PHM model verification management method based on digital twinning

Abstract

The embodiment of the invention provides a PHM model verification management method based on digital twinning. The PHM model verification management method is applied to the technical field of signal monitoring and analysis. The method comprises the following steps: receiving and analyzing real-time operation data of a physical system through a digital twinborn model, and dynamically adjusting parameters of the digital twinborn model based on a system health state vector; injecting a fault evolution algorithm into the digital twin model after parameter adjustment, and generating a high-simulation fault data sequence containing continuous fault features and intermittent fault features; running the PHM model in a pre-established virtual verification environment, enabling the PHM model to be coupled with the digital twin model, and receiving and running real-time simulation data from the digital twin model; and performing quantitative evaluation on the operation performance of the PHM model by adopting a dynamic weight evaluation system, and adaptively adjusting a fault injection strategy and model parameters in the digital twin model according to an evaluation result. In this way, precision, high efficiency and intelligence of PHM model verification can be achieved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of signal monitoring and analysis, in particular to a PHM model verification management method based on digital twinning. BACKGROUND

[0002] As a key support system for ensuring the normal take-off and landing and flight safety of the aircraft, the mechanical and electrical system of the civil aircraft covers multiple core subsystems such as the brake system, hydraulic system, landing gear system, etc., and its running state is directly related to the safety and economy of flight operations. With the rapid development of civil aircraft towards large-scale and intelligentization, through the construction of a fault prediction and health management (PHM) model to realize the early detection, accurate diagnosis and residual life prediction of mechanical and electrical system faults, it has become a core technical requirement to reduce unplanned maintenance downtime, reduce operating costs and avoid safety risks caused by hidden faults. The industry's demand for efficient and accurate PHM model verification technology is increasingly urgent.

[0003] In the current industry practice of PHM model verification, traditional technologies generally adopt the implementation path of static modeling combined with fixed fault injection, which specifically presents three aspects of technical status: Firstly, in the aspect of digital twinning model construction and calibration, the industry mostly sets constant temperature, standard load and other fixed environmental parameters and system operating conditions according to the design manual, and only performs model calibration through manual collection of physical system data with a cycle of 1-3 months, lacking the dynamic response capability to the real-time health state changes of the physical system, resulting in a deviation of the model parameters from the actual parameters of the physical system often exceeding ±5%, and the generated verification data being seriously out of line with the real flight conditions; Secondly, in the aspect of fault data generation, the traditional method mainly uses a single fault mode, generates a simple fault sequence with "normal-fault" two-state switching by manually setting fixed parameters such as voltage drop amplitude and pressure leakage rate, and the mixing ratio of fault data and normal data is fixed, which cannot simulate the intermittent faults (such as the cycle state of landing gear jamming and recovery) and multiple fault concurrent composite fault scenarios frequently occurring in the actual operation of civil aircraft, and the parameter fluctuations lack asynchronous coordination characteristics, making the verification of the PHM model under complex fault scenarios obviously missing; Thirdly, in the aspect of virtual verification environment construction, the traditional scheme mostly adopts a single machine deployment mode, allocates unified specification of computing resources for all verification tasks, and realizes verification of different fault scenarios by manually switching configuration files, and the data injection and model interaction rely on serial processing at the software level, lacking hardware acceleration support, which not only leads to low resource utilization efficiency and high scene switching delay, but also is difficult to meet the verification needs of multiple parameter concurrent monitoring.

[0004] The inherent defects of the aforementioned traditional technical solutions directly lead to two key issues: First, the digital twin model lacks dynamic consistency with the physical system, failing to accurately reflect changes in the physical system's health status (such as increased power supply resistance and increased hydraulic leakage coefficient), resulting in a significant reduction in the authenticity and effectiveness of the verification data, severely impacting the accuracy of PHM model verification results and the reliability of subsequent fault diagnosis. Second, the limited authenticity and coverage of fault data make it impossible to fully simulate complex fault scenarios in real flight, hindering the PHM model from forming an effective response capability to real faults through sufficient scenario verification. Consequently, it cannot guarantee the safety and stability of civil aircraft electromechanical systems in actual operation, and also restricts the in-depth application and effectiveness of PHM technology in the civil aircraft field. Summary of the Invention

[0005] This disclosure provides a digital twin-based PHM model verification management method, which solves the technical problem that existing methods have data that is out of sync with real working conditions, seriously affecting the accuracy, reliability and authenticity of PHM model verification.

[0006] According to a first aspect of this disclosure, a PHM model validation management method based on digital twins is provided. The method includes: The system receives and parses real-time operational data of the physical system through a digital twin model, and dynamically adjusts the parameters of the digital twin model based on the system health status vector. A pre-defined fault evolution algorithm is injected into the digital twin model after parameter adjustment to generate a highly simulated fault data sequence containing both continuous and intermittent fault characteristics. Run the PHM model in a pre-built virtual verification environment, couple it with the digital twin model, and receive and run real-time simulation data from the digital twin model. A dynamic weighted evaluation system is used to quantitatively evaluate the performance of the PHM model, and the fault injection strategy and model parameters in the digital twin model are adaptively adjusted based on the evaluation results.

[0007] In addition to the aspects described above and any possible implementation, a further implementation is provided in which the system health status vector is composed of multiple health status index components, and the health status of the health status index components is calculated through real-time operating data of the physical system.

[0008] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the dynamic adjustment of the digital twin model parameters based on the system health state vector includes: When the health status of at least two health status index components is simultaneously lower than the first preset threshold, the comprehensive adjustment coefficient of the parameters is calculated based on the weight superposition and total amount capping mechanism, and the adjustment priority is allocated according to the safety criticality of the parameters. The health status changes of each health status index component are periodically calculated. If the health status change of any health status index component is greater than or equal to the second preset threshold, a parameter adjustment command is triggered. If the change in health status of all health status index components is less than the second preset threshold, then parameter adjustment is paused. After the parameters are updated, the deviation between the output of the digital twin model and the parameters of the physical system is verified, and the updated parameters or the state before adjustment is decided based on the verification results.

[0009] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein the comprehensive adjustment coefficient for the parameters calculated based on the weight superposition and total amount capping mechanism includes: If multiple health status index components adjust the same digital twin parameter in the same direction, the comprehensive adjustment coefficient is calculated according to the influence weight of each health status index component. If multiple health status index components adjust the same digital twin parameter in opposite directions, the adjustment direction of the health status index component with the greater influence weight shall prevail, and the adjustment magnitude of the health status index component with the smaller influence weight shall be superimposed at a predetermined ratio. When the overall adjustment exceeds the third preset threshold, the adjustment of safety-critical parameters is prioritized, and the adjustment amount of non-safety-critical parameters is adjusted based on the influence weight.

[0010] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein injecting a preset fault evolution algorithm into the adjusted digital twin model to generate a highly simulated fault data sequence containing continuous fault characteristics and intermittent fault characteristics includes: Inject a fault evolution algorithm based on a reinforcement learning model into the adjusted digital twin model; Construct a state vector with the deviation between the output parameters of the digital twin model and the parameters of the physical system as the core indicator and the system health state vector as the core, and define the action space; The fault evolution rate is dynamically controlled based on the real-time calculated system performance degradation gradient, generating a highly simulated fault data sequence that includes both continuous and intermittent fault characteristics.

[0011] As described above and in any possible implementation, a further implementation is provided, wherein the step of dynamically controlling the fault evolution rate based on the real-time calculated system performance degradation gradient to generate a highly simulated fault data sequence containing continuous fault characteristics includes: The system calculates the probability of performance degradation in real time. If the degradation probability is greater than or equal to a fourth preset threshold, a fault acceleration mechanism is triggered to adjust the fault evolution rate. Under the dynamic control of the fault evolution rate, the simulation of the fault degree gradually develops from the initial value to the complete path of system failure, generating a continuous fault data sequence.

[0012] In addition to the aspects described above and any possible implementations, a further implementation is provided in which generating a highly simulated fault data sequence containing intermittent fault characteristics includes: The intermittent behavior of faults is simulated based on a multi-state Markov jump model, and the fault states are extended to normal state, mild abnormal state, severe abnormal state and fault recovery state. Establish a mapping relationship between the fault state and the fluctuation of system parameters, define differentiated fluctuation ranges for system parameters under different states, and realize dynamic switching between states through a preset state transition probability matrix; Based on the multi-state Markov transition model and the mapping relationship between the fault state and parameter fluctuations, the cyclical transfer process of the fault between normal, mildly abnormal, severely abnormal, and fault recovery states is simulated to generate an intermittent fault data sequence containing parameter fluctuation sequences under each state.

[0013] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein the virtual verification environment includes a data injection module, a PHM interface module, and an evaluation and analysis module; wherein, The data injection module is used to take the highly simulated fault data sequence as input, mix the fault data stream and the normal data stream in a proportional manner, and input the mixed data stream into the PHM model in real time. The PHM interface module is used to control API latency and ensure the real-time data interaction between the PHM model and the virtual environment. The evaluation and analysis module is used to compare the analysis results output by the PHM model with the preset fault injection information in real time, quantify and calculate the performance indicators of detection delay, diagnostic accuracy and prediction error, and support concurrent processing of multi-parameter monitoring scenarios.

[0014] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein running the PHM model in a pre-built virtual verification environment, coupling it with the digital twin model, and receiving and running real-time simulation data from the digital twin model includes: The data injection module mixes the highly simulated fault data sequence with the normal operation data of the physical system according to a preset ratio to generate a mixed data stream. The mixing process is accelerated using FPGA hardware acceleration technology, and the mixed data stream is input to the PHM model in real time through the PHM interface module.

[0015] In addition to the aspects and any possible implementations described above, a further implementation is provided, wherein the step of using a dynamic weight evaluation system to quantitatively evaluate the performance of the PHM model, and adaptively adjusting the fault injection strategy and model parameters in the digital twin model based on the evaluation results includes: After receiving the mixed data stream, the PHM model processes it based on the built-in health status analysis algorithm and outputs analysis results data including fault detection status, diagnosis type and predicted remaining service life. The evaluation and analysis module receives the analysis results data output by the PHM model in real time, compares the analysis results data with the preset fault injection information, and quantitatively calculates the performance indicators of the PHM model based on the comparison results. The weights of each performance indicator are dynamically assigned based on the fault type, system operation stage, and system health status vector, and a comprehensive performance score is calculated. The overall performance score is compared with a preset threshold. If the overall performance score is lower than the preset threshold, an adjustment instruction is generated to adaptively adjust the fault injection strategy and model parameters in the digital twin model.

[0016] According to a second aspect of this disclosure, a PHM model verification and management device based on digital twins is provided. The device includes: The digital twin module is used to receive and parse real-time operating data of the physical system through the digital twin model, and dynamically adjust the parameters of the digital twin model based on the system health status vector; A pre-defined fault evolution algorithm is injected into the digital twin model after parameter adjustment to generate a highly simulated fault data sequence containing both continuous and intermittent fault characteristics. The PHM model is used to run the PHM model in a pre-built virtual verification environment, thereby coupling it with the digital twin model and receiving and running real-time simulation data from the digital twin model. The feedback optimization module is used to quantitatively evaluate the performance of the PHM model using a dynamic weight evaluation system, and adaptively adjust the fault injection strategy and model parameters in the digital twin model based on the evaluation results.

[0017] According to a third aspect of this disclosure, an electronic device is provided. The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.

[0018] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the methods according to the first and / or second aspects of this disclosure.

[0019] In this disclosure, firstly, the model receives and analyzes real-time operating data of the physical system through a digital twin model and dynamically adjusts the model parameters based on the system health state vector. This breaks through the limitations of traditional fixed parameter modeling and long-cycle calibration, enabling the digital twin model to match the actual working conditions of the physical system in real time. This greatly improves the dynamic consistency between the digital twin model and the physical system, and provides basic data support that fits the real working conditions for subsequent PHM model verification. Secondly, by injecting a pre-set fault evolution algorithm into the adjusted digital twin model, a highly realistic fault data sequence covering both continuous and intermittent fault characteristics is generated. This breaks through the limitations of traditional single fault modes, not only fully covering the continuous evolution path from minor faults to system failure, but also accurately simulating the intermittent fault cycle state of "normal-mild anomaly-severe anomaly-fault recovery". This fills the data gap in the verification of complex fault scenarios, allowing the PHM model to withstand more comprehensive and realistic fault scenario tests. Furthermore, the PHM model and the digital twin model are coupled and run through a pre-built virtual verification environment. Containerized cluster management enables dynamic allocation and elastic scaling of computing resources. Combined with FPGA hardware acceleration technology, low latency and efficient parallel processing of mixed data are ensured. At the same time, fast scene switching is achieved through Restful API, which significantly improves the concurrent processing capability, resource utilization efficiency and real-time performance of the verification environment, providing a stable and reliable operating platform for the efficient verification of the PHM model. Finally, a dynamic weighted evaluation system was adopted to quantitatively evaluate the performance of the PHM model. Based on the evaluation results, the fault injection strategy and model parameters of the digital twin model were adaptively adjusted to form a closed-loop optimization mechanism. This mechanism can specifically strengthen the verification of weak fault types in the PHM model and dynamically optimize the configuration of various parameters in the verification process, effectively improving the accuracy and effectiveness of PHM model verification. Ultimately, the performance of the PHM model in fault detection, diagnosis, and remaining life prediction was significantly optimized, meeting the application requirements of civil aircraft electromechanical systems for high reliability and high accuracy of PHM technology. This provides a strong guarantee for reducing unplanned maintenance downtime, lowering operating costs, and mitigating safety risks.

[0020] It should be understood that the description in the Summary of the Invention section is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0021] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. The drawings are provided for a better understanding of the invention and are not intended to limit the scope of this disclosure. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein: Figure 1 A flowchart illustrating a digital twin-based PHM model verification management method provided by an embodiment of this disclosure is shown. Figure 2 A structural diagram of a digital twin-based PHM model verification and management device is shown, according to an embodiment of the present disclosure. Figure 3 A structural diagram of an exemplary electronic device capable of implementing embodiments of the present disclosure is shown. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0023] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0024] In this disclosure, firstly, the model receives and analyzes real-time operating data of the physical system through a digital twin model and dynamically adjusts the model parameters based on the system health state vector. This breaks through the limitations of traditional fixed parameter modeling and long-cycle calibration, enabling the digital twin model to match the actual working conditions of the physical system in real time. This greatly improves the dynamic consistency between the digital twin model and the physical system, and provides basic data support that fits the real working conditions for subsequent PHM model verification. Secondly, by injecting a pre-set fault evolution algorithm into the adjusted digital twin model, a highly realistic fault data sequence covering both continuous and intermittent fault characteristics is generated. This breaks through the limitations of traditional single fault modes, not only fully covering the continuous evolution path from minor faults to system failure, but also accurately simulating the intermittent fault cycle state of "normal-mild anomaly-severe anomaly-fault recovery". This fills the data gap in the verification of complex fault scenarios, allowing the PHM model to withstand more comprehensive and realistic fault scenario tests. Furthermore, the PHM model and the digital twin model are coupled and run through a pre-built virtual verification environment. Containerized cluster management enables dynamic allocation and elastic scaling of computing resources. Combined with FPGA hardware acceleration technology, low latency and efficient parallel processing of mixed data are ensured. At the same time, fast scene switching is achieved through Restful API, which significantly improves the concurrent processing capability, resource utilization efficiency and real-time performance of the verification environment, providing a stable and reliable operating platform for the efficient verification of the PHM model. Finally, a dynamic weighted evaluation system was adopted to quantitatively evaluate the performance of the PHM model. Based on the evaluation results, the fault injection strategy and model parameters of the digital twin model were adaptively adjusted to form a closed-loop optimization mechanism. This mechanism can specifically strengthen the verification of weak fault types in the PHM model and dynamically optimize the configuration of various parameters in the verification process, effectively improving the accuracy and effectiveness of PHM model verification. Ultimately, the performance of the PHM model in fault detection, diagnosis, and remaining life prediction was significantly optimized, meeting the application requirements of civil aircraft electromechanical systems for high reliability and high accuracy of PHM technology. This provides a strong guarantee for reducing unplanned maintenance downtime, lowering operating costs, and mitigating safety risks.

[0025] The following detailed description, with reference to the accompanying drawings and specific embodiments, illustrates a PHM model verification and management method based on digital twins provided in this disclosure.

[0026] Figure 1 A flowchart illustrating a digital twin-based PHM model validation management method provided by an embodiment of this disclosure is shown, such as... Figure 1 As shown, a PHM model validation management method 100 based on digital twins may include the following steps: S110 receives and parses real-time operating data of the physical system through a digital twin model, and dynamically adjusts the parameters of the digital twin model based on the system health status vector.

[0027] In some embodiments, the system health status vector consists of multiple health status index components, and the health status of the health status index components is calculated using real-time operating data of the physical system.

[0028] In some embodiments, dynamically adjusting the parameters of the digital twin model based on the system health state vector includes: When the health status of at least two health status index components is simultaneously lower than the first preset threshold, the comprehensive adjustment coefficient of the parameters is calculated based on the weight superposition and total amount capping mechanism, and the adjustment priority is allocated according to the safety criticality of the parameters. The health status changes of each health status index component are periodically calculated. If the health status change of any health status index component is greater than or equal to the second preset threshold, a parameter adjustment command is triggered. If the change in health status of all health status index components is less than the second preset threshold, then parameter adjustment is paused. After the parameters are updated, the deviation between the output of the digital twin model and the parameters of the physical system is verified, and the updated parameters or the state before adjustment is decided based on the verification results.

[0029] In some embodiments, the comprehensive adjustment coefficient for calculating parameters based on the weighted superposition and total amount capping mechanism includes: If multiple health status index components adjust the same digital twin parameter in the same direction, the comprehensive adjustment coefficient is calculated according to the influence weight of each health status index component. If multiple health status index components adjust the same digital twin parameter in opposite directions, the adjustment direction of the health status index component with the greater influence weight shall prevail, and the adjustment magnitude of the health status index component with the smaller influence weight shall be superimposed at a predetermined ratio. When the overall adjustment exceeds the third preset threshold, the adjustment of safety-critical parameters is prioritized, and the adjustment amount of non-safety-critical parameters is adjusted based on the influence weight.

[0030] Specifically, the digital twin model of the target civil aircraft's electromechanical system (including subsystems such as braking system, hydraulic system, and landing gear system) receives real-time operational data transmitted from the physical system. This data includes core parameters such as 28VDC power supply bus voltage, hydraulic system pressure, and landing gear deflection angle. To ensure data quality and support the accuracy of subsequent modeling and adjustments, the received real-time data needs to be preprocessed. Sliding window filtering combined with the 3σ criterion is used to remove outliers from the data. The timestamp alignment accuracy is strictly controlled within ±10ms. For parameters such as voltage signals that are susceptible to interference, Kalman filtering is additionally implemented to reduce noise. At the same time, the dynamic time warping (DTW) algorithm is used to align multi-source asynchronous data streams and compensate for transmission delays caused by differences in installation location and transmission links of different sensors, ensuring the reliability of subsequent calculations from the data source.

[0031] Specifically, based on the preprocessed real-time operational data, a system health status vector is constructed, which is defined as follows: ,in, Represents voltage health status, Represents the pressure health of the hydraulic system. Representing the landing gear deflection angle health, the health of each Health Status Index (HSI) component is calculated by comparing the real-time monitored values ​​of the corresponding physical parameters with their rated values ​​and thresholds, such as voltage health. The deviation between the actual value of the pre-processed 28VDC bus voltage and the rated 28V can be quantitatively determined, while the overall health status vector is used to intuitively reflect the current health level of the physical system, providing a core basis for adjusting the parameters of the digital twin model.

[0032] Specifically, when dynamically adjusting the parameters of the digital twin model based on the system health status vector, the first preset threshold is defined as 0.5 (i.e., when the health status is ≤0.5, the component health status needs to be adjusted). When the health status of at least two HSI components is simultaneously below this threshold, parameter adjustment calculation is initiated: the comprehensive adjustment coefficient λ of the parameters is calculated based on the weight superposition and total amount capping mechanism, and the specific formula is as follows: in, The weight of the influence of the i-th HSI component on the target adjustment parameter; Defined as 0.5- Only when The value is taken when ≤0.5. >0.5 =0; n is the number of HSI components involved in the adjustment.

[0033] At the same time, a total limit constraint is set, meaning that the final adjustment range of the parameters must meet the following requirements. ;in, The adjusted parameter values, (To adjust the parameter values ​​before adjustment) to avoid the digital twin model deviating from the actual working conditions of the physical system due to excessive adjustment.

[0034] Specifically, regarding the adjustment direction, if multiple HSI components adjust the same digital twin parameter in the same direction (e.g., both trigger parameters increase or decrease), the final adjustment amount is directly calculated using the aforementioned comprehensive adjustment coefficient λ. The adjustment formula is as follows: If multiple HSI components adjust the same parameter in opposite directions (e.g.) Increase the trigger parameter If the trigger parameter decreases, then the HSI component with a larger weight (referred to as the high-weight component) will be affected. , The adjustment direction shall be based on the predetermined direction, and at the same time, the HSI component with a smaller impact weight (referred to as the low-weight component) shall be added at a predetermined ratio of 30%. , The adjustment range is determined by the parameter's impact on system security. When the overall adjustment range calculated as described above exceeds 20% (i.e., the third preset threshold), the adjustment priority is determined based on the parameter's impact on system security. The first level is the critical safety parameter: landing gear friction coefficient. Hydraulic pump output pressure coefficient ; Level 2 is the core performance parameter: internal resistance of the braking system. Hydraulic system leakage coefficient ; The third level is an auxiliary protection parameter: bus filter capacitor value. Servo response delay .

[0035] Prioritize retaining the adjustment range of Level 1 and Level 2 parameters, and truncate the adjustment amount of Level 3 non-safety critical parameters according to their impact weight from large to small, to ensure that the parameter adjustments of high-weight HSI components are more in line with the changes in the health status of the physical system.

[0036] Specifically, the parameter adjustment trigger mechanism combines periodic monitoring with threshold judgment. The HSI vector H is calculated every 500ms, comparing the health changes of each HSI component within previous and subsequent periods. If any HSI component ( If the change in the health status of the physical system is ≥0.05 (i.e., the second preset threshold), it is determined that the health status of the physical system has changed significantly, and the parameter adjustment instruction is immediately triggered to generate. The parameters of the digital twin model are updated in order of priority from high to low. If the change in the health status of all HSI components is less than 0.05 and the stable state lasts for more than 10 seconds, the parameter adjustment process is suspended in order to reduce unnecessary fluctuations in the digital twin model.

[0037] After the parameters are updated, the effect of the adjustment needs to be verified in real time—calculate the output parameters of the digital twin model. Real-time parameters of physical system deviation ,like If the adjustment is valid, the updated parameters are retained; otherwise... If the current adjustment fails to match the physical system's operating conditions, the model should be immediately reverted to the parameters before the adjustment. The influence weights of each HSI component and the target parameter should be recalculated to provide a basis for the next adjustment and optimization calculation. This will ensure that the digital twin model always maintains a high degree of dynamic consistency with the physical system and provide an accurate simulation basis for subsequent PHM model verification.

[0038] S120 injects a preset fault evolution algorithm into the digital twin model after parameter adjustment to generate a highly simulated fault data sequence containing continuous fault characteristics and intermittent fault characteristics.

[0039] In some embodiments, injecting a preset fault evolution algorithm into the adjusted digital twin model to generate a highly simulated fault data sequence containing both continuous and intermittent fault characteristics includes: Inject a fault evolution algorithm based on a reinforcement learning model into the adjusted digital twin model; Construct a state vector with the deviation between the output parameters of the digital twin model and the parameters of the physical system as the core indicator and the system health state vector as the core, and define the action space; The fault evolution rate is dynamically controlled based on the real-time calculated system performance degradation gradient, generating a highly simulated fault data sequence that includes both continuous and intermittent fault characteristics.

[0040] In some embodiments, dynamically controlling the fault evolution rate based on the real-time calculated system performance degradation gradient to generate a highly simulated fault data sequence containing continuous fault characteristics includes: The system calculates the probability of performance degradation in real time. If the degradation probability is greater than or equal to the fourth preset threshold, a fault acceleration mechanism is triggered to adjust the fault evolution rate. Under the dynamic control of the fault evolution rate, the simulation of the fault degree gradually develops from the initial value to the complete path of system failure, generating a continuous fault data sequence.

[0041] In some embodiments, generating a highly simulated fault data sequence containing intermittent fault characteristics includes: The intermittent behavior of faults is simulated based on a multi-state Markov jump model, and the fault states are extended to normal state, mild abnormal state, severe abnormal state and fault recovery state. Establish a mapping relationship between fault states and system parameter fluctuations, define differentiated fluctuation ranges for system parameters under different states, and realize dynamic switching between states through a preset state transition probability matrix; Based on the multi-state Markov transition model and the mapping relationship between fault state and parameter fluctuation, the cyclical transfer process of fault between normal, mildly abnormal, severely abnormal and fault recovery states is simulated to generate intermittent fault data sequences containing parameter fluctuation sequences under each state.

[0042] Specifically, a pre-defined fault evolution algorithm is injected into the parameter-adjusted digital twin model. This algorithm, built with a dual deep Q-network (DDQN) as the core reinforcement learning model, aims to accurately simulate the evolution of actual faults in civil aircraft electromechanical systems, providing algorithmic support for subsequent high-fidelity fault data generation. In the algorithm construction process, the definition of the state vector is first clarified. Combining the dynamic relationship between the digital twin model and the physical system, the state vector is set as... in, Represents the real-time parameters of the physical system at time t. Output parameters of digital twin model The absolute difference (i.e.) ), used to quantify the consistency deviation between the two; while The system health state vector constructed in S110 is continued to ensure the continuity between the state description and the previous parameter adjustment. At the same time, the action space is defined based on the common fault types of civil aircraft electromechanical systems, covering core fault injection modes such as "voltage drop", "pressure leakage" and "landing gear jamming". Each fault mode is further subdivided into different severity levels and the parameter change rules are clearly defined. For example, voltage drop is divided into mild (28V→25V, lasting 5s), moderate (28V→22V, lasting 10s), and severe (28V→18V, lasting 20s), with a uniform drop rate of 5V / s. Pressure leakage decreases at a rate of 0.5MPa / min for mild and 2MPa / min for severe. Landing gear jamming corresponds to the deflection angle change threshold under different jamming degrees, so that the fault injection logic of the action space fits the actual system fault characteristics.

[0043] Specifically, by calculating the system performance degradation gradient in real time. The dynamic control of fault evolution rate uses a gradient to reflect the fault deterioration rate, which is calculated based on the variation of mean failure time, and is defined by the following formula: in, The default calculation step size is 100ms; Let be the average failure time of the system at time t, and satisfy . ( (where k is the initial mean time to failure and k is the decay coefficient, which is related to the characteristics of the system components).

[0044] In fault evolution control, a fourth preset threshold is set as follows: ( (where β is the preset maximum system performance degradation gradient, and β is the trigger threshold coefficient, set according to the verification scenario requirements). When the real-time calculation... When the fault is determined to have entered a rapid deterioration phase, the fault acceleration mechanism is immediately triggered, updating the evolution speed to: in The initial evolution velocity, This is a fault acceleration factor, and its dynamic adjustment ensures that key data during the fault deterioration phase are not missed.

[0045] When generating a continuous fault data sequence based on this, it fully covers everything from "minor faults" (… ) → Moderate fault ( ) → Severe Fault ( System failure The complete evolutionary path of "=0"; where, The maximum permissible deviation threshold for the corresponding parameters is used to enable continuous fault data to simulate the entire process of a fault from its inception to complete failure.

[0046] Specifically, for the generation of intermittent fault data sequences, a multi-state Markov jump model is used to extend the traditional "normal-fault" two-state model, clearly defining the normal state ( =0), mild abnormality ( =1), severe abnormalities ( =2), Fault Recovery ( =3) Four states are defined, and a state transition probability matrix is ​​preset based on civil aircraft fault statistics. For example, the probability of transitioning from a normal state to a slightly abnormal state is 0.1, and the probability of transitioning from a severely abnormal state to a fault recovery state is 0.2, ensuring that state switching conforms to the actual intermittent characteristics of faults. Simultaneously, a mapping relationship between "fault state - system parameter fluctuation" is established, defining differentiated fluctuation ranges for key parameters under different states: taking voltage as an example... For example: Normal state ( =0): =28±0.5V (fluctuation range) =0.5V). Mild abnormality ( =1): =28±(0.5+0.3⋅ V (fluctuation amplitude varies with the degree of evolution) Increase); Severe abnormality ( =2): =28±(1.0+0.5⋅ V (The fluctuation range has further expanded); Fault recovery ( =3): =28±(0.5+0.2⋅(1− V (The fluctuation amplitude decreases as the recovery process progresses); Specifically, the duration for a parameter to jump from a stable value to an outlier. It follows a Weibull distribution, and its probability density function is: Furthermore, different parameters correspond to differentiated distributed parameters—the shape parameter m=1.2 and the scale parameter of the voltage jump. =0.8s, shape parameter m=1.5, scale parameter of pressure jump =1.2s, to simulate the real characteristics of asynchronous coordinated fluctuations of multiple parameters in intermittent faults. Finally, based on the above multi-state Markov transition model and parameter fluctuation rules, the fault is simulated in the cyclical transition process of "normal → mild anomaly → severe anomaly → fault recovery → normal", generating intermittent fault data containing parameter fluctuation sequences in each state. Together with continuous fault data, this constitutes a highly realistic fault data sequence covering typical fault scenarios of civil aircraft electromechanical systems, providing a comprehensive and realistic data source for subsequent PHM model verification.

[0047] S130 runs the PHM model in a pre-built virtual verification environment, couples it with the digital twin model, and receives real-time simulation data from the digital twin model.

[0048] In some embodiments, the virtual verification environment includes a data injection module, a PHM interface module, and an evaluation and analysis module; wherein... The data injection module is used to take a highly simulated fault data sequence as input, mix the fault data stream and the normal data stream in a proportional manner, and input the mixed data stream into the PHM model in real time. The PHM interface module is used to control API latency and ensure the real-time data interaction between the PHM model and the virtual environment. The evaluation and analysis module is used to compare the analysis results output by the PHM model with the preset fault injection information in real time, quantify and calculate the performance indicators of detection delay, diagnostic accuracy and prediction error, and support concurrent processing of multi-parameter monitoring scenarios.

[0049] In some embodiments, the PHM model is run in a pre-built virtual verification environment and coupled with the digital twin model, receiving real-time simulation data from the digital twin model includes: The data injection module mixes the high-fidelity fault data sequence with the normal operation data of the physical system according to a preset ratio to generate a mixed data stream. The mixing process is accelerated using FPGA hardware acceleration technology, and the mixed data stream is input to the PHM model in real time through the PHM interface module.

[0050] Specifically, the pre-built virtual verification environment uses the highly simulated fault data sequence generated by S120 as the core input. The verification carrier is built based on containerization technology. The container resource pool is managed uniformly using a Kubernetes cluster to adapt to the different complexity differences of different tasks in the verification of the PHM model of civil aircraft electromechanical system: For simple verification tasks (such as basic fault detection verification of a single subsystem), a resource specification of 1 CPU core + 2GB memory is allocated; for complex tasks (such as composite fault diagnosis verification of multiple subsystems), a resource specification of 4 CPU cores + 8GB memory is allocated; and for ultra-complex tasks (such as joint verification of fault evolution and remaining life prediction of the whole system), a resource specification of 8 CPU cores + 16GB memory is allocated. At the same time, dynamic resource scheduling rules are set: when the container CPU utilization rate is ≥80% for 30 seconds, the cluster automatically triggers expansion to supplement computing resources and avoid task lag; when the CPU utilization rate is ≤30% for 1 minute, it automatically performs shrinkage to release redundant resources, ensuring the resource utilization efficiency and operational stability of the entire verification environment.

[0051] Specifically, this virtual verification environment incorporates three core functional modules: a data injection module, a PHM interface module, and an evaluation and analysis module. These modules work together to support the coupled operation of the PHM model and the digital twin model. The core function of the data injection module is to achieve accurate mixing and efficient transmission of fault data and normal data. It first receives the highly simulated fault data sequence output by the S120, simultaneously accesses the normal operation data of the physical system, and then calculates the system performance degradation gradient in real time within the S120. Dynamically adjust the mixing ratio and define the mixing ratio coefficient: in, This is the preset maximum performance degradation gradient for the system.

[0052] When the degree of failure evolution intensifies ( When (increase), As the synchronous increase occurs, the proportion of fault data in the hybrid stream increases accordingly, making the injected data more consistent with the actual data distribution at the current fault evolution stage, thus avoiding the problem of insufficient representativeness of fault data caused by traditional fixed-proportion mixing. To ensure mixing efficiency and real-time performance, this module adopts FPGA hardware acceleration technology, which realizes the rapid fusion of fault data stream and normal data stream through parallel computing at the hardware level, strictly controlling the mixing delay to ≤10ms, and ensuring that the data input to the PHM model can reflect the simulation state of the digital twin model in real time.

[0053] Specifically, the PHM interface module acts as a "bridge" for data interaction between the PHM model and the virtual verification environment and digital twin model. Its core objective is to control API latency to ensure real-time performance. On the one hand, the module uses a Restful API architecture to implement data transmission, keeping the interface latency within the range required for verification. This ensures that the real-time simulation data output by the digital twin model can be quickly transmitted to the PHM model, and the analysis results of the PHM model can be promptly fed back to the evaluation and analysis module. On the other hand, to meet the need for continuous verification of multiple fault scenarios, the module adopts a three-layer nested dictionary structure for the configuration file. The top layer associates the fault scenario ID with different fault evolution paths in S120, the middle layer corresponds to the environmental parameters and system operating condition settings of the digital twin model in S110, and the bottom layer presets the evaluation index weights and thresholds. Before switching scenarios, the configuration files of the first three scenarios to be switched are preloaded into memory. Combined with the efficient calling characteristics of the Restful API, a fast scenario switching of ≤0.5 seconds is achieved, avoiding the impact of configuration file loading time on the continuity of model coupling operation.

[0054] Specifically, the evaluation and analysis module is primarily responsible for the preliminary data preparation for subsequent performance quantification evaluation. It not only receives real-time analysis results from the PHM model, including fault detection status, diagnostic type, and predicted remaining life, but also retrieves pre-set fault injection information (such as fault type, occurrence time, and severity level) from S120 and performs precise comparison between the two. Simultaneously, this module supports concurrent processing of multi-parameter monitoring scenarios in civil aircraft electromechanical systems. It can simultaneously process core parameter verification data from multiple subsystems, such as braking system voltage, hydraulic system pressure, and landing gear deflection angle, thus reducing subsequent calculations and detection delays. Diagnostic accuracy Prediction error This lays the foundation for performance indicators such as [the performance indicators].

[0055] Specifically, during the coupled operation of the PHM model and the digital twin model, the data injection module first merges the highly simulated fault data sequence with the normal operation data of the physical system according to a dynamically adjusted mixing ratio to generate a mixed data stream. After FPGA hardware acceleration to ensure that the mixing delay is ≤10ms, the mixed data stream is transmitted to the PHM model in real time through the Restful API of the PHM interface module. After receiving the mixed data stream, the PHM model processes it based on its built-in health status analysis algorithms (such as fault feature extraction and remaining life prediction models). At the same time, the digital twin model continuously outputs real-time parameters reflecting the current simulation conditions (such as voltage and pressure change data in the simulated fault evolution), which are synchronously transmitted to the PHM model through the PHM interface module. This allows the PHM model to obtain the simulation status of the digital twin model in real time, achieving deep coupling between the two. This ensures that the verification process of the PHM model is always based on dynamic simulation data that closely resembles the real operating conditions, thus improving the reliability of the verification results.

[0056] S140 employs a dynamic weight evaluation system to quantitatively evaluate the performance of the PHM model and adaptively adjusts the fault injection strategy and model parameters in the digital twin model based on the evaluation results.

[0057] In some embodiments, a dynamic weight evaluation system is used to quantitatively evaluate the performance of the PHM model, and the fault injection strategy and model parameters in the digital twin model are adaptively adjusted based on the evaluation results, including: After receiving the mixed data stream, the PHM model processes it based on the built-in health status analysis algorithm and outputs analysis results data including fault detection status, diagnosis type and predicted remaining service life. The evaluation and analysis module receives the analysis results data output by the PHM model in real time, compares the analysis results data with the preset fault injection information, and quantifies the performance indicators of the PHM model based on the comparison results. The weights of each performance indicator are dynamically assigned based on the fault type, system operation stage, and system health status vector, and a comprehensive performance score is calculated. The overall performance score is compared with a preset threshold. If the overall performance score is lower than the preset threshold, an adjustment instruction is generated to adaptively adjust the fault injection strategy and model parameters in the digital twin model.

[0058] Specifically, the PHM model receives the mixed data stream generated in S130 and processes it based on the built-in health status analysis algorithm. The algorithm first extracts features from the mixed data stream (such as voltage fluctuation features, pressure change slope, landing gear deflection angle change features, etc., to maintain consistency with the preprocessed parameter features in S110), then identifies the fault type through the fault classification model, calculates the remaining service life in combination with the remaining service life prediction model, and finally outputs complete analysis results data including fault detection status (such as "detected" "not detected"), diagnosis type (such as "severe voltage drop" "minor hydraulic leakage"), predicted remaining service life (such as "120 minutes"), and false alarm indicators (such as "no false alarm" "false alarm"). This ensures that the output information can cover the evaluation dimensions of the core functions of the PHM model.

[0059] The evaluation and analysis module receives the output of the PHM model in real time and compares it precisely with the preset fault injection information. This preset fault injection information originates from the configuration parameters of the fault evolution algorithm in S120, including key information such as fault type, fault occurrence time, fault severity level, and actual remaining lifespan. Based on this comparison result, the core performance indicators of the PHM model are quantitatively calculated: detection latency is the time difference between the actual fault occurrence time and the moment the PHM model first outputs the "detected" status; if it exceeds a preset duration (e.g., 100ms), it is considered that the detection latency exceeds the standard. Diagnostic accuracy is the proportion of cases that correctly diagnose the fault type and severity level out of the total number of fault cases; false alarm cases (i.e., PHM falsely reporting faults when there is no fault, requiring a false alarm rate ≤ 0.1%) must be excluded. Prediction error is |predicted remaining lifespan - actual remaining lifespan| / actual remaining lifespan × 100%; if the error exceeds a preset threshold (e.g., 5%), it requires close attention.

[0060] Specifically, in the dynamic weight allocation stage, weighting rules are strictly formulated based on fault type, system operation stage, and system health state vector: First, faults are classified into Class I (severe faults leading to mission failure, such as complete landing gear jamming) and Class II (moderate faults causing performance degradation, such as minor hydraulic leakage) according to the GJB / Z1391-2006 standard. The basic weight allocation is the Class I fault detection weight. =0.4, diagnostic weight =0.3, Class II fault detection weight =0.3, diagnostic weight =0.4; then, combined with adjustments during the verification phase—in the early verification stage (when the system health is relatively stable and HSI is generally ≥0.7), the focus is on fault detection capabilities, increasing the weight of Type I fault detection to 0.5 and decreasing the prediction weight to 0.2; in the middle verification stage (when the system health fluctuates and HSI is between 0.3 and 0.7), the focus is on diagnostic accuracy, uniformly increasing the diagnostic weight to 0.4-0.5; in the later verification stage (when the system health is poor and HSI ≤0.3), the focus is on remaining lifetime prediction capabilities, increasing the prediction weight to 0.5-0.6; at the same time, if the HSI of a certain component drops significantly (≤0.4), the diagnostic and prediction weights of faults associated with that component are further increased to ensure that the weight allocation accurately matches the actual health risks of the system.

[0061] Specifically, based on the above dynamic weights and quantitative indicators, the formula is adopted: Calculate the overall performance score of the PHM model, where To detect the dynamic weights corresponding to the delay (and) Positive correlation, initial stage =0.5), The dynamic weights corresponding to the diagnostic accuracy (i.e.) Mid-term =0.5), The dynamic weights corresponding to the prediction error (positively correlated with the prediction weights, later...) =0.6), this formula converts performance indicators from different dimensions into a unified score, and the higher the score, the better the performance of the PHM model. The evaluation must consider both specific indicator thresholds and the overall score threshold: the preset thresholds are: Type I fault detection rate ≥99.5%, Type II fault diagnosis accuracy ≥95%, prediction error ≤5%, and the overall score threshold is set at 85 points. If any specific indicator fails to meet the standard or the overall score is below 85 points, the PHM model is deemed to have failed to meet the performance requirements.

[0062] At this point, adjustment instructions need to be generated to adaptively optimize the fault injection strategy and model parameters of the digital twin model: Regarding the fault injection strategy adjustment, for fault types with low scores (such as "intermittent voltage drops"), their injection frequency is automatically increased by 20%-50%, and the reinforcement learning reward function of the fault evolution algorithm in S120 is also adjusted. Coefficient of the term: in, Score the current fault scenario. The average score is the average score of the last 10 similar failure scenarios.

[0063] Will The coefficient of the term was increased from the initial 0.2 to 0.3-0.4, enhancing the sensitivity of the reward function to the performance deviation of the PHM model and guiding the fault evolution path to tilt towards the weak links of the PHM model. In terms of adjusting the parameters of the digital twin model, the influence matrix of the HSI component and the digital twin parameters in S110 was recalibrated. If there is still a deviation after adjusting the internal resistance of the braking system, the influence weight of the voltage health can be finely adjusted from 0.6 to 0.65. At the same time, the feature extraction weight of the PHM model was adjusted (such as increasing the weight of voltage jump characteristics). After all adjustments are completed, 10 consecutive verification cycles need to be started. If the average comprehensive score of the 10 verifications increases by ≥5 points and the specific indicators meet the standards, the adjustment plan is retained. Otherwise, it is reverted to the state before adjustment and the optimization direction is explored again, forming a closed-loop optimization mechanism of "evaluation-adjustment-verification" to ensure that the performance of the PHM model continues to approach the actual application requirements, while making the fault simulation of the digital twin model more in line with the verification pain points of the PHM model.

[0064] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this disclosure is not limited to the described order of actions, because according to this disclosure, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this disclosure.

[0065] The above is an introduction to the method embodiments. The following describes the present disclosure solution further through device embodiments.

[0066] Figure 2 A structural diagram of a digital twin-based PHM model verification and management device provided in an embodiment of this disclosure is shown, as follows: Figure 2 As shown, a PHM model verification and management device 200 based on digital twins may include: The digital twin module 210 is used to receive and parse real-time operating data of the physical system through the digital twin model, and dynamically adjust the parameters of the digital twin model based on the system health status vector. A pre-defined fault evolution algorithm is injected into the digital twin model after parameter adjustment to generate a highly simulated fault data sequence containing both continuous and intermittent fault characteristics.

[0067] PHM Model 220 is used to run the PHM model in a pre-built virtual verification environment, couple it with the digital twin model, and receive real-time simulation data from the digital twin model.

[0068] The feedback optimization module 230 is used to quantitatively evaluate the performance of the PHM model using a dynamic weight evaluation system, and adaptively adjust the fault injection strategy and model parameters in the digital twin model based on the evaluation results.

[0069] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0070] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0071] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0072] Figure 3 A schematic block diagram of an electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0073] Electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes according to a computer program stored in ROM 302 or a computer program loaded into RAM 303 from storage unit 308. RAM 303 can also store various programs and data required for the operation of electronic device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via bus 304. I / O interface 305 is also connected to bus 304.

[0074] Multiple components in electronic device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of displays, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows electronic device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0075] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as method 100. For example, in some embodiments, method 100 may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of method 100 described above may be performed. Alternatively, in other embodiments, the computing unit 301 may be configured to perform method 100 by any other suitable means (e.g., by means of firmware).

[0076] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0077] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0078] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0079] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including voice input, speech input, or tactile input).

[0080] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0081] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0082] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this disclosure can be achieved, and this is not limited herein.

[0083] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A PHM model validation and management method based on digital twins, characterized in that, include: The system receives and parses real-time operational data of the physical system through a digital twin model, and dynamically adjusts the parameters of the digital twin model based on the system health status vector. A pre-defined fault evolution algorithm is injected into the digital twin model after parameter adjustment to generate a highly simulated fault data sequence containing both continuous and intermittent fault characteristics. Run the PHM model in a pre-built virtual verification environment, couple it with the digital twin model, and receive and run real-time simulation data from the digital twin model. A dynamic weighted evaluation system is used to quantitatively evaluate the performance of the PHM model, and the fault injection strategy and model parameters in the digital twin model are adaptively adjusted based on the evaluation results.

2. The method according to claim 1, characterized in that, The system health status vector is composed of multiple health status index components, and the health status of the health status index components is calculated using real-time operating data of the physical system.

3. The method according to claim 2, characterized in that, The dynamic adjustment of the digital twin model parameters based on the system health state vector includes: When the health status of at least two health status index components is simultaneously lower than the first preset threshold, the comprehensive adjustment coefficient of the parameters is calculated based on the weight superposition and total amount capping mechanism, and the adjustment priority is allocated according to the safety criticality of the parameters. The health status changes of each health status index component are periodically calculated. If the health status change of any health status index component is greater than or equal to the second preset threshold, a parameter adjustment command is triggered. If the change in health status of all health status index components is less than the second preset threshold, then parameter adjustment is paused. After the parameters are updated, the deviation between the output of the digital twin model and the parameters of the physical system is verified, and the updated parameters or the state before adjustment is decided based on the verification results.

4. The method according to claim 3, characterized in that, The comprehensive adjustment coefficient for the parameters calculated based on the weighted superposition and total amount capping mechanism includes: If multiple health status index components adjust the same digital twin parameter in the same direction, the comprehensive adjustment coefficient is calculated according to the influence weight of each health status index component. If multiple health status index components adjust the same digital twin parameter in opposite directions, the adjustment direction of the health status index component with the greater influence weight shall prevail, and the adjustment magnitude of the health status index component with the smaller influence weight shall be superimposed at a predetermined ratio. When the overall adjustment exceeds the third preset threshold, the adjustment of safety-critical parameters is prioritized, and the adjustment amount of non-safety-critical parameters is adjusted based on the influence weight.

5. The method according to claim 4, characterized in that, The step of injecting a pre-defined fault evolution algorithm into the adjusted digital twin model to generate a highly simulated fault data sequence containing both continuous and intermittent fault characteristics includes: Inject a fault evolution algorithm based on a reinforcement learning model into the adjusted digital twin model; Construct a state vector with the deviation between the output parameters of the digital twin model and the parameters of the physical system as the core indicator and the system health state vector as the core, and define the action space; The fault evolution rate is dynamically controlled based on the real-time calculated system performance degradation gradient, generating a highly simulated fault data sequence that includes both continuous and intermittent fault characteristics.

6. The method according to claim 5, characterized in that, The process of dynamically controlling the fault evolution rate based on the real-time calculated system performance degradation gradient to generate a highly simulated fault data sequence containing continuous fault characteristics includes: The system calculates the probability of performance degradation in real time. If the degradation probability is greater than or equal to a fourth preset threshold, a fault acceleration mechanism is triggered to adjust the fault evolution rate. Under the dynamic control of the fault evolution rate, the simulation of the fault degree gradually develops from the initial value to the complete path of system failure, generating a continuous fault data sequence.

7. The method according to claim 6, characterized in that, The generation of highly simulated fault data sequences containing intermittent fault characteristics includes: The intermittent behavior of faults is simulated based on a multi-state Markov jump model, and the fault states are extended to normal state, mild abnormal state, severe abnormal state and fault recovery state. Establish a mapping relationship between the fault state and the fluctuation of system parameters, define differentiated fluctuation ranges for system parameters under different states, and realize dynamic switching between states through a preset state transition probability matrix; Based on the multi-state Markov transition model and the mapping relationship between the fault state and parameter fluctuations, the cyclical transfer process of the fault between normal, mildly abnormal, severely abnormal, and fault recovery states is simulated to generate an intermittent fault data sequence containing parameter fluctuation sequences under each state.

8. The method according to claim 7, characterized in that, The virtual verification environment includes a data injection module, a PHM interface module, and an evaluation and analysis module; wherein... The data injection module is used to take the highly simulated fault data sequence as input, mix the fault data stream and the normal data stream in a proportional manner, and input the mixed data stream into the PHM model in real time. The PHM interface module is used to control API latency and ensure the real-time data interaction between the PHM model and the virtual environment. The evaluation and analysis module is used to compare the analysis results output by the PHM model with the preset fault injection information in real time, quantify and calculate the performance indicators of detection delay, diagnostic accuracy and prediction error, and support concurrent processing of multi-parameter monitoring scenarios.

9. The method according to claim 8, characterized in that, The step of running the PHM model in a pre-built virtual verification environment, coupling it with the digital twin model, and receiving and running real-time simulation data from the digital twin model includes: The data injection module mixes the highly simulated fault data sequence with the normal operation data of the physical system according to a preset ratio to generate a mixed data stream. The mixed data stream is input into the PHM model in real time through the PHM interface module.

10. The method according to claim 9, characterized in that, The process of using a dynamic weighted evaluation system to quantitatively evaluate the performance of the PHM model and adaptively adjusting the fault injection strategy and model parameters in the digital twin model based on the evaluation results includes: After receiving the mixed data stream, the PHM model processes it based on the built-in health status analysis algorithm and outputs analysis results data including fault detection status, diagnosis type and predicted remaining service life. The evaluation and analysis module receives the analysis results data output by the PHM model in real time, compares the analysis results data with the preset fault injection information, and quantitatively calculates the performance indicators of the PHM model based on the comparison results. The weights of each performance indicator are dynamically assigned based on the fault type, system operation stage, and system health status vector, and a comprehensive performance score is calculated. The overall performance score is compared with a preset threshold. If the overall performance score is lower than the preset threshold, an adjustment instruction is generated to adaptively adjust the fault injection strategy and model parameters in the digital twin model.

Images