realizing the all-round and in-depth identification of complex failure modes in aviation systems.
[0019] (2) The method of the present invention introduces a dynamic time-series evolution modeling method for performance parameters, systematically incorporates multiple key influencing factors such as aging loss, maintenance intervention, and environmental fluctuations, establishes a dynamic model of fault development that fits the actual working conditions, significantly improves the accuracy and timeliness of the prediction of the fault evolution process, and provides reliable data support for subsequent intervention decisions.
[0020] (3) The method of the present invention achieves accurate quantification and characterization of superlinear, sublinear and threshold triggering effects under the synergistic effect of multiple factors through the original coupling amplification factor CAF and non-additive CPT, effectively breaking the technical bottleneck of traditional Bayesian networks in handling complex nonlinear interactions, and can greatly improve the ability to analyze and predict multi-factor coupled faults.
[0021] (4) This invention innovatively defines and predicts the risk transition critical point, which can detect the evolution trend of a sharp increase in risk in advance, actively issue early warning signals and match and optimize intervention measures, successfully promote the transformation of aviation system operation and maintenance from passive response maintenance to predictive and proactive intervention, significantly enhance the safety and reliability of system operation, and reduce the losses caused by failure. Attached Figure Description
[0022] Figure 1 This is a schematic diagram of the process for analyzing and warning of emergent faults in aviation systems according to the present invention; Figure 2 This is a flowchart illustrating the emergent fault analysis and early warning method for aviation systems according to the present invention. Figure 3 This is an SCM-DBN model skeleton diagram of a pitch rate sensor failure of a certain type of aircraft in an embodiment of the present invention; Figure 4 This is a simulation curve of the dynamic evolution of key performance parameters and the evolution of system risk in an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the prediction of the risk transition critical point of nonlinear coupling effect in an embodiment of the present invention; Figure 6This is a schematic block diagram of the system of the present invention. Detailed Implementation
[0023] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.
[0024] This invention provides a method for analyzing and warning of emergent failures in aviation systems, such as... Figure 1 and Figure 2 As shown, it includes the following steps: S1. Constructing a Structured Emergent Model Integrating SCM and DBN: The Swiss Cheese Model (SCM) of aviation systems, based on its multi-layered defense failure theory, is integrated with the causal reasoning mechanism of Dynamic Bayesian Network (DBN) to obtain the SCM-DBN fused structured emergent model. In step S1, the different levels of defense in the SCM and their potential vulnerabilities are mapped to different levels of nodes in the DBN. The defense failure mechanisms of each defense level in the SCM are transformed into parent nodes in the DBN, system vulnerabilities constitute child nodes, and the concept of vulnerabilities is visualized as fault precursor nodes or intermediate nodes.
[0025] S2. Probabilistic and Dynamic Temporal Evolution Modeling of Key Performance Parameters: Key performance parameters (KPIs) related to system failures in aviation systems are probabilistically represented and dynamically evolved over time. The constructed state transition matrix reflects the dynamic changes in system state. The performance margin Pt of the key performance parameters decays or deteriorates with the influence of time t, operating environment Et, load Lt, maintenance history Mt, and uncertainty factors Ut. The dynamic evolution modeling of key performance parameters is expressed as follows: ; Where f(·) is the performance degradation / deterioration function; Δt is the time step; These are the key performance parameters at time t. These are the key performance parameters at time t+1. In this embodiment, the state transition matrix in step S2 is dynamically updated based on actual operating data, aging patterns, maintenance strategies, and environmental disturbance models.
[0026] S3. Generate a non-additive conditional probability table (CPT) for the fusion coupling amplification factor (CAF): To address the nonlinear coupling effect under the synergistic effect of multiple factors in aerospace systems, a non-additive conditional probability table (CPT) for the fusion coupling amplification factor (CAF) is constructed. The CAF is used to quantitatively characterize the parent nodes of multiple fault factors at different levels. Simultaneously, when in an abnormal state, the superlinear or sublinear synergistic effect on the failure probability of downstream child node S, and the failure probability of downstream child node S.
[0027] ; in, This represents the failure probability of the downstream child node S; Indicates the parent node; F i The CAF represents the child node under independent action; S is the failure probability, and CAF is the coupling amplification factor. The value of CAF can be greater than 1 (superlinear), less than 1 (sublinear), or equal to 1 (linear), which is used to accurately capture cooperative enhancement, threshold triggering, or suppression effects, overcoming the limitations of traditional Bayesian networks in dealing with complex interactions of multiple factors.
[0028] S4. Deep Fault Diagnosis and Risk Transition Critical Point Prediction Based on Bidirectional Reasoning: Based on the constructed SCM-DBN fusion model, combined with key performance parameters and fusion coupling amplification factor, bidirectional reasoning is used to diagnose and predict system faults. Specifically, this includes: using backward reasoning to trace the source of existing system anomalies at multiple levels to locate the cause of the fault; and using forward reasoning to dynamically predict the future risk evolution of the system and identify the upcoming risk transition critical point to trigger an early warning.
[0029] Step S4 specifically includes the following sub-steps: S41. Backward diagnosis: When a system malfunctions or detects anomalies, such as sensor readings exceeding limits or system alarms, these observed fault phenomena or anomaly data are input as Evidence Nodes into the constructed SCM-DBN fusion model. Subsequently, the model utilizes backward inference algorithms of dynamic Bayesian networks, such as Belief Propagation or Markov Chain Monte Carlo (MCMC) methods, to calculate the posterior probability of each upstream parent node (potential cause) in the network, given the evidence. This tracing process proceeds layer by layer upwards along the defense layers defined by SCM, such as the operational layer, maintenance layer, and design layer, until the probability distribution of all potential root causes is calculated. Finally, the system identifies the node path with the highest posterior probability; this path is the most likely fault propagation chain, and the node at the highest level of this chain is identified as the root cause of the fault, thus obtaining the fault cause and achieving deep, multi-level fault diagnosis.
[0030] S42, Forward Prediction and Predictive Reasoning: Based on the current state of the system, i.e., the current probability distribution of each node, and combined with the current values of the key performance parameters (KPIs) obtained from step S2 and their predicted evolution trajectories, the model uses a forward inference algorithm of a dynamic Bayesian network, such as the ForwardAlgorithm, to extrapolate the probability distribution of the system state step by step Δt. At each future time step t+k, the model will: S421. Update parameter status: Predict the performance margin P_t+k at time t+k based on the performance decay function f(·).
[0031] S422. Applying the coupling effect: Based on the predicted parent node state, the non-additive conditional probability table CPT is dynamically calculated using the coupling amplification factor CAF defined in step S3 to accurately reflect the nonlinear coupling effects that may occur in the future.
[0032] S423. Calculate future risks: Based on the above information, calculate the failure probability of the system's top-level failure event at time t+k. .
[0033] A future failure probability evolution curve is obtained through continuous calculation. The system analyzes the changing trend of the curve, especially its second derivative with respect to time as defined in claim 3, to accurately identify the "risk transition critical point", thereby achieving dynamic and forward-looking prediction and early warning of system risks.
[0034] The specific process for identifying the risk transition critical point in step S423 is as follows: This process is a forward-looking, model-driven, dynamic identification process, rather than a simple threshold judgment. The specific implementation is as follows: S4231, Performance Trajectory Prediction: First, utilize the dynamic evolution model of key performance parameters established in step S2 ( Based on the current system state and the preset future operating conditions, iterative calculations are performed to predict the evolution trajectory P(t) of key performance parameters (KPIs) over a future period.
[0035] S4232, Risk Curve Derivation: Then, this predicted performance evolution trajectory P(t) is used as the key dynamic input to the forward inference process in step S42, and substituted into the SCM-DBN fusion model. Based on this, the model will deduce the future failure probability evolution curve of the system's top-level failure event. .
[0036] S4233, Critical Point Identification: Finally, the system analyzes the predicted two curves P(t) and... By performing synchronous analysis, the first time point in the future that simultaneously meets the following two conditions is defined as the risk transition critical point T_critical: (1) Performance margin condition: The predicted key performance parameter value reaches or falls below its preset safety threshold for the first time, i.e., P(t)≤P_threshold.
[0037] (2) Risk acceleration condition: While satisfying condition (1), the acceleration of the rate of change of the predicted failure probability curve P_failure(t) at that point, i.e. its second derivative with respect to time, increases significantly or exceeds the preset threshold θ for the first time.
[0038] Its mathematical expression is: finding the smallest Make: , where t satisfies and When the risk transitions to a critical point... When a risk is predicted, the system automatically generates an early warning message, indicating the key performance parameters and coupling paths leading to the risk transition, in order to recommend specific preventative intervention measures. In the formula, At the current point in time, As a safety margin threshold, for Threshold.
[0039] Secondly, the present invention also provides an emergent failure analysis system for aviation systems, such as... Figure 6 As shown, it includes: a structured emergent model construction module 1, a performance parameter dynamic evolution modeling module 2, a non-additive CPT construction and application module 3, and a bidirectional reasoning and early warning module 4. Each module is used to execute the corresponding steps in the above method.
[0040] The Structured Emergent Model Construction Module 1 is used to construct a structured emergent model fused with SCM-DBN; the Performance Parameter Dynamic Evolution Modeling Module 2 is used to complete the probabilistic and dynamic temporal evolution modeling of key performance parameters. The Non-Additive CPT Construction and Application Module 3 is used to generate a non-additive conditional probability table (CPT) for application and coupling amplification factor CAF. The Bidirectional Reasoning and Early Warning Module 4 is used for deep fault diagnosis and risk transition critical point prediction based on bidirectional reasoning. Specific Implementation This embodiment provides a method for analyzing and warning of emergent failures in aviation systems, the execution flow of which is as follows: Figure 2 As shown, Figure 3 The SCM-DBN model skeleton diagram for aircraft pitch rate sensor failure is shown below. The specific steps are as follows: Step S1: Construct a structured emergent model that integrates SCM and DBN.
[0042] SCM Model Deconstruction: First, conduct an in-depth analysis of the aviation system of interest to identify its multi-layered defense mechanisms. For example, for an aircraft engine system, its defense layers can be divided into: a design and manufacturing defect defense layer, a routine maintenance and inspection defense layer, a pilot operation defense layer, and an automatic control system defense layer. Within each defense layer, identify potential "holes," i.e., points or mechanisms where that defense layer may fail. Examples include design defects, maintenance omissions, human error, or sensor malfunctions.
[0043] DBN Node Mapping: This maps the various defense mechanisms and their "holes" in the SCM to nodes in the DBN. Specifically, the defense levels of the SCM can correspond to different time slices (for the dynamic nature of the DBN) or different parent node groups in the DBN. Defense failure mechanisms such as "maintenance omissions" are modeled as parent nodes. System vulnerabilities such as "bearing fatigue" are treated as child nodes, whose states are affected by their parent nodes. The "holes" concept in the SCM is visualized as fault precursor nodes or intermediate nodes in the DBN.
[0044] Precursor nodes for failure include, but are not limited to, the initiation of structural fatigue cracks, initial wear of components, abnormal software logs, and slight sensor drift. These are early warning signals before the system develops into a complete failure, and their probability and severity can be monitored in real time through data.
[0045] Intermediate nodes: characterize the accumulation of system vulnerabilities in the fault propagation path. For example, stress concentration caused by design flaws may lead to crack propagation after a certain operating time.
[0046] Causal chain construction: Based on the actual failure mechanisms of aviation systems and expert knowledge, directed edges are established between nodes in the DBN to represent causal relationships. For example, "maintenance omission" may lead to "initial component wear," "initial component wear" may further lead to "bearing fatigue," and "bearing fatigue" may ultimately lead to "engine failure." Simultaneously, temporal evolution relationships are introduced, enabling the DBN to describe the development of faults over time.
[0047] For example, Figure 4 A simplified schematic diagram of the SCM-DBN fusion model is shown, in which the layered defenses of SCM, such as human operation or mechanical defense, are mapped to different node layers of DBN, and the “holes” of SCM are visualized as precursory faults or intermediate vulnerable nodes in DBN.
[0048] Step S2: Probabilistic and dynamic time-series evolution modeling of key performance parameters.
[0049] KPI Identification and Quantification: Identify key performance parameters (KPIs) in aviation systems that are closely related to emergent failures. These KPIs include, but are not limited to: Structural fatigue: Accumulated fatigue damage is calculated using data from stress sensors and vibration sensors.
[0050] Component wear level: inferred through oil analysis, ultrasonic testing, or performance degradation curves.
[0051] Software aging indicators include: memory leaks, abnormal CPU utilization, and number of process crashes.
[0052] Sensor drift: Quantified by comparison with redundant sensors or by reference calibration data.
[0053] Performance margin definition and initial assessment: Define a performance margin Pt for each KPI, representing the distance between its current performance and the failure threshold. Perform an initial assessment using historical data and system design specifications.
[0054] Dynamic Update of State Transition Matrix: An innovative state transition matrix is constructed to describe the probability of state changes of KPIs within different time steps. The update of this matrix comprehensively considers the following factors: Actual operational data: obtained from flight data recorders, maintenance logs, and real-time sensor data.
[0055] Aging law function g(t): A component aging model established based on materials science, mechanical engineering principles and historical fault data.
[0056] Maintenance strategy function h: Considers the impact of different maintenance types (such as preventive maintenance, periodic inspection, and overhaul) on performance recovery or delaying degradation.
[0057] Environmental disturbance model k: Considers the impact of external environmental factors such as temperature, humidity, air pressure, and vibration on the performance of KPIs.
[0058] The dynamic update of the state transition matrix can be expressed as: ; Where M is the dynamically updated state transition matrix function.
[0059] Dynamic evolution of performance margin: Based on the above factors, the performance margin Pt is updated in real time. Its dynamic evolution relationship is as follows: ; Where f(·) is the performance degradation / deterioration function, which describes the amount of performance degradation per unit time under the influence of the current environment Et, load Lt, maintenance history Mt and uncertainty factor Ut, and Δt is the time step.
[0060] Step S3: Design and apply the non-additive conditional probability table CPT of the fusion coupling amplification factor CAF.
[0061] CAF Definition and Calibration: A Coupling Amplification Factor (CAF) is introduced to address the multi-factor synergistic effects in aviation systems. When multiple failure factors have parent nodes... When both are in an abnormal state, CAF quantitatively characterizes their superlinear or sublinear impact on the failure probability of downstream child nodes S.
[0062] The value of CAF is calibrated and dynamically adjusted based on expert knowledge, historical fault data, or simulation models. Machine learning-based regression analysis: Using historical failure data, a regression model is used to analyze the nonlinear correlation between the states of multiple parent nodes and the failure probabilities of child nodes, thereby fitting a CAF function.
[0063] Based on an expert scoring system: Organizational domain experts evaluate and score the synergistic effects of different combinations of parent nodes, quantifying their enhancing or inhibiting effects.
[0064] Based on physical simulation model: High-precision simulation model is used to simulate the system response under the combined effect of multiple factors, obtain the true failure probability, and then deduce the CAF value.
[0065] Non-additive CPT Construction: Traditional Bayesian network CPTs typically assume independent contributions from parent nodes or simple additive effects, failing to capture collaborative effects. The non-additive CPT constructed in this invention incorporates CAF (Contribution-Oriented Effect), making the failure probability of child node S... Represented as: ; in, This represents the parent node. A CAF value greater than 1 indicates synergistic enhancement (superlinearity), less than 1 indicates synergistic inhibition (sublinearity), and equal to 1 indicates a linear relationship. For example, if two slightly worn components operate simultaneously under high temperature and high pressure, their combined failure probability may be much greater than the sum of their individual failure probabilities; in this case, the CAF will be greater than 1.
[0066] Step S4: Deep fault diagnosis and risk transition critical point prediction based on bidirectional reasoning.
[0067] S41. Backward diagnosis: Fault occurrence or anomaly detection: When the aviation system detects faults such as engine shutdown or abnormal signals such as excessive vibration or excessive oil temperature, these observational evidences are input into the SCM-DBN fusion model.
[0068] Multi-level attribution: The model utilizes the backward inference capabilities of DBNs, such as the Viterbi algorithm or particle filtering in HMMs, to perform probabilistic inference and trace back to the possible causes of the failure. First, it identifies the direct cause, such as sensor failure, and then traces further back to deeper root causes, such as maintenance omissions or design flaws. The hierarchical structure of SCM helps guide this attribution process, tracing back from surface-level failures to the highest-level potential root causes, such as tracing from aircraft loss of control to pilot error, and then to insufficient training or flawed user interface design.
[0069] S42. Forward prediction and risk transition critical point identification: Future risk profile prediction: Utilizing the forward inference capabilities of DBN, combined with a dynamic evolution model of the current system state and KPIs, the probability of system failure at different points in the future can be predicted.
[0070] Definition and Identification of Pfailure Risk Transition Critical Point: An innovative definition of the risk transition critical point Tcritical is presented. This point is the abrupt change where the system failure probability Pfailure transitions from a slow increase to a sharp rise. This typically occurs when the system performance margin Pt and the coupling amplification factor CAF reach specific thresholds.
[0071] Mathematically defined as: when the system performance margin That is, when performance is severely degraded and the coupling amplification factor CAF ≥ CAFthreshold (i.e., the synergistic effect of multiple factors is significantly enhanced), the second derivative of the failure probability Pfailure with respect to time t. Significantly increasing or exceeding the preset threshold θ, i.e.: ; Once a critical risk transition point is predicted, the system automatically generates early warning information and recommends specific intervention measures. Figure 5 The diagram illustrates the prediction of the risk transition critical point. When the performance margin decreases and the CAF increases, the rate of increase of the failure probability accelerates sharply.
[0072] Early warning and intervention measures: Once a risk transition threshold is predicted, the system automatically generates a high-level early warning and recommends specific intervention measures based on the diagnostic results and the predicted risk path. These measures include, but are not limited to: Preventive maintenance: Replacing or repairing critical components before failure occurs. Component replacement: Replacing components that have reached the end of their lifespan or are severely worn. Software upgrades: Fixing potential software defects or optimizing algorithms. Operating procedure adjustments: Optimizing flight operations or maintenance procedures based on risk assessment results. Personnel training: Improving personnel skills for identified high-risk areas. In this way, a paradigm shift from "reactive maintenance" to "predictive, proactive intervention" is achieved.
[0073] 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 method for analyzing and warning of emergent failures in aviation systems, characterized in that, It includes the following steps: S1. Constructing a structured emergent model fused with SCM-DBN: The multi-layer defense failure theory of the Swiss cheese model (SCM) of aviation systems is fused with the causal reasoning mechanism of dynamic Bayesian network (DBN) to obtain a structured emergent model fused with SCM-DBN. S2. Probabilistic and Dynamic Temporal Evolution Modeling of Key Performance Parameters: Key performance parameters (KPIs) related to system failures in aviation systems are probabilistically represented and dynamically evolved over time. The constructed state transition matrix reflects the dynamic changes in the aviation system state. The performance margin Pt of the key performance parameters varies with time t and the operating environment E. t Load L t Maintenance history M t and uncertainties U t The dynamic evolution model of key performance parameters, which is affected by factors such as decay or deterioration, can be represented as follows: ； Where f(·) is the performance degradation / deterioration function; Δt is the time step; These are the key performance parameters at time t. These are the key performance parameters at time t+1; S3. Generate a non-additive conditional probability table (CPT) for the fusion coupling amplification factor (CAF): To address the nonlinear coupling effect under the synergistic effect of multiple factors in aerospace systems, a non-additive conditional probability table (CPT) for the fusion coupling amplification factor (CAF) is constructed. The CAF is used to quantitatively characterize the parent nodes of multiple fault factors at different levels. When both are in an abnormal state, the superlinear or sublinear collaborative effect on the failure probability of downstream child node S is expressed as follows: ； in, This represents the failure probability of the downstream child node S. Indicates the parent node; F i This represents a child node under independent action; S is the failure probability, and CAF is the coupling amplification factor. S4. Deep Fault Diagnosis and Risk Transition Critical Point Prediction Based on Bidirectional Reasoning: Based on the constructed SCM-DBN fusion structured emergent model, combined with key performance parameters and fusion coupling amplification factor, bidirectional reasoning is used to diagnose and predict aviation system faults. Specifically, this includes: using backward reasoning to trace the source of existing system anomalies at multiple levels to locate the cause of the fault; and using forward reasoning to dynamically predict the future risk evolution of the system, predict the risk transition critical point, and provide fault early warning.
2. The method for analyzing and warning of emergent failures in aviation systems according to claim 1, characterized in that: Step S4 specifically includes the following sub-steps: S41. Backward diagnosis: When a failure occurs in the aviation system or an anomaly is detected, the failure phenomenon or abnormal data is input as evidence into the constructed SCM-DBN fusion structured emergent model. Using the backward inference algorithm of dynamic Bayesian network, the posterior probability of the potential cause is calculated in reverse under the condition of input evidence. S42. Forward prediction: Based on the current state of the aviation system, and combining the current values of each key performance parameter (KPI) obtained from step S2 with the predicted evolution trajectory, the forward inference algorithm of the dynamic Bayesian network is used to extrapolate the probability distribution of the aviation system state step by step.
3. The method for analyzing and warning of emergent failures in aviation systems according to claim 2, characterized in that: In step S42, at each future time step t+k, the structured emergent model fused with SCM-DBN performs the following operations: S421. Predict the performance margin Pt at time t+k based on the performance decay function f(·). +k ; S422. Based on the predicted parent node state, the non-additive conditional probability table CPT is dynamically calculated using the coupling amplification factor CAF. S423. Calculate the failure probability of a top-level failure event in an aviation system at a future time t+k. The evolution curve of future failure probability is obtained through continuous calculation. By analyzing the changing trend of the future failure probability evolution curve, the risk transition critical point is predicted, and a fault warning is triggered based on the risk transition critical point.
4. The method for analyzing and warning of emergent failures in aviation systems according to claim 3, characterized in that: The specific process of predicting the risk transition critical point in step S423 is as follows: S4231. Using the dynamic evolution model of key performance parameters established in step S2, based on the current state of the aviation system and the preset future operating conditions, iteratively calculate forward to predict the performance evolution trajectory P(t) of key performance parameters KPIs in the future period. S4232. Using the predicted performance evolution trajectory P(t) as the key dynamic input for the forward inference in step S42, and substituting it into the structured emergent model fused with SCM-DBN, we obtain the future failure probability evolution curve of the top-level failure event of the aviation system. ; S4233, for the two predicted curves P(t) and By performing synchronous analysis, the first time point in the future that simultaneously meets the following two conditions is defined as the risk transition critical point T_critical: Find the smallest Make: , where t satisfies and In the formula, At the current point in time, As a safety margin threshold, for Threshold.
5. The method for analyzing and warning of emergent faults in aviation systems according to claim 2, characterized in that: In step S41, the back-calculation of the posterior probability of potential causes proceeds upwards along the defense hierarchy defined by the SCM until the probability distribution of all potential root causes is calculated. The node path with the highest posterior probability is identified as the fault propagation chain, and the node at the highest level of the fault propagation chain is the fault cause.
6. The method for analyzing and warning of emergent failures in aviation systems according to claim 1, characterized in that: In step S1, the different levels of defense in SCM and their potential vulnerabilities are mapped to different levels of nodes in DBN. The defense failure mechanism of each defense level of SCM is transformed into the parent node in DBN, the system vulnerability constitutes the child node, and the concept of vulnerabilities is visualized as fault precursor nodes or intermediate nodes.
7. The method for analyzing and warning of emergent failures in aviation systems according to claim 1, characterized in that: The state transition matrix in step S2 is dynamically updated based on actual operating data, aging patterns, maintenance strategies, and environmental disturbance models.
8. The method for analyzing and warning of emergent failures in aviation systems according to claim 1, characterized in that: In step S3, the value of CAF can be greater than 1, less than 1, or equal to 1.
9. An aviation system emergent fault analysis system for use in the aviation system emergent fault analysis and early warning method according to claim 1, characterized in that: It includes: a structured emergent model construction module, a performance parameter dynamic evolution modeling module, a non-additive CPT construction and application module, and a bidirectional reasoning and early warning module. The structured emergent model building module is used to build a structured emergent model that integrates SCM-DBN; the performance parameter dynamic evolution modeling module is used to complete the probabilistic and dynamic temporal evolution modeling of key performance parameters. The non-additive CPT construction and application module is used to generate a non-additive conditional probability table (CPT) for application-coupled amplification factor (CAF): The bidirectional reasoning and early warning module is used for in-depth fault diagnosis and risk transition critical point prediction based on bidirectional reasoning.