Aircraft hydraulic pipeline system fault prediction method and system based on digital twin

Abstract

The present application relates to the field of data processing technology, and discloses a fault prediction method and system for an aircraft hydraulic pipeline system based on digital twins. The method comprises: obtaining flight parameters through the aircraft data bus for stage identification to obtain four flight stage identifiers; mapping and processing the load spectrum according to the identifiers to construct a staged digital twin pipeline model; weighting the multi-sensor data according to the identifiers, extracting the fault feature vector according to the twin model; and obtaining the fault prediction result through staged BP neural network prediction processing. The present application solves the problem that the existing digital twin model cannot adapt to the load differences of the entire flight envelope by using fixed parameters, resulting in a lack of flight stage adaptability in the weight distribution of sensor data, which in turn makes the fault prediction algorithm unable to accurately identify the differentiated fault characteristics of different flight stages.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a fault prediction method and system for an aircraft hydraulic pipeline system based on digital twins. Background Art

[0002] Existing fault prediction methods for aircraft hydraulic pipeline systems are primarily based on traditional prognostic health management (PHM) technology. These methods collect data by installing pressure, temperature, and flow sensors at key points in the hydraulic pipelines, and then utilize a single machine learning model to diagnose and predict hydraulic pipeline faults. Digital twin technology, as an emerging intelligent maintenance tool, has been widely adopted in the industrial sector. By constructing virtual models that map to physical entities, it enables real-time data interaction between physical and virtual spaces, providing technical support for online status monitoring, fault diagnosis, and predictive maintenance of equipment. Traditional digital twin hydraulic pipeline fault prediction methods typically establish a unified digital twin model based on ground test data or a single operating condition, using a virtual model with fixed parameters for fault analysis.

[0003] However, the load spectrum and operating conditions experienced by the hydraulic piping system during different aircraft flight phases (takeoff, cruise, maneuvering, and landing) vary significantly. During takeoff, the system must withstand transient high-pressure shocks caused by landing gear retraction and extension. During cruise, it faces changes in hydraulic oil viscosity in the high-altitude, low-temperature environment. During maneuvering, it experiences stress concentrations due to high overloads. Finally, during landing, the system undergoes frequent hydraulic system actuation. Existing single digital twin models are unable to adapt to these diverse load characteristics. Virtual models with fixed physical parameters struggle to accurately reflect the true operating conditions of hydraulic piping during different flight phases, resulting in insufficient fault prediction accuracy. Furthermore, existing methods lack specialized fault feature extraction and prediction algorithms for different flight phases, making them unable to effectively identify the unique failure modes of each phase.

[0004] Further analysis revealed deeper technical problems with existing technologies: the weight allocation strategy for sensor data lacks adaptability to flight phases and cannot be dynamically adjusted according to the monitoring focus of different phases, resulting in the concealment of key fault characteristics; the physical parameter configuration of the digital twin model lacks an adaptive switching mechanism, and the model parameters cannot be updated in real time according to changes in flight status; the fault prediction algorithm lacks a dedicated phased design, and a single BP neural network model finds it difficult to accurately learn and identify differentiated fault characteristics in different flight phases. In particular, when dealing with multi-field, multi-scale, and strongly nonlinear pipeline system faults, the prediction accuracy and real-time performance cannot meet the fault prediction requirements of the entire flight envelope. Summary of the Invention

[0005] The present application provides a digital twin-based aircraft hydraulic pipeline system fault prediction method and system, which is used to solve the problem that the existing digital twin model uses fixed parameters and cannot adapt to the load differences in the entire flight envelope, resulting in the lack of flight stage adaptability of sensor data weight distribution, and thus making the fault prediction algorithm unable to accurately identify the differentiated fault characteristics of different flight stages.

[0006] In the first aspect, the present application provides a digital twin-based aircraft hydraulic pipeline system fault prediction method, which includes: obtaining flight parameters through the aircraft data bus, performing stage identification processing on the flight envelope according to the flight parameters, and obtaining four flight stage identifiers of takeoff, cruise, maneuvering and landing; performing stage-by-stage mapping processing on the hydraulic pipeline load spectrum according to the flight stage identifier to obtain a differentiated physical parameter set for each stage, and constructing a staged digital twin pipeline model through the differentiated physical parameter set for each stage; performing weight distribution processing on multi-sensor data according to the flight stage identifier to obtain differentiated monitoring data, and performing fault feature extraction processing on the differentiated monitoring data according to the staged digital twin pipeline model to obtain a fault feature vector; and performing prediction processing on the fault feature vector through a staged BP neural network to obtain a fault prediction result.

[0007] In a second aspect, the present application provides an aircraft hydraulic pipeline system fault prediction system based on digital twins, the aircraft hydraulic pipeline system fault prediction system based on digital twins comprising:

[0008] The identification module is used to obtain flight parameters through the aircraft data bus, perform phase identification processing on the flight envelope based on the flight parameters, and obtain the four flight phase identifiers of takeoff, cruise, maneuvering and landing;

[0009] a mapping module, configured to perform phased mapping processing on the hydraulic pipeline load spectrum according to the flight phase identifier to obtain a differentiated physical parameter set for each phase, and construct a phased digital twin pipeline model using the differentiated physical parameter set for each phase;

[0010] an allocation module, configured to perform weighted allocation processing on the multi-sensor data according to the flight phase identifier to obtain differentiated monitoring data, and perform fault feature extraction processing on the differentiated monitoring data according to the staged digital twin pipeline model to obtain a fault feature vector;

[0011] The prediction module is used to perform prediction processing on the fault feature vector through a staged BP neural network to obtain a fault prediction result.

[0012] In a third aspect, a digital twin-based aircraft hydraulic pipeline system fault prediction device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory to enable the digital twin-based aircraft hydraulic pipeline system fault prediction device to execute the above-mentioned digital twin-based aircraft hydraulic pipeline system fault prediction method.

[0013] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, which, when executed on a computer, enables the computer to execute the above-mentioned digital twin-based aircraft hydraulic pipeline system fault prediction method.

[0014] In the technical solution provided in this application, by establishing a staged digital twin hydraulic pipeline fault prediction system for the entire flight envelope, the key technical problem that the existing technology cannot adapt to the load differences in different flight phases is effectively solved. The technical feature of obtaining flight parameters through the aircraft data bus and performing stage identification processing on the flight envelope according to the flight parameters to obtain four flight stage identifiers has broken through the limitation of traditional fault prediction methods that ignore flight state differences, and established an accurate identification mechanism for the four typical flight stages of take-off, cruising, maneuvering and landing, laying the foundation for subsequent differential analysis. The technical feature of performing staged mapping processing on the hydraulic pipeline load spectrum according to the flight stage identifier to obtain a set of differentiated physical parameters for each stage and construct a staged digital twin pipeline model has innovatively established a dynamic mapping relationship between the load spectrum and the flight stage, so that the digital twin model can adaptively adjust the physical parameters according to the actual flight state, significantly improving the synchronization accuracy of the virtual model and the physical entity, and greatly improving the prediction accuracy of the traditional fixed parameter model in different flight stages. The technical features of differentiated monitoring data are derived by weighting multi-sensor data according to flight phase identifiers. This addresses the problem of key fault features being masked by fixed sensor data weights in traditional methods. A dynamic weighting mechanism highlights monitoring priorities for each flight phase, focusing on pressure and flow changes during takeoff, temperature changes during cruise, and vibration and stress during maneuvering, enabling targeted fault feature extraction. Fault feature extraction of differentiated monitoring data based on a staged digital twin pipeline model yields technical features of fault feature vectors. A fault feature extraction mechanism based on the digital twin model is established, accurately identifying abnormal patterns through virtual-to-real comparison analysis. Compared to traditional feature extraction methods based on historical data, this method offers greater real-time and accuracy. The technical feature of predicting fault feature vectors through a phased BP neural network to obtain fault prediction results reflects the important contribution of artificial intelligence algorithms in specific application fields. The design of the phased BP neural network fully considers the differences in fault characteristics of aircraft hydraulic pipeline systems in different flight phases. Each dedicated network is specially trained for the physical characteristics and failure modes of a specific phase. The network structure and parameter configuration are optimized. The takeoff-dedicated network focuses on learning transient pressure shock patterns, the cruise-dedicated network focuses on learning temperature gradient change patterns, and the maneuvering-dedicated network focuses on learning stress concentration and vibration anomaly patterns. Compared with the traditional single BP network model, this phased artificial intelligence algorithm design has significant improvements in key performance indicators such as fault identification accuracy, false alarm rate control, and prediction response time. In particular, when dealing with the complex mechanisms, multi-source excitation, multi-field and multi-scale characteristics of hydraulic pipeline systems, the phased BP neural network is able to better learn and express complex fault evolution patterns, and the algorithm's generalization ability and robustness are significantly enhanced.Overall, the present invention achieves a technological leap from passive maintenance to predictive maintenance by constructing a phased and adaptive digital twin fault prediction system, providing technical support for the full life cycle health management of aircraft hydraulic pipeline systems, effectively reducing maintenance costs and improving flight safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 This is a schematic diagram of an embodiment of a method for predicting faults in an aircraft hydraulic pipeline system based on digital twins in an embodiment of the present application;

[0017] Figure 2 This is a schematic diagram of an embodiment of a digital twin-based aircraft hydraulic pipeline system fault prediction system in an embodiment of the present application;

[0018] Figure 3 It is a structural schematic block diagram of an aircraft hydraulic pipeline system fault prediction device based on digital twin in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The embodiments of the present application provide a method and system for predicting faults in an aircraft hydraulic pipeline system based on digital twins. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0020] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiments of the present application, an embodiment of a method for predicting faults in an aircraft hydraulic pipeline system based on digital twins includes:

[0021] Step S101: Acquire flight parameters through the aircraft data bus, perform phase identification processing on the flight envelope based on the flight parameters, and obtain four flight phase identifiers: takeoff, cruise, maneuvering, and landing;

[0022] Step S102: mapping the hydraulic pipeline load spectrum in stages according to the flight stage identifier to obtain a differentiated physical parameter set for each stage, and constructing a staged digital twin pipeline model based on the differentiated physical parameter set for each stage;

[0023] Step S103: Weighting the multi-sensor data according to the flight phase identifier to obtain differentiated monitoring data, and extracting fault features from the differentiated monitoring data based on the staged digital twin pipeline model to obtain a fault feature vector;

[0024] Step S104: performing prediction processing on the fault feature vector through a phased BP neural network to obtain a fault prediction result.

[0025] It is understandable that the execution subject of this application can be a digital twin-based aircraft hydraulic pipeline system fault prediction system, or a terminal or server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0026] Specifically, the aircraft data bus collects key flight parameters such as altitude, airspeed, angle of attack, overload coefficient, and control surface deflection angle in real time. These parameters are normalized to form a flight state feature vector. The flight phase discrimination matrix is a pre-set threshold judgment table. The flight state feature vector is compared with this matrix. For example, when the flight altitude is less than 500 meters and the airspeed is greater than 80 meters per second, it is determined to be the takeoff phase. When the flight altitude exceeds 3,000 meters and the speed change rate is less than 0.5 meters per second squared, it is determined to be the cruise phase. Based on the discrimination results, the state machine algorithm generates a state transition sequence, which records the changes in the flight phase and ultimately assigns a corresponding flight phase identifier. Secondly, the load spectrum parameter library stores the physical characteristic data of the hydraulic pipelines for each flight phase. The corresponding parameters are retrieved based on the flight phase identifier. During takeoff, the transient pressure peak parameter is extracted to handle the sudden high pressure during landing gear retraction and extension. During cruise, the hydraulic oil viscosity temperature coefficient parameter is extracted to cope with the high-altitude low-temperature environment. During maneuvering, the pipeline stress concentration factor parameter is extracted to cope with high overload conditions. These parameters are aggregated and configured into the pipeline geometry model. For example, the transient peak pressure parameter during takeoff is matched with pipeline wall thickness data, and the corresponding wall thickness configuration data is calculated using the pressure bearing capacity formula. The temperature coefficient parameter during cruise phase is used to compensate for temperature in the pipeline inner diameter. The virtual mapping construction process integrates these physical parameters into the 3D digital model, forming a phased digital twin pipeline model that automatically switches parameters according to the flight phase. The weighting of multi-sensor data is adjusted based on the monitoring priorities of different flight phases. During takeoff, the pressure sensor weight is set to 0.4, and the flow sensor weight is set to 0.4. During cruise, the temperature sensor weight is set to 0.4 and the pressure sensor weight is set to 0.3. During maneuvering, the vibration sensor weight is set to 0.4 and the pressure sensor weight is set to 0.4. A weighted fusion calculation multiplies the data from each sensor by its corresponding weight and sums them, generating differentiated monitoring data that reflects the key monitoring characteristics of the current phase. The fault feature extraction process compares the differentiated monitoring data with the theoretical data of the staged digital twin pipeline model point by point, calculates the data deviation value, and determines it as an abnormal state when the deviation value exceeds the set threshold. Through feature vector quantization processing, a fault feature vector containing pressure anomaly, temperature anomaly, flow anomaly and vibration anomaly features is generated.

[0027] The phased BP neural network selects the corresponding dedicated network model based on the flight phase identifier. For example, a network trained for transient pressure changes is selected for takeoff, and a network trained for temperature gradient changes is selected for cruise. The BP neural network inputs the fault feature vector into the hidden layer through forward propagation. After weight matrix calculation and activation function processing, the output layer generates the probability distribution of various fault types, including the probability of pipeline rupture, seal failure, blockage, loosening, corrosion, and fatigue cracks. The maximum probability value is ultimately selected as the fault prediction result.

[0028] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0029] The flight altitude, airspeed, angle of attack, overload coefficient and control surface deflection angle data are collected through the aircraft data bus, and the flight altitude, airspeed, angle of attack, overload coefficient and control surface deflection angle data are normalized to obtain the flight state feature vector;

[0030] A flight phase discrimination matrix is established based on the flight state feature vector, and a threshold comparison process is performed on the flight state feature vector to obtain the flight phase discrimination result. The discrimination condition for the takeoff phase is low altitude and high speed, and the discrimination condition for the cruise phase is high altitude and stable speed.

[0031] The flight phase identification result is input into the state machine algorithm for logic determination processing to obtain the state transition sequence of the current flight phase, and the corresponding flight phase code is generated according to the state transition sequence;

[0032] Identifier allocation processing is performed based on the flight phase coding to obtain four flight phase identifiers: takeoff, cruise, maneuvering and landing. The takeoff phase identifier is the takeoff identifier, the cruise phase identifier is the cruise identifier, the maneuvering phase identifier is the maneuvering identifier, and the landing phase identifier is the landing identifier.

[0033] Specifically, the aircraft data bus, serving as a high-speed digital communication network within the aircraft, collects key flight parameters necessary for hydraulic line fault prediction in real time. During data collection, the altitude sensor measures the aircraft's vertical height above sea level, the airspeed sensor detects the aircraft's velocity relative to the surrounding air, the angle of attack sensor monitors the angle between the aircraft's longitudinal axis and the direction of the relative airflow, the load factor sensor records the ratio of the aircraft's acceleration to the acceleration due to gravity, and the control surface deflection angle sensor obtains the deflection angles of each control surface. Normalization converts these raw data of varying dimensions to a standard range of zero to one. This process involves subtracting the minimum value from each parameter value and then dividing it by the difference between the maximum and minimum values. This eliminates dimensional differences and numerical ranges between the parameters, resulting in a flight state feature vector. The flight phase discrimination matrix is a pre-established two-dimensional data table, with rows representing different flight phases and columns representing the discrimination thresholds for each flight parameter. The threshold comparison process compares each element in the flight state feature vector with the corresponding threshold in the discriminant matrix. The takeoff phase is determined by the flight altitude being less than a preset low-altitude threshold and the airspeed being greater than a preset high-speed threshold. The cruise phase is determined by the flight altitude being greater than a preset high-altitude threshold and the airspeed rate of change being less than a preset stability threshold. If the flight state feature vector satisfies all the criteria for a particular phase, that phase is marked as the current flight state, forming the flight phase determination result.

[0034] The state machine algorithm is a logic judgment method based on finite state automaton theory that processes the flight phase judgment result as an input signal. The logic judgment process first checks the consistency of the current judgment result with the flight phase at the previous moment. If the same phase is determined at multiple consecutive time points, the state switch is confirmed. If the judgment result changes frequently, the previous state is maintained to avoid misjudgment. The state transition sequence records the changing trajectory of the flight phase, such as the complete process of switching from the takeoff mark to the cruise mark and then to the maneuver mark. The flight phase code generates the corresponding digital code based on the state transition sequence. The takeoff stage is assigned code one, the cruise stage is assigned code two, the maneuvering stage is assigned code three, and the landing stage is assigned code four.

[0035] The identifier assignment process converts digital codes into character identifiers that are easily recognized by the hydraulic pipeline fault prediction system. The takeoff phase identifier is the takeoff identifier, which triggers the hydraulic pipeline system's monitoring mode for transient high pressure during landing gear retraction and extension. The cruise phase identifier is the cruise identifier, which activates the analysis mode for changes in hydraulic oil viscosity in high-altitude, low-temperature environments. The maneuvering phase identifier is the maneuver identifier, which activates the detection mode for pipeline stress concentration under high-g conditions. The landing phase identifier is the landing identifier, which activates the monitoring mode for frequent hydraulic system movements during landing.

[0036] For example, during a normal flight, a civil airliner collects raw data from its data bus: an altitude of 8,500 meters, an airspeed of 250 meters per second, an angle of attack of 3 degrees, an overload factor of 1.02, and a rudder deflection angle of 0 degrees. Normalization is performed by dividing the altitude of 8,500 meters by the maximum altitude of 12,000 meters to obtain 0.708, and dividing the airspeed of 250 meters per second by the maximum airspeed of 300 meters per second to obtain 0.833. Similarly, other parameters are processed to form a flight state eigenvector. The cruise phase discrimination matrix identifies the cruise phase as an altitude greater than 0.5 and an airspeed rate of change less than 0.01. The current flight state eigenvector satisfies these conditions, resulting in a cruise phase discrimination. The state machine algorithm detects that the first three time points were all cruise states, confirms the state transition to cruise, and generates a state transition sequence from takeoff to cruise, corresponding to a flight phase code of 2. The identifier allocation process converts code 2 into a cruise identifier. After receiving the cruise identifier, the hydraulic pipeline fault prediction system automatically switches to the cruise phase monitoring mode, focusing on the impact of the viscosity change of the hydraulic oil in the high-altitude and low-temperature environment on the pipeline pressure distribution.

[0037] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0038] Based on the flight phase identifier, the preset load spectrum parameter library is queried and the hydraulic pipeline load spectrum is extracted and processed in stages. The transient pressure peak parameter of the takeoff phase, the hydraulic oil viscosity temperature coefficient parameter of the cruise phase, the pipeline stress concentration factor parameter of the maneuvering phase, and the flow pulsation frequency parameter of the landing phase are obtained.

[0039] The parameters of the transient pressure peak during takeoff, the hydraulic oil viscosity temperature coefficient during cruise, the pipeline stress concentration coefficient during maneuvering, and the flow pulsation frequency during landing are aggregated to obtain differentiated physical parameter sets for each stage.

[0040] The physical properties of the pipeline geometry model are configured according to the differentiated physical parameter sets of each stage to obtain a staged pipeline physical model. The high dynamic response properties are configured for the takeoff stage, the temperature sensitivity properties are configured for the cruise stage, and the stress analysis properties are configured for the maneuvering stage.

[0041] Based on the staged pipeline physical model, virtual mapping construction is performed to obtain a staged digital twin pipeline model. The staged digital twin pipeline model includes an adaptive switching mechanism for physical parameters corresponding to each flight stage.

[0042] Specifically, the load spectrum parameter library is a pre-built multi-dimensional data storage structure that stores the physical characteristic parameters of the hydraulic pipeline under different working conditions according to the flight phase classification. The query processing based on the flight phase identifier locates the corresponding data record in the parameter library through the identifier matching algorithm. The takeoff identifier triggers access to the takeoff phase data table to extract the transient pressure peak parameter, which reflects the instantaneous high-pressure impact characteristics of the hydraulic pipeline when the landing gear is retracted or extended. The cruise identifier corresponds to the cruise phase data table and extracts the hydraulic oil viscosity temperature coefficient parameter, which describes the change law of the hydraulic oil flow performance in high-altitude and low-temperature environments. The maneuvering identifier triggers access to the maneuvering phase data table to extract the pipeline stress concentration coefficient parameter, which quantifies the stress distribution characteristics of the key nodes of the pipeline under large overload conditions. The landing identifier corresponds to the landing phase data table and extracts the flow pulsation frequency parameter, which reflects the flow fluctuation characteristics caused by the frequent actions of the hydraulic system during landing.

[0043] Parameter aggregation processing structures and organizes the single parameters extracted from different flight phases to form differentiated physical parameter sets for each phase. The processing first establishes a parameter classification index table, classifying the transient pressure peak parameter into the pressure category, the hydraulic oil viscosity temperature coefficient parameter into the temperature category, the pipeline stress concentration coefficient parameter into the stress category, and the flow pulsation frequency parameter into the flow category. Parameter correlation analysis is then performed to examine the coupling relationship between different parameters, such as the correlation between the transient pressure peak and the pipeline stress concentration coefficient, and the mutual influence between the hydraulic oil viscosity temperature coefficient and the flow pulsation frequency. Aggregation processing combines related parameters into a parameter matrix, where each row represents a flight phase and each column represents a type of physical parameter. The matrix elements record the specific parameter values, forming a differentiated physical parameter set covering the entire flight envelope.

[0044] The physical property configuration process assigns properties to the pipeline geometry model based on the differentiated physical parameter sets for each stage. The pipeline geometry model contains the three-dimensional geometric structure information of the pipeline, including geometric parameters such as pipe diameter, wall thickness, length, and bending radius, as well as physical parameters such as material properties and boundary conditions. The high dynamic response property configuration for the takeoff stage writes the transient pressure peak parameter into the pressure bearing property field of the pipeline model and adjusts the pipeline wall thickness parameter to withstand transient high pressure. The temperature sensitive property configuration for the cruise stage writes the hydraulic oil viscosity temperature coefficient parameter into the fluid property field and modifies the pipeline inner diameter parameter to accommodate viscosity changes. The stress analysis property configuration for the maneuvering stage writes the pipeline stress concentration factor parameter into the structural mechanics property field and adjusts the material density parameters of key nodes to analyze stress distribution. The configuration process generates staged pipeline physical models, each of which is specifically optimized for the load characteristics of a specific flight stage.

[0045] The virtual mapping construction process converts the physical model of the phased pipeline into a virtual model in the digital twin environment. The mapping process first performs geometric data conversion, converting the three-dimensional geometric information of the pipeline physical model into a format compatible with the digital twin platform to establish the geometric structure of the virtual pipeline. Next, physical attribute mapping is performed to map the physical attribute parameters configured in each stage to the corresponding attribute fields of the virtual model, establishing an association between the virtual model and the physical parameters. The physical parameter adaptive switching mechanism is the key control logic. This mechanism monitors changes in the flight phase identifier. When the identifier switches, it automatically calls the physical parameter set of the corresponding stage and updates the attribute configuration of the virtual model. The switching mechanism includes a parameter verification module to ensure the continuity and consistency of the parameters during the switching process and avoid model distortion caused by parameter mutations.

[0046] In a specific embodiment, the process of performing the physical attribute configuration processing on the pipeline geometric model according to the differentiated physical parameter sets of each stage may specifically include the following steps:

[0047] Matching and calculating the transient pressure peak parameter during the takeoff phase with the wall thickness data of the pipeline geometric model to obtain the pipeline wall thickness configuration data during the takeoff phase, and updating the wall thickness attribute value of the pipeline geometric model according to the pipeline wall thickness configuration data during the takeoff phase;

[0048] Perform temperature compensation calculation on the inner diameter of the pipeline based on the temperature coefficient parameter of the hydraulic oil viscosity during the cruise phase, obtain the temperature-corrected inner diameter data during the cruise phase, and write the temperature-corrected inner diameter data during the cruise phase into the inner diameter parameter field of the pipeline geometric model;

[0049] Perform stress-weighted calculation on the position coordinates of the pipeline nodes according to the stress concentration coefficient parameters of the pipeline in the maneuvering stage to obtain the stress weight data of the nodes in the maneuvering stage, and adjust the material density parameters of each node in the pipeline geometric model based on the stress weight data of the nodes in the maneuvering stage;

[0050] The flow pulsation frequency parameter of the landing phase and the pipeline length data are used to calculate the flow resistance to obtain the pipeline flow resistance coefficient data of the landing phase. The surface roughness parameters of the pipeline geometric model are modified according to the pipeline flow resistance coefficient data of the landing phase.

[0051] The updated pipeline wall thickness attribute value, inner diameter parameter field, material density parameter and surface roughness parameter are processed through data integration to obtain a staged pipeline physical model.

[0052] Specifically, the initial wall thickness values for each pipe segment are read from the pipeline geometry model. The maximum pressure load each segment must withstand during takeoff is then calculated based on the transient pressure peak parameters. The matching calculation uses a pressure bearing capacity assessment algorithm to compare the transient pressure peak with the material strength limit of each pipe segment. When the transient pressure peak exceeds the bearing capacity of the existing wall thickness, the required wall thickness is calculated. When the transient pressure peak is below the bearing capacity, the wall thickness reduction is calculated, thereby obtaining the takeoff pipeline wall thickness configuration data. This configuration data contains the optimized wall thickness values for each pipe segment. These values are then written into the wall thickness attribute field of the pipeline geometry model, replacing the original wall thickness values and completing the wall thickness attribute value update operation.

[0053] The temperature compensation calculation process corrects and adjusts the inner diameter of the pipeline based on the hydraulic oil viscosity temperature coefficient parameter during the cruise phase. The calculation process first obtains the ambient temperature data during the cruise phase, and then calculates the actual viscosity value of the hydraulic oil at that temperature using the hydraulic oil viscosity temperature coefficient parameter. The temperature compensation calculation uses the pipeline flow theory in fluid mechanics. When the viscosity of the hydraulic oil increases due to a decrease in temperature, the inner diameter of the pipeline needs to be appropriately increased to maintain the same flow rate. When the viscosity of the hydraulic oil decreases due to an increase in temperature, the inner diameter of the pipeline can be appropriately reduced to save material. The calculation process converts the viscosity change into an inner diameter adjustment amount through the Reynolds number correction formula to obtain the temperature-corrected inner diameter data for the cruise phase. This data contains the optimal inner diameter values of each pipe section during the cruise phase. These values are then written into the inner diameter parameter field of the pipeline geometry model, overwriting the original inner diameter parameters to complete the temperature compensation adjustment of the inner diameter size.

[0054] The stress-weighted calculation process performs stress analysis and weight assignment on the pipeline node locations based on the pipeline stress concentration factor parameters for the maneuvering phase. This process first identifies all key nodes in the pipeline geometric model, including elbows, tees, reducers, and other locations prone to stress concentration. The stress-weighted calculation applies the pipeline stress concentration factor parameters for the maneuvering phase to the position coordinates of each node. The stress distribution state of each node under maneuvering flight overload conditions is calculated using the finite element analysis method. Different weight coefficients are assigned to each node based on the stress magnitude. The greater the stress, the higher the weight coefficient of the node. Based on the calculated maneuvering phase node stress weight data, the material density parameters of the pipe sections corresponding to each node in the pipeline geometric model are adjusted. The pipe sections corresponding to high-weight nodes use higher-density materials to withstand greater stress, while the pipe sections corresponding to low-weight nodes use lower-density materials to reduce overall weight.

[0055] The flow resistance calculation process combines the landing phase flow pulsation frequency parameters with the pipeline length data to perform flow resistance analysis. The calculation process first extracts the length data of each pipe segment from the pipeline geometry model, and then analyzes the flow characteristics of the hydraulic oil in the pipeline based on the landing phase flow pulsation frequency parameters. The flow resistance calculation uses a variation of the Darcy-Weisbach formula to convert the flow pulsation frequency into an equivalent Reynolds number change. The flow resistance coefficient of each pipe segment is calculated based on the relationship between pipeline length and flow resistance. When the flow pulsation frequency is high, it indicates that the hydraulic system is frequently operating and the flow resistance of the pipeline needs to be reduced. When the flow pulsation frequency is low, the flow resistance can be appropriately increased to improve system stability. Based on the calculated landing phase pipeline flow resistance coefficient data, the surface roughness parameters of the pipeline geometry model are modified, and the flow resistance characteristics are changed by adjusting the roughness of the pipeline inner surface.

[0056] The data integration process integrates the parameters updated in the above steps to form a complete phased pipeline physical model. The integration process first checks the compatibility between the updated pipeline wall thickness attribute values, inner diameter parameter fields, material density parameters and surface roughness parameters to ensure the physical rationality and consistency of each parameter. Data integration adopts the parameter priority sorting method. When there is a conflict between different parameters, the final value of the parameter is determined in the order of safety, economy and performance optimization. The integration process also includes the verification of the correlation between parameters. For example, when the wall thickness increases, the inner diameter size needs to be adjusted accordingly, when the material density changes, the stress distribution needs to be recalculated, and when the surface roughness changes, the flow resistance coefficient needs to be corrected.

[0057] For example, during an intercontinental flight, the hydraulic piping system of a wide-body passenger aircraft needs to adapt to the changes from takeoff to cruise. Transient pressure peak parameters during takeoff indicate that the pressure surge experienced by the piping during main landing gear retraction and extension far exceeds the normal operating pressure during cruise. A matching calculation process compares this peak pressure with the existing wall thickness of the landing gear hydraulic piping and finds that the existing wall thickness is insufficient to withstand the transient surge. The calculation then determines that the wall thickness of the main landing gear piping needs to be increased from its original thickness to a thicker one. Once the aircraft enters cruise, the low temperatures at cruising altitude cause the viscosity of the hydraulic oil to increase significantly. The temperature compensation calculation process calculates the specific increase in viscosity based on the hydraulic oil viscosity temperature coefficient parameter. To maintain normal flow, the inner diameter of the main hydraulic piping needs to be increased from its original size. During maneuvering, the lateral overloads generated by the aircraft's turns place additional stress on the hydraulic piping's elbows. The stress-weighted calculation process identifies the highest stress concentration at the elbow node and assigns the highest weight coefficient to this node. Accordingly, the piping material density at the elbow is adjusted to a high-strength alloy. During landing, the hydraulic system frequently operates components like flaps and landing gear, resulting in significantly higher flow pulsation frequency than during cruise. Based on this change, the flow resistance calculation process determines the need to reduce flow resistance in the pipelines, thereby minimizing flow losses by reducing the roughness of the pipeline's internal surface. Data integration integrates all these parameter adjustment results to form a phased pipeline physical model optimized for the entire flight phase.

[0058] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0059] The pressure sensor data, temperature sensor data, flow sensor data, and vibration sensor data of the key nodes of the hydraulic pipeline are processed by weight coefficients according to the flight phase identifier to obtain a stage sensor weight coefficient table, where the pressure weight coefficient for the takeoff phase is 0.4, and the temperature weight coefficient for the cruise phase is 0.4;

[0060] Based on the stage-by-stage sensor weight coefficient table, the pressure sensor data, temperature sensor data, flow sensor data, and vibration sensor data are weighted and fused to obtain differentiated monitoring data. The differentiated monitoring data includes the key monitoring parameter values for each flight stage.

[0061] The differentiated monitoring data is input into the phased digital twin pipeline model for data comparison and analysis to obtain the deviation value between the monitoring data and the model data. The abnormality degree value of each monitoring point is calculated based on the deviation value;

[0062] The abnormality degree value is subjected to feature vector quantization processing to obtain a fault feature vector, which includes a numerical combination of pressure abnormality characteristics, temperature abnormality characteristics, flow abnormality characteristics and vibration abnormality characteristics.

[0063] Specifically, a correspondence table between sensor types and flight phases was established, and weights were assigned to pressure sensors, temperature sensors, flow sensors, and vibration sensors according to the monitoring priorities of different flight phases. During takeoff, due to the frequent retraction and extension of the landing gear and the transient high-pressure impact on the hydraulic system, the pressure sensor was assigned the highest weight coefficient of 0.4. The flow sensor weight coefficient was also set to 0.4 to monitor the rapid flow of hydraulic oil. The temperature sensor weight coefficient was set to 0.1, and the vibration sensor weight coefficient was set to 0.1. During the cruise phase, due to the significant impact of the high-altitude and low-temperature environment on the viscosity of the hydraulic oil, the temperature sensor was assigned the highest weight coefficient of 0.4. The pressure sensor weight coefficient was set to 0.3 to monitor steady-state pressure. The flow sensor weight coefficient was set to 0.2, and the vibration sensor weight coefficient was set to 0.1. During the maneuvering phase, due to the high overload conditions that intensify pipeline vibration, the vibration sensor weight coefficient was set to 0.4, the pressure sensor weight coefficient was set to 0.4, and the temperature and flow sensors weight coefficients were each 0.1. During the landing phase, the hydraulic system operates frequently. The flow sensor's weight coefficient is 0.4, the pressure sensor's weight coefficient is 0.3, and the temperature and vibration sensors' weight coefficients are each 0.15, forming a complete table of sensor weight coefficients for each phase. A weighted fusion calculation process performs a numerical fusion operation on the sensor data based on this table. This calculation process first reads the real-time data values of each sensor and then queries the corresponding weight coefficient based on the current flight phase identifier. The fusion calculation uses a linear weighted summation method, multiplying each sensor's data value by its corresponding weight coefficient and then summing all weighted values to produce the fusion result. For example, during takeoff, the pressure sensor data is multiplied by a weight coefficient of 0.4, the flow sensor data is multiplied by a weight coefficient of 0.4, the temperature sensor data is multiplied by a weight coefficient of 0.1, and the vibration sensor data is multiplied by a weight coefficient of 0.1. The sum of these four weighted values yields differentiated monitoring data for that monitoring point. This data reflects the comprehensive operating status of the hydraulic pipeline during the current flight phase, highlighting the most important changes in the physical parameters during that phase, and forms a differentiated monitoring data set containing the values of the key monitored parameters for each flight phase.

[0064] The data comparison and analysis process compares the differentiated monitoring data point-by-point with the theoretical data from the phased digital twin pipeline model. This process first extracts a theoretical data sequence corresponding to the current flight phase from the digital twin pipeline model. This sequence contains the expected numerical range for each monitoring point under normal operating conditions. The comparative analysis calculates the numerical difference between the actual monitoring data and the theoretical data to obtain a deviation value for each monitoring point. A positive deviation value indicates that the actual data exceeds the theoretical data, while a negative deviation value indicates that the actual data falls below the theoretical data. The degree of abnormality is quantitatively assessed based on the magnitude and duration of the deviation value. When a deviation value exceeds the preset normal fluctuation range, an abnormality severity value is calculated based on the degree of the excess. A larger deviation value indicates a higher abnormality severity value, and a longer duration indicates a greater increase in the abnormality severity value. The abnormality severity value is standardized on a scale of zero to one, with zero indicating a completely normal state, one indicating a severe abnormality, and intermediate values indicating varying degrees of abnormality.

[0065] Feature vectorization categorizes and organizes the abnormality values of each monitoring point by sensor type, forming a multi-dimensional fault feature vector. This process first divides all monitoring points into four groups based on sensor type: pressure, temperature, flow, and vibration. Then, statistical characteristics of the abnormality values of each monitoring point within each group are calculated, including statistical quantities such as mean, maximum, variance, and trend. The pressure abnormality feature is derived from the abnormality values of all pressure monitoring points and reflects the pressure anomaly pattern of the hydraulic pipeline. The temperature abnormality feature is derived from the temperature monitoring point data and reflects the thermodynamic anomaly pattern. The flow and vibration abnormality features are derived from the corresponding monitoring point data. The fault feature vector arranges the four abnormality features in a fixed order to form a four-dimensional numerical vector. This vector contains the numerical combination of the pressure, temperature, flow, and vibration abnormality features. The numerical range of each dimension is also zero to one. The overall pattern of the vector reflects the distribution of fault characteristics in the hydraulic pipeline system.

[0066] For example, when a regional airliner is operating a regional flight, the hydraulic pipeline system transitions from takeoff to cruise, and the flight phase identifier switches from takeoff to cruise. The weight coefficient assignment process queries the weight coefficient table based on the cruise identifier and adjusts the temperature sensor weight from 0.1 for takeoff to 0.4 for cruise, the pressure sensor weight from 0.4 to 0.3, and the flow sensor weight from 0.4 to 0.2, while the vibration sensor weight remains unchanged at 0.1. The weighted fusion calculation process reads the hydraulic oil temperature data displayed by the main hydraulic pipeline temperature sensor. This data drops significantly due to the low temperature environment at cruise altitude. By multiplying it by a high weight coefficient of 0.4, the temperature data dominates the fusion result. Simultaneously, the pressure sensor data is multiplied by a weight coefficient of 0.3, the flow sensor data by a weight coefficient of 0.2, and the vibration sensor data by a weight coefficient of 0.1. The sum of these four weighted values yields the differentiated monitoring data for that monitoring point. Data comparison and analysis compared this differentiated monitoring data with the theoretical temperature data of the digital twin pipeline model during the cruise phase. The actual temperature data fell below the normal range of the theoretical data, resulting in a negative deviation. Based on the degree and duration of the deviation, the abnormality level for this monitoring point was calculated to be 0.3. Feature vector quantization statistically analyzed the abnormality levels of all temperature monitoring points, yielding a temperature anomaly characteristic value of 0.25. Simultaneously, statistics for other types of monitoring points yielded a pressure anomaly characteristic value of 0.1, a flow anomaly characteristic value of 0.05, and a vibration anomaly characteristic value of 0.02. The resulting fault feature vector is a combination of temperature anomalies, pressure anomalies, flow anomalies, and vibration anomalies. This vector reflects the characteristic fault pattern of the hydraulic pipeline system during the cruise phase, primarily characterized by temperature anomalies.

[0067] In a specific embodiment, the step of inputting the differentiated monitoring data into the staged digital twin pipeline model for data comparison and analysis may specifically include the following steps:

[0068] The differentiated monitoring data are classified and processed according to the location of the monitoring points to obtain the real-time data sequence of each monitoring point. The corresponding theoretical data sequence is extracted from the phased digital twin pipeline model according to the flight phase identifier;

[0069] Based on the point-by-point difference calculation between the real-time data sequence and the theoretical data sequence, the data deviation value of each monitoring point is obtained, and the data deviation value is classified and stored according to pressure deviation, temperature deviation, flow deviation and vibration deviation;

[0070] The abnormality degree of each monitoring point is quantitatively calculated based on the data deviation value to obtain the abnormality degree score value. The monitoring points with a deviation value exceeding the normal range by 10% are classified as mild abnormalities, and the monitoring points with a deviation value exceeding the normal range by 20% are classified as moderate abnormalities.

[0071] The abnormality score values are normalized to obtain the abnormality value of each monitoring point. The abnormality value ranges from zero to one.

[0072] Specifically, the classification and sorting process performs spatial grouping operations on the differentiated monitoring data according to the physical installation position of each sensor in the hydraulic pipeline. The processing process first establishes a monitoring point location index table, and numbers and classifies all sensors in the pipeline system according to their specific locations in the pipeline network. The location classification adopts the pipeline topology analysis method to classify the monitoring points according to structural features such as main pipelines, branch pipelines, elbow positions, tee positions, valve positions, etc. Each location category contains multiple specific monitoring point numbers. The sorting process combines the differentiated monitoring data of the same location category and arranges them in chronological order to form a real-time data sequence. The sequence records the continuous data changes of each monitoring point over a period of time. At the same time, the phased digital twin pipeline model is queried according to the current flight phase identifier, and a theoretical data sequence that completely corresponds to the real-time monitoring point position is extracted from the model database. The theoretical sequence contains the standard numerical change pattern that each monitoring point should present under normal working conditions in the current flight phase.

[0073] The point-by-point difference calculation process performs a one-to-one numerical comparison between the real-time data sequence and the theoretical data sequence. This calculation process uses a point-by-point comparison method to subtract the value of each time point in the real-time data sequence from the value of the corresponding time point in the theoretical data sequence. The specific operation of the difference calculation is to subtract the theoretical standard value from the real-time monitoring value to obtain positive and negative deviation values. A positive deviation value indicates that the actual data is higher than the theoretical standard, a negative deviation value indicates that the actual data is lower than the theoretical standard, and a zero deviation value indicates that the actual data is completely consistent with the theoretical standard. After the calculation is completed, the data deviation values of all monitoring points are classified and stored according to the sensor type. The deviation value of the pressure sensor is classified into the pressure deviation category, the deviation value of the temperature sensor is classified into the temperature deviation category, the deviation value of the flow sensor is classified into the flow deviation category, and the deviation value of the vibration sensor is classified into the vibration deviation category, forming four independent deviation data storage areas.

[0074] The quantification of abnormality severity is performed based on the magnitude of the data deviations. This calculation process first establishes an abnormality severity scoring standard and compares the deviation values with a preset normal operating range. The normal operating range is a numerical range determined based on the hydraulic pipeline system's design parameters and historical operating data. A monitoring point's deviation value is considered normal when it falls within this range. The quantitative calculation uses a graded scoring method: when the deviation value exceeds the normal range by a small amount, it is scored as a mild abnormality; when the deviation value exceeds the normal range by a large amount, it is scored as a moderate abnormality; and when the deviation value exceeds the normal range by a significant amount, it is scored as a severe abnormality. The score is an integer scale, with a normal score of zero, a mild abnormality of one, a moderate abnormality of two, and a severe abnormality of three, forming the abnormality severity score for each monitoring point. Normalization converts the abnormality severity score to a standardized range. This process uses maximum normalization, dividing the score by the maximum score to obtain a normalized value between zero and one. Normalization involves dividing the score of each monitoring point by the maximum value of three in the scoring system. This results in a value between zero and one for the degree of abnormality, where zero represents a completely normal state, 0.33 represents a mild abnormality, 0.67 represents a moderate abnormality, and one represents a severe abnormality. Normalization eliminates jumps between different scoring levels, resulting in a continuously varying degree of abnormality.

[0075] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0076] According to the flight phase identifier, the corresponding BP neural network model is selected for model matching processing to obtain a BP neural network dedicated to the current phase. The takeoff phase selects the takeoff-specific BP neural network, the cruise phase selects the cruise-specific BP neural network, and the maneuver phase selects the maneuver-specific BP neural network;

[0077] The fault feature vector is input into the current stage dedicated BP neural network for forward propagation calculation processing to obtain the hidden layer output data, and the hidden layer output data is subjected to nonlinear transformation calculation through the activation function;

[0078] Based on the results of nonlinear transformation calculation, the output layer weight calculation is performed to obtain the fault type probability distribution data, which includes the probability of pipeline rupture, sealing failure, blockage, loosening, corrosion and fatigue cracking;

[0079] The maximum probability selection processing is performed on the fault type probability distribution data to obtain a fault prediction result, which includes the predicted fault type and the corresponding confidence value.

[0080] Specifically, the model matching process selects the corresponding dedicated network model from the pre-trained BP neural network model library based on the flight phase identifier. The process first establishes a correspondence table between the flight phase and the BP neural network model. Each flight phase corresponds to a specially trained BP neural network model. The takeoff-specific BP neural network is a network model trained for the transient pressure shocks and frequent landing gear movements of the hydraulic pipeline during the takeoff phase. The number of input layer nodes and the hidden layer structure of this network are optimized for the fault feature vector dimension of the takeoff phase. The cruise-specific BP neural network is a network model trained for the viscosity changes and temperature anomalies of the hydraulic oil in the high-altitude and low-temperature environment during the cruise phase. This network focuses on learning temperature-related failure modes. The maneuvering-specific BP neural network is a network model trained for the stress concentration and vibration anomalies of the pipeline under high overload conditions during the maneuvering flight phase. This network specifically identifies stress and vibration-related failure types. The matching process automatically loads the corresponding dedicated BP neural network from the model library by querying the current flight phase identifier and sets it as the dedicated BP neural network for the current phase.

[0081] The forward propagation calculation process passes the fault feature vector as input data to the current stage's dedicated BP neural network for calculation. This calculation process first inputs each element of the fault feature vector into the network's input layer nodes, with each input layer node receiving one feature vector element. The forward propagation calculation uses a weighted summation method, multiplying the input layer data by the connection weights and then summing the results to obtain the weighted input values for each hidden layer node. The hidden layer output data is calculated by substituting the weighted input values into the hidden layer's net input function, which numerically combines all weighted connections from the input layer to the hidden layer. The activation function performs a nonlinear transformation on the hidden layer output data. The activation function uses a Sigmoid function or a Reluctant Unit (ReLU) function to convert the linear combination results of the hidden layer into a nonlinear output. The purpose of the nonlinear transformation calculation is to introduce nonlinear characteristics, enabling the BP neural network to learn and identify complex fault patterns. The transformed data is then passed to the output layer as the final output of the hidden layer.

[0082] The output layer weight calculation process performs the final fault type prediction based on the nonlinear transformation results. This process multiplies the nonlinear transformation results of the hidden layer by the connection weights of the output layer, and then sums the weighted connections from all hidden layers to the output layer. The output layer contains six nodes, corresponding to the six fault types: pipe rupture, seal failure, blockage, loosening, corrosion, and fatigue cracking. The calculation result of each output node represents the probability of occurrence of the corresponding fault type. The weight calculation uses a softmax function to normalize the raw output layer results to ensure that the sum of the probabilities of the six fault types equals one, while also limiting all probability values to the range of zero to one. The fault type probability distribution data consists of six numerical values, each representing the probability of a specific fault type. A larger numerical value indicates a higher probability of that fault type. The maximum probability selection process compares and analyzes the fault type probability distribution data to identify the fault type with the highest probability value as the prediction result. This process uses a numerical comparison algorithm to compare the probability values of the six fault types one by one, identifying the highest probability value and its corresponding fault type. Fault prediction results consist of two parts: the first is the predicted fault type, which is the name of the fault type with the highest probability. The second is the corresponding confidence value, which is the probability value for that fault type. The confidence value reflects the reliability of the prediction result. A higher confidence value indicates a more reliable prediction result, while a lower confidence value indicates uncertainty in the prediction result. When the difference between the highest probability value and the second highest probability value is small, it indicates a mixture of multiple fault types, requiring further diagnostic analysis.

[0083] The above describes the aircraft hydraulic pipeline system fault prediction method based on digital twin in the embodiment of the present application. The following describes the aircraft hydraulic pipeline system fault prediction system based on digital twin in the embodiment of the present application. Figure 2 In the embodiments of the present application, an embodiment of a digital twin-based aircraft hydraulic pipeline system fault prediction system includes:

[0084] Identification module 201 is used to obtain flight parameters through the aircraft data bus, perform phase identification processing on the flight envelope based on the flight parameters, and obtain identifiers for the four flight phases of takeoff, cruise, maneuvering, and landing;

[0085] A mapping module 202 is configured to perform phase-by-phase mapping processing on the hydraulic pipeline load spectrum according to the flight phase identifier to obtain a differentiated physical parameter set for each phase, and construct a phased digital twin pipeline model using the differentiated physical parameter set for each phase;

[0086] an allocation module 203 for performing weighted allocation processing on the multi-sensor data according to the flight phase identifier to obtain differentiated monitoring data, and performing fault feature extraction processing on the differentiated monitoring data according to the staged digital twin pipeline model to obtain a fault feature vector;

[0087] The prediction module 204 is configured to perform prediction processing on the fault feature vector through a phased BP neural network to obtain a fault prediction result.

[0088] above Figure 2 The aircraft hydraulic pipeline system fault prediction system based on digital twin in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The aircraft hydraulic pipeline system fault prediction device based on digital twin in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0089] Reference Figure 3 In an embodiment of the present invention, a digital twin-based aircraft hydraulic pipeline system fault prediction device is provided. The digital twin-based aircraft hydraulic pipeline system fault prediction device can be a server, and its internal structure can be as follows: Figure 3 As shown. The aircraft hydraulic pipeline system fault prediction device based on digital twin includes a processor, a memory, a display screen, an input device, a network interface and a database connected via a system bus. Among them, the computer-designed processor is used to provide computing and control capabilities. The memory of the aircraft hydraulic pipeline system fault prediction device based on digital twin includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the aircraft hydraulic pipeline system fault prediction device based on digital twin is used to store the corresponding data in this embodiment. The network interface of the aircraft hydraulic pipeline system fault prediction device based on digital twin is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0090] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the digital twin-based aircraft hydraulic pipeline system fault prediction device to which the solution of the present invention is applied.

[0091] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions. When the instructions are executed on a computer, the computer executes the steps of the aircraft hydraulic pipeline system fault prediction method based on digital twins.

[0092] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0093] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a digital twin-based aircraft hydraulic pipeline system fault prediction device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0094] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A fault prediction method for aircraft hydraulic pipeline system based on digital twin, characterized by: The method comprises: The flight parameters are obtained through the aircraft data bus, and the flight envelope is processed according to the flight parameters to obtain the four flight phase identifiers of takeoff, cruise, maneuvering and landing; Performing stage-by-stage mapping processing on the hydraulic pipeline load spectrum according to the flight stage identifier to obtain a differentiated physical parameter set for each stage, and constructing a staged digital twin pipeline model using the differentiated physical parameter set for each stage; performing weight distribution processing on the multi-sensor data according to the flight phase identifier to obtain differentiated monitoring data, and performing fault feature extraction processing on the differentiated monitoring data according to the staged digital twin pipeline model to obtain a fault feature vector; The fault feature vector is predicted and processed by a phased BP neural network to obtain a fault prediction result.

2. The aircraft hydraulic pipeline system fault prediction method based on digital twin according to claim 1 is characterized in that: The method of obtaining flight parameters through the aircraft data bus and performing phase identification processing on the flight envelope according to the flight parameters to obtain four flight phase identifiers of takeoff, cruise, maneuvering and landing includes: collecting flight altitude, airspeed, angle of attack, overload coefficient, and control surface deflection angle data through an aircraft data bus, and normalizing the flight altitude, airspeed, angle of attack, overload coefficient, and control surface deflection angle data to obtain a flight state feature vector; Establishing a flight phase discrimination matrix based on the flight state feature vector, performing threshold comparison processing on the flight state feature vector, and obtaining a flight phase discrimination result, wherein the takeoff phase discrimination condition is a low altitude and high speed state, and the cruising phase discrimination condition is a high altitude and stable speed state; Inputting the flight phase discrimination result into a state machine algorithm for logic determination processing to obtain a state transition sequence of the current flight phase, and generating a corresponding flight phase code according to the state transition sequence; Identifier allocation processing is performed based on the flight phase code to obtain four flight phase identifiers: takeoff, cruise, maneuvering, and landing. The takeoff phase identifier is a takeoff identifier, the cruise phase identifier is a cruise identifier, the maneuvering phase identifier is a maneuvering identifier, and the landing phase identifier is a landing identifier.

3. The aircraft hydraulic pipeline system fault prediction method based on digital twin according to claim 1 is characterized in that: The hydraulic pipeline load spectrum is mapped in stages according to the flight stage identifier to obtain a differentiated physical parameter set for each stage, and a staged digital twin pipeline model is constructed using the differentiated physical parameter set for each stage, including: Based on the flight phase identifier, a preset load spectrum parameter library is searched, and the hydraulic pipeline load spectrum is extracted and processed in stages to obtain a transient pressure peak parameter in the takeoff phase, a hydraulic oil viscosity temperature coefficient parameter in the cruise phase, a pipeline stress concentration factor parameter in the maneuvering phase, and a flow pulsation frequency parameter in the landing phase; The transient pressure peak parameter of the takeoff phase, the hydraulic oil viscosity temperature coefficient parameter of the cruise phase, the pipeline stress concentration coefficient parameter of the maneuvering phase, and the flow pulsation frequency parameter of the landing phase are processed into parameter sets to obtain differentiated physical parameter sets for each phase; Performing physical attribute configuration processing on the pipeline geometric model according to the differentiated physical parameter sets of each stage to obtain a staged pipeline physical model, wherein a high dynamic response attribute is configured for the takeoff stage, a temperature sensitive attribute is configured for the cruise stage, and a stress analysis attribute is configured for the maneuvering stage; A virtual mapping construction process is performed based on the staged pipeline physical model to obtain a staged digital twin pipeline model, which includes an adaptive switching mechanism for physical parameters corresponding to each flight stage.

4. The aircraft hydraulic pipeline system fault prediction method based on digital twin according to claim 3 is characterized in that: The physical attribute configuration processing of the pipeline geometric model according to the differentiated physical parameter sets of each stage to obtain the staged pipeline physical model includes: Matching and calculating the transient pressure peak parameter during the takeoff phase with the wall thickness data of the pipeline geometric model to obtain pipeline wall thickness configuration data during the takeoff phase, and updating the wall thickness attribute value of the pipeline geometric model according to the pipeline wall thickness configuration data during the takeoff phase; performing temperature compensation calculation processing on the inner diameter of the pipeline based on the temperature coefficient parameter of the hydraulic oil viscosity in the cruise phase to obtain temperature-corrected inner diameter data for the cruise phase, and writing the temperature-corrected inner diameter data for the cruise phase into the inner diameter parameter field of the pipeline geometric model; performing stress-weighted calculation processing on the position coordinates of the pipeline nodes according to the stress concentration coefficient parameter of the pipeline in the maneuvering stage to obtain stress weight data of the nodes in the maneuvering stage, and adjusting the material density parameters of each node in the pipeline geometric model based on the stress weight data of the nodes in the maneuvering stage; performing flow resistance calculation processing on the landing phase flow pulsation frequency parameter and the pipeline length data to obtain pipeline flow resistance coefficient data for the landing phase, and modifying the surface roughness parameter of the pipeline geometric model according to the pipeline flow resistance coefficient data for the landing phase; The updated pipeline wall thickness attribute value, inner diameter parameter field, material density parameter and surface roughness parameter are processed through data integration to obtain a staged pipeline physical model.

5. The aircraft hydraulic pipeline system fault prediction method based on digital twin according to claim 1 is characterized in that: The multi-sensor data is weighted according to the flight phase identifier to obtain differentiated monitoring data, and the differentiated monitoring data is subjected to fault feature extraction processing according to the staged digital twin pipeline model to obtain a fault feature vector, including: The pressure sensor data, temperature sensor data, flow sensor data, and vibration sensor data of the key nodes of the hydraulic pipeline are processed by weight coefficient assignment according to the flight phase identifier to obtain a stage-by-stage sensor weight coefficient table, wherein the pressure weight coefficient for the takeoff phase is 0.4, and the temperature weight coefficient for the cruise phase is 0.4; performing weighted fusion calculation processing on the pressure sensor data, temperature sensor data, flow sensor data, and vibration sensor data based on the staged sensor weight coefficient table to obtain differentiated monitoring data, wherein the differentiated monitoring data includes key monitoring parameter values for each flight stage; Inputting the differentiated monitoring data into the staged digital twin pipeline model for data comparison and analysis, obtaining a deviation value between the monitoring data and the model data, and calculating the abnormality degree value of each monitoring point based on the deviation value; The abnormality degree value is subjected to feature vector quantization processing to obtain a fault feature vector, wherein the fault feature vector includes a numerical combination of pressure abnormality feature, temperature abnormality feature, flow abnormality feature and vibration abnormality feature.

6. The aircraft hydraulic pipeline system fault prediction method based on digital twin according to claim 5 is characterized in that: The step of inputting the differentiated monitoring data into the staged digital twin pipeline model for data comparison and analysis to obtain a deviation value between the monitoring data and the model data, and calculating the abnormality degree value of each monitoring point based on the deviation value includes: Classifying and processing the differentiated monitoring data according to the location of the monitoring points to obtain a real-time data sequence for each monitoring point, and extracting a corresponding theoretical data sequence from the staged digital twin pipeline model according to the flight phase identifier; Performing point-by-point difference calculation processing based on the real-time data sequence and the theoretical data sequence to obtain data deviation values of each monitoring point, and classifying and storing the data deviation values according to pressure deviation, temperature deviation, flow deviation and vibration deviation; A quantitative calculation of the abnormality degree of each monitoring point is performed based on the data deviation value to obtain an abnormality degree score value, wherein a monitoring point with a deviation value exceeding the normal range by 10% is classified as a mild abnormality, and a monitoring point with a deviation value exceeding the normal range by 20% is classified as a moderate abnormality; The abnormality degree score values are normalized to obtain abnormality degree values for each monitoring point, where the abnormality degree values range from zero to one.

7. The aircraft hydraulic pipeline system fault prediction method based on digital twin according to claim 1, characterized in that: The predictive processing of the fault feature vector by the staged BP neural network to obtain a fault prediction result includes: Selecting a corresponding BP neural network model according to the flight phase identifier to perform model matching processing to obtain a BP neural network dedicated to the current phase, wherein the takeoff-specific BP neural network is selected for the takeoff phase, the cruise-specific BP neural network is selected for the cruise phase, and the maneuvering-specific BP neural network is selected for the maneuvering phase; Inputting the fault feature vector into the current stage dedicated BP neural network for forward propagation calculation processing to obtain hidden layer output data, and performing nonlinear transformation calculation on the hidden layer output data through an activation function; Based on the results of the nonlinear transformation calculation, an output layer weight calculation process is performed to obtain fault type probability distribution data, wherein the fault type probability distribution data includes pipeline rupture probability, seal failure probability, blockage probability, loosening probability, corrosion probability and fatigue crack probability; The maximum probability selection processing is performed on the fault type probability distribution data to obtain a fault prediction result, which includes a predicted fault type and a corresponding confidence value.

8. A digital twin-based aircraft hydraulic pipeline system fault prediction system, characterized in that: For implementing the aircraft hydraulic pipeline system fault prediction method based on digital twin according to any one of claims 1 to 7, the aircraft hydraulic pipeline system fault prediction system based on digital twin comprises: The identification module is used to obtain flight parameters through the aircraft data bus, perform phase identification processing on the flight envelope based on the flight parameters, and obtain the four flight phase identifiers of takeoff, cruise, maneuvering and landing; a mapping module, configured to perform phased mapping processing on the hydraulic pipeline load spectrum according to the flight phase identifier to obtain a differentiated physical parameter set for each phase, and construct a phased digital twin pipeline model using the differentiated physical parameter set for each phase; an allocation module, configured to perform weighted allocation processing on the multi-sensor data according to the flight phase identifier to obtain differentiated monitoring data, and perform fault feature extraction processing on the differentiated monitoring data according to the staged digital twin pipeline model to obtain a fault feature vector; The prediction module is used to perform prediction processing on the fault feature vector through a staged BP neural network to obtain a fault prediction result.

9. A digital twin-based aircraft hydraulic pipeline system fault prediction device, characterized in that: It includes a memory and a processor, the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the aircraft hydraulic pipeline system fault prediction method based on digital twin according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the processor executes the aircraft hydraulic pipeline system fault prediction method based on digital twin according to any one of claims 1 to 7.

Images