Locomotive engine state synchronization and self-calibration method, system, equipment and medium

Abstract

The invention discloses a locomotive engine state synchronization and self-calibration method, system and device and a medium, and the method comprises the steps: real-time synchronization and data preprocessing: receiving multi-mode original data of a locomotive engine, and after preprocessing, driving a multi-domain coupling mechanism model in a digital twin to operate for a simulation step length; multi-modal weighted fusion synchronization error calculation: respectively calculating a vibration error and a thermodynamic error, and obtaining a comprehensive synchronization error through weight weighted fusion; intelligent self-calibration triggering: setting a threshold triggering condition, a continuous overrun triggering condition and a synchronous error change rate triggering condition, and triggering a parameter self-calibration instruction if any triggering condition is met; and automatic parameter calibration and security updating: starting an optimization engine to construct an optimization model with the purpose of minimizing the comprehensive synchronization error, obtaining an optimal parameter set through an optimization algorithm, updating the optimal parameter set to a multi-domain coupling mechanism model, and performing verification after updating. According to the invention, efficient and intelligent locomotive engine state synchronization and self-calibration can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of locomotive engine state monitoring, and particularly relates to a locomotive engine state synchronization and self-calibration method, system, device and medium. BACKGROUND

[0002] With the development of locomotive engines towards high efficiency and intelligence, digital twin technology has become the core support for state monitoring and health management due to its precise mapping capability of physical entities. However, in the current engine management system based on digital twin, there is a common problem of desynchronization between static models and dynamic entities. The physical characteristics of real engines will slowly and continuously drift due to factors such as wear, aging, and carbon deposition during long-term service, while the key parameters of traditional simulation models remain fixed after initial setting and cannot be dynamically adjusted to follow the changes of physical entities. At the same time, the boundary conditions on which the model runs, such as environmental factors and operating loads, cannot be accurately matched with the complex and variable actual working conditions in real time. This desynchronization makes the output of the digital twin gradually deviate from the real state of the physical entity, and the simulation accuracy continuously decreases with the running time, eventually leading to the loss of the core value of the twin as a state monitoring benchmark, and the inability to provide reliable basis for engine health assessment.

[0003] On this basis, although a large number of sensors are deployed in existing systems to collect multi-dimensional real-time data such as vibration, temperature, and pressure, the application level of these data is relatively shallow and their core value is not fully utilized. Most systems only display or compare the measured data of sensors and the simulation data of the twin in time sequence, and rely on manual judgment to determine whether there is an abnormality, lacking deep mining and intelligent application of data. More importantly, the data flow is mostly one-way transmission from the physical entity to the virtual model, and the virtual model is only a passive data receiver, rather than a living model that can evolve itself using data. It fails to form a closed-loop mechanism of perception-decision-adjustment, and cannot automatically correct model errors through data, making it difficult to improve the intelligent level of the system.

[0004] In addition, existing technologies lack efficient and reliable automatic calibration mechanisms, and the adjustment of model parameters often relies on manual intervention by engineers. This manual calibration method not only consumes time and effort, but also has a strong dependence on the personal experience of experts, making it difficult to achieve real-time calibration to cope with dynamic changes in working conditions. At the same time, the trigger mechanism for calibration is relatively simple, mostly using a single threshold trigger method, which is easily affected by data noise to cause false triggering, or misses critical calibration opportunities due to insufficient sensitivity, and cannot adapt to the complex and variable operating conditions of the engine. More importantly, there is a lack of effective safety verification and rollback mechanism during parameter updating, which may cause simulation interruption or introduce incorrect parameters to cause model divergence, posing a high risk to system operation.

[0005] In the related art field, although some patent technologies have explored engine fault diagnosis or troubleshooting, there are still obvious limitations.

[0006] For example, CN110378034A discloses a locomotive engine fault diagnosis method, including the following steps: collecting engine state parameter information; performing state classification on the engine state parameters; and giving the fault root cause causing the state classification according to a pre-prepared fusion diagnosis model. The locomotive engine fault diagnosis method establishes a fusion diagnosis model, combines weight coefficients to comprehensively judge the fault root cause, and focuses on the identification of the fault cause. However, it does not focus on the synchronization problem between the digital twin model and the physical entity, nor does it design a model parameter self-calibration mechanism. Therefore, it cannot solve the precision decay problem caused by the model drifting over time. The sensor data is only used for auxiliary judgment of fault diagnosis, and cannot drive the model to realize dynamic updating.

[0007] CN117150884A discloses an engine fault troubleshooting method and system based on digital twinning, including: establishing an engine digital twin model according to engine normal working condition historical operation data, segmenting the engine digital twin model according to functional characteristics to obtain segmented different characteristic digital twin models; establishing a first fault diagnosis model according to different characteristic digital twin models, combining engine fault working condition historical operation data and fault information; performing fault traversal on different characteristic digital twin models, combining machine learning to establish a second fault diagnosis model; and completing engine fault troubleshooting according to the first fault diagnosis model and the second fault diagnosis model. The engine fault troubleshooting method and system realize fault troubleshooting by segmenting the digital twin model and establishing a fault diagnosis model. However, its core goal is to identify and locate faults, and it does not involve real-time synchronization and automatic calibration of digital twins. The model parameters are still in a relatively fixed state, making it difficult to adapt to dynamic changes in engine physical properties. At the same time, this technology does not form a data-driven model optimization closed loop, and lacks a safety guarantee mechanism during calibration. Therefore, it cannot meet the requirements of engine long-term and stable operation for the precision and reliability of the digital twin model.

[0008] In summary, the deficiencies of these existing technologies further highlight the necessity and urgency of constructing an efficient, intelligent, and safe locomotive engine state synchronization and self-calibration mechanism. SUMMARY

[0009] To solve the above problems, the application provides a locomotive engine state synchronization and self-calibration method, system, device and medium, which takes the multi-modal weighted fusion synchronization error as the core index, automatically starts the optimization algorithm to adjust the key parameters of the digital twin in reverse when the error exceeds the limit through the multi-condition intelligent triggering mechanism, and ensures the reliability of the calibration process through the safe update and rollback mechanism, finally realizes the self-calibration and long-term high fidelity of the digital twin.

[0010] The technical solutions adopted by the application are as follows: A locomotive engine state synchronization and self-calibration method, comprising: Real-time synchronization and data preprocessing: receiving the multi-modal original data of the locomotive engine collected by the physical sensor, and after preprocessing, the data is used as boundary conditions or input excitation to drive the multi-domain coupled mechanism model in the digital twin to run for one simulation step; Multi-modal weighted fusion synchronization error calculation: respectively calculating the vibration error and the thermodynamic error, and then obtaining the comprehensive synchronization error through weighted fusion; Intelligent self-calibration trigger: setting threshold trigger conditions, continuous over-limit trigger conditions and synchronization error change rate trigger conditions, and triggering the parameter self-calibration instruction when any trigger condition is met; Parameter automatic calibration and safe update: after receiving the parameter self-calibration instruction, starting the optimization engine to build an optimization model with the objective of minimizing the comprehensive synchronization error, obtaining the optimal parameter set through the optimization algorithm, and updating the multi-domain coupled mechanism model in the digital twin when the engine working condition is stable after effectiveness verification, and performing post-update verification.

[0011] Further, the preprocessing of the multi-modal original data of the locomotive engine includes noise filtering, invalid value elimination and unit unification processing.

[0012] Further, the calculation method of the vibration error and the thermodynamic error comprises: calculating the root mean square error of the measured vibration signal and the simulated vibration signal in the same measuring point and the same time period, and then normalizing the root mean square error by dividing the maximum amplitude of the measured vibration signal, thereby obtaining the vibration error; calculating the average absolute percentage error of the measured temperature and the simulated temperature of the plurality of thermocouple measuring points, the average absolute percentage error being the average value of the ratio of the absolute value of the temperature difference of each measuring point to the corresponding measured temperature, thereby obtaining the thermodynamic error.

[0013] Further, the threshold trigger condition includes that the comprehensive synchronization error is greater than a first preset threshold; the continuous over-limit trigger condition includes that the comprehensive synchronization error is greater than a second preset threshold and the duration exceeds a preset holding time; and the synchronization error change rate trigger condition includes that the change rate of the comprehensive synchronization error is greater than a preset change rate threshold.

[0014] Further, the optimization algorithm includes a particle swarm optimization algorithm, and the parameter types in the optimal parameter set include a damping coefficient and a thermal convection coefficient.

[0015] Further, the multi-domain coupled mechanism model in the digital twin is updated at a stable engine operating condition after the effectiveness verification, including: in the simulation environment, verifying whether the optimal parameter set can significantly reduce the comprehensive synchronization error by using historical data, and if so, under the idle state of the engine, the multi-domain coupled mechanism model in the digital twin is updated according to the optimal parameter set by using the write-ahead backup or atomic operation.

[0016] Further, the post-update verification includes: monitoring the comprehensive synchronization error in a period of time after the update, and if the comprehensive synchronization error compared with the comprehensive synchronization error before calibration does not decrease by more than a preset value or the error increases, it is determined that the calibration fails, and the last set of effective parameters is automatically rolled back.

[0017] A locomotive engine state synchronization and self-calibration system, comprising: A real-time synchronization and data preprocessing module configured to receive multi-modal raw data of a locomotive engine collected by a physical sensor, and after preprocessing, the multi-modal raw data is used as a boundary condition or an input excitation to drive a multi-domain coupled mechanism model in a digital twin to run for one simulation step; A multi-modal weighted fusion synchronization error calculation module configured to calculate vibration errors and thermodynamic errors, and then obtain a comprehensive synchronization error by weighted fusion through weights; An intelligent self-calibration triggering module configured to set threshold triggering conditions, continuous over-limit triggering conditions, and synchronization error change rate triggering conditions, and trigger a parameter self-calibration instruction when any triggering condition is met; A parameter automatic calibration and safe update module configured to, after receiving the parameter self-calibration instruction, start an optimization engine to build an optimization model with the objective of minimizing the comprehensive synchronization error, obtain an optimal parameter set through an optimization algorithm, update the multi-domain coupled mechanism model in the digital twin to a stable engine operating condition after effectiveness verification, and perform post-update verification.

[0018] A computer device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the locomotive engine state synchronization and self-calibration method when executing the computer program.

[0019] A computer readable storage medium, storing a computer program, the computer program is executed by a processor to implement the locomotive engine state synchronization and self-calibration method.

[0020] The beneficial effects of the present application are: 1. The application takes the multi-modal weighted fusion synchronous error as the core index, through the multi-condition intelligent triggering mechanism, when the error is out of limit, the optimization algorithm is automatically started to adjust the key parameters of the digital twin, and the safety update and rollback mechanism is used to ensure the reliability of the calibration process, finally realizing the self-calibration and long-term high accuracy of the digital twin.

[0021] 2. In the real-time synchronization and data preprocessing link, the application does not simply use sensor data as the basis for comparison, but directly converts the preprocessed multi-modal data into real-time boundary conditions and input excitations of the multi-domain coupled mechanism model in the digital twin, so that the model is no longer a passive static framework that receives data, but an active model that dynamically responds to the running state of the physical engine, thus reducing the real-time deviation between the virtual and physical entities, and enabling the digital twin to accurately map the real running state of the engine at all times, completely changing the situation of past model accuracy decay over time.

[0022] 3. Based on the comprehensive synchronous error formed by multi-modal weighted fusion, the application not only integrates the state information of key dimensions during engine operation, but also dynamically adjusts the influence weight of each dimension according to the monitoring focus under different working conditions through a flexible weight configuration mechanism, so that the error judgment is comprehensive and targeted, avoiding the one-sidedness of single-dimensional error evaluation, making the health status evaluation more in line with actual operation requirements, and providing accurate and reliable core basis for subsequent calibration. The intelligent self-calibration triggering logic, by integrating multiple judgment conditions, not only avoids the false triggering problem caused by data noise interference in traditional single threshold triggering, but also solves the calibration omission problem caused by insufficient sensitivity, and can adaptively identify the calibration opportunity according to the complex changes of engine working conditions, ensuring that the calibration action neither excessively frequent to affect system stability, nor delays the correction opportunity to cause error expansion, greatly improving the adaptability and reliability of the self-calibration system.

[0023] 4. In the parameter automatic calibration and safety update link, the optimization engine minimizes the comprehensive synchronous error to optimize the parameters, ensuring that the optimal parameter set obtained can accurately match the physical characteristics of the current engine, and the design of effective verification and working condition stable time update further avoids the interference of parameter update on simulation continuity, and in combination with the perfect rollback mechanism, completely solves the problem of time-consuming and labor-intensive traditional manual calibration and dependence on experience, while eliminating the model divergence risk caused by parameter update, ensuring the stability and accuracy of long-term continuous operation of the digital twin.

[0024] 5、Compared with the prior art of CN110378034A, CN117150884A and the like, the present application not only fills the gap of the prior art in real-time synchronization and self-calibration of digital twin models, but also fully excavates the value of sensor data through data-driven closed-loop design, realizes the transformation of the model from passive monitoring to active evolution, and at the same time, with a perfect safety mechanism to ensure the reliability of the calibration process, so that the digital twin can not only be used for fault diagnosis and troubleshooting, but also can be used as a precise benchmark for engine state monitoring for a long time, providing more comprehensive, more sustainable and more reliable technical support for engine health management. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is a locomotive engine state synchronization and self-calibration method flow chart of embodiment 1 of the present application.

[0026] Figure 2 is a locomotive engine state synchronization and self-calibration method flow chart of embodiment 2 of the present application.

[0027] Figure 3 is a locomotive engine state synchronization and self-calibration system principle diagram of embodiment 3 of the present application. DETAILED DESCRIPTION

[0028] In order to have a more clear understanding of the technical features, purposes and effects of the present application, the specific embodiments of the present application will be described. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application, that is, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0029] Embodiment 1 As shown in Figure 1 , the present embodiment provides a locomotive engine state synchronization and self-calibration method, comprising: Real-time synchronization and data preprocessing: receiving the multi-modal original data of the locomotive engine collected by the physical sensor, after preprocessing, as the boundary condition or input excitation, driving the multi-domain coupled mechanism model in the digital twin to run one simulation step in real time; Multi-modal weighted fusion synchronization error calculation: calculate the vibration error and the thermodynamic error respectively, and then get the comprehensive synchronization error by weighted fusion through the weight; Intelligent self-calibration trigger: set threshold trigger condition, continuous overrun trigger condition and synchronization error change rate trigger condition, and trigger the parameter self-calibration instruction when any trigger condition is met; Parameter automatic calibration and safe update: after receiving the parameter self-calibration instruction, the optimization engine is started to build an optimization model with the goal of minimizing the integrated synchronization error, and the optimal parameter set is obtained through the optimization algorithm. After effectiveness verification, the parameter set is updated to the multi-domain coupled mechanism model in the digital twin at a stable engine operating condition, and the updated model is verified.

[0030] It should be noted that the method ensures real-time correspondence between the digital twin and the physical engine state by real-time synchronization of data and driving of model operation; comprehensively reflects the synchronization deviation between the two through multi-modal error fusion calculation; the multi-trigger condition design ensures the timeliness and accuracy of self-calibration; and the parameter calibration and safe update process realizes dynamic optimization of model parameters, improves the modeling accuracy and reliability of the digital twin, and thus provides stronger support for state monitoring and fault diagnosis of the locomotive engine.

[0031] Preferably, the preprocessing of the multi-modal raw data of the locomotive engine includes noise filtering, invalid value elimination, and unit unification processing. Specifically, after obtaining the multi-modal raw data of the locomotive engine, first, an appropriate filtering technique is used to suppress and remove noise signals in the data, reducing the influence of irrelevant interference on subsequent processing; then, invalid information in the data is identified, including data beyond a reasonable range, data lacking key features, etc., and is eliminated from the original data set; finally, the data after noise reduction and invalid value elimination is processed to unify different sources and units of data into consistent measurement standards, ensuring the consistency and comparability of the data.

[0032] It should be noted that through noise filtering processing, the purity of the original data is improved, and the interference of noise on the simulation results of the model is reduced; invalid value elimination avoids model running deviation caused by abnormal data, ensuring data quality; unit unification processing eliminates calculation errors caused by data format differences, laying a solid foundation for accurate operation and error calculation of the subsequent multi-domain coupled mechanism model.

[0033] Preferably, the calculation method of the vibration error and the thermodynamic error includes: calculating the root mean square error of the measured vibration signal and the simulated vibration signal at the same measuring point and in the same time period, and then normalizing the root mean square error by dividing it by the maximum amplitude of the measured vibration signal to obtain the vibration error; calculating the average absolute percentage error of the measured temperature and the simulated temperature of multiple thermocouple measuring points, the average absolute percentage error being the average value of the ratio of the absolute value of the temperature difference of each measuring point to the corresponding measured temperature, thereby obtaining the thermodynamic error.

[0034] Specifically, when calculating the vibration error, the same measurement point position and time interval are determined first, and the measured vibration signal and the simulated vibration signal in the range are extracted. The root mean square error of the two types of signals is calculated to reflect the dispersion degree of the two types of signals. Then, the root mean square error is subjected to ratio operation with the maximum amplitude of the measured vibration signal to complete the normalization processing, and finally the vibration error is obtained. When calculating the thermodynamic error, multiple thermocouple measurement points are selected, and the measured temperature and the simulated temperature of each measurement point are obtained respectively. The absolute value of the temperature difference of each measurement point is calculated, and then the absolute value is subjected to ratio operation with the measured temperature of the corresponding measurement point. Finally, the average value of the ratio results of all measurement points is taken to obtain the thermodynamic error.

[0035] It should be noted that the calculation of the vibration error eliminates the influence of signal amplitude difference on error evaluation through normalization processing, making the error result more comparable. The thermodynamic error is calculated by using the average absolute percentage error, which can directly reflect the relative deviation between the simulated temperature and the measured temperature, and takes into account the error of multiple measurement points. The calculation methods of the two types of errors are scientific and reasonable, providing a guarantee for the accurate acquisition of subsequent comprehensive synchronization error.

[0036] Preferably, the threshold trigger condition includes that the comprehensive synchronization error is greater than a first preset threshold value; the continuous overrun trigger condition includes that the comprehensive synchronization error is greater than a second preset threshold value and the duration exceeds a preset retention time; and the synchronization error change rate trigger condition includes that the change rate of the comprehensive synchronization error is greater than a preset change rate threshold value.

[0037] Specifically, after setting the trigger conditions, the calculated comprehensive synchronization error is monitored in real time to determine whether it is greater than the first preset threshold value. If it is satisfied, the threshold trigger condition is triggered. At the same time, whether the comprehensive synchronization error is greater than the second preset threshold value is monitored, and the duration of this state is recorded. When the duration exceeds the preset retention time, the continuous overrun trigger condition is triggered. In addition, the change rate of the comprehensive synchronization error is calculated in real time. The change rate is obtained by the ratio of the difference value of the comprehensive synchronization error of adjacent time periods to the time difference. Whether the change rate is greater than the preset change rate threshold value is determined. If it is satisfied, the synchronization error change rate trigger condition is triggered. When any one of the above three trigger conditions is satisfied, the parameter self-calibration instruction is immediately triggered.

[0038] It should be noted that the threshold trigger condition can quickly respond to large synchronization error and start calibration in time. The continuous overrun trigger condition avoids false triggering caused by short-term fluctuations and ensures that the calibration is aimed at the persistent deviation problem. The synchronization error change rate trigger condition can capture the situation of rapid error change and avoid the risk of further expansion of error in advance. The three types of trigger conditions complement each other and cover comprehensively, effectively guaranteeing the timeliness, accuracy and rationality of self-calibration.

[0039] Preferably, the optimization algorithm comprises a particle swarm optimization algorithm, and the parameter types in the optimal parameter set comprise a damping coefficient and a thermal convection coefficient. Specifically, after starting the optimization engine, the particle swarm optimization algorithm is selected as the core optimization algorithm to build the corresponding optimization model, with the objective of minimizing the comprehensive synchronization error. In the optimization process, the damping coefficient and the thermal convection coefficient are taken as the key optimization parameters and are included in the search range of the optimal parameter set. Through the iterative search mechanism of the particle swarm optimization algorithm, the parameter values are continuously adjusted, and the parameter combination that can minimize the comprehensive synchronization error is gradually found, so as to finally determine the optimal parameter set comprising the damping coefficient and the thermal convection coefficient.

[0040] It should be noted that the particle swarm optimization algorithm has the characteristics of fast convergence speed and strong global search ability, and can efficiently find the optimal solution in the parameter space to improve the efficiency of parameter calibration. The damping coefficient and the thermal convection coefficient are key parameters that affect the simulation accuracy of the multi-domain coupled mechanism model of the locomotive engine. Optimizing these two types of parameters can accurately improve the synchronization deviation between the model and the actual engine and significantly improve the modeling accuracy of the digital twin.

[0041] Preferably, the multi-domain coupled mechanism model in the digital twin is updated at the stable engine operating condition after effectiveness verification, comprising: verifying in the simulation environment whether the optimal parameter set can significantly reduce the comprehensive synchronization error, and if so, updating the multi-domain coupled mechanism model in the digital twin according to the optimal parameter set under the engine idle state by using the write-ahead backup or atomic operation method.

[0042] Specifically, after obtaining the optimal parameter set, the historical running data of the engine is called in the simulation environment, the optimal parameter set is substituted into the multi-domain coupled mechanism model, the model is run and the corresponding comprehensive synchronization error is calculated, and the optimal parameter set is verified whether it can significantly reduce the comprehensive synchronization error by comparing with the previous comprehensive synchronization error. If the verification is passed, the running condition of the engine is continuously monitored, and when the engine enters the idle state, i.e., the stable operating condition, the write-ahead backup or atomic operation method is used to protect the current parameters of the model, and then the parameter values corresponding to the optimal parameter set are written into the multi-domain coupled mechanism model to complete the update of the model parameters.

[0043] It should be noted that the effectiveness verification by using the historical data ensures the practicality and effectiveness of the optimal parameter set, avoiding the impact of invalid parameter update on the model. The selection of the stable operating condition of the engine idle state for updating reduces the interference of operating condition fluctuation on the parameter update process, ensuring the stability of the update operation. The use of write-ahead backup or atomic operation provides safety protection for parameter update, preventing the loss or disorder of model parameters caused by abnormality in the update process.

[0044] Preferably, the post-update verification comprises: monitoring the integrated synchronization error in a period of time after the update, and determining that the calibration fails and automatically rolling back to the last set of valid parameters if the integrated synchronization error does not decrease by more than a preset value or the error increases compared with the integrated synchronization error before the calibration.

[0045] Specifically, after completing the multi-domain coupling mechanism model parameter update, the integrated synchronization error generated by the model running in a subsequent period of time is continuously monitored, and the integrated synchronization error data before the calibration is called to calculate the decrease amplitude of the integrated synchronization error after the update; the decrease amplitude is compared with a preset standard, and if the decrease amplitude does not reach the preset standard or the integrated synchronization error after the update increases, it is determined that the parameter calibration fails this time; at this time, the system automatically starts the parameter rollback program to restore the model parameters to the last set of verified valid parameters.

[0046] It should be noted that the post-update verification process can timely verify the actual effect of the parameter calibration, avoid the influence of invalid parameters on the accuracy of the digital twin for a long time, and provide a remedy for the calibration failure, thereby guaranteeing the continuity and reliability of the model running and preventing the system performance from being degraded due to improper calibration.

[0047] Correspondingly, the embodiment also provides a locomotive engine state synchronization and self-calibration system, comprising: The real-time synchronization and data preprocessing module is configured to receive the multi-modal raw data of the locomotive engine collected by the physical sensor, and pre-process the multi-modal raw data as boundary conditions or input excitations to drive the multi-domain coupling mechanism model in the digital twin to run for one simulation step; The multi-modal weighted fusion integrated synchronization error calculation module is configured to calculate the vibration error and the thermodynamic error, and then obtain the integrated synchronization error by weighted fusion through the weights; The intelligent self-calibration triggering module is configured to set threshold triggering conditions, continuous over-limit triggering conditions and synchronization error change rate triggering conditions, and trigger the parameter self-calibration instruction if any triggering condition is met; The parameter automatic calibration and safe update module is configured to, after receiving the parameter self-calibration instruction, start the optimization engine to build an optimization model with the objective of minimizing the integrated synchronization error, obtain the optimal parameter set through the optimization algorithm, update the multi-domain coupling mechanism model in the digital twin at the stable engine working condition after the effectiveness verification, and perform post-update verification.

[0048] Specifically, the real-time synchronization and data preprocessing module continuously receives the multi-modal raw data of the locomotive engine transmitted by the physical sensor, pre-processes the data, and then transmits the pre-processed data as boundary conditions or input excitations to the multi-domain coupling mechanism model of the digital twin to drive the model to complete one simulation step of running.

[0049] The multi-modal weighted fusion synchronous error calculation module processes the relevant data to obtain vibration errors and thermodynamic errors according to a preset calculation method, and then performs a weighted fusion operation on the two types of errors according to a preset weight rule, and outputs a comprehensive synchronous error.

[0050] The intelligent self-calibration trigger module pre-configures three types of trigger conditions, receives the comprehensive synchronous error data in real time and performs condition judgment, and once any trigger condition is met, generates and sends a parameter self-calibration instruction immediately.

[0051] After receiving the self-calibration instruction, the parameter automatic calibration and safety update module starts the optimization engine to build an optimization model, searches for an optimal parameter set using an optimization algorithm, verifies the effectiveness of the optimal parameter set, and when the engine operating condition is stable, updates the parameters to the multi-domain coupled mechanism model, and performs verification after updating.

[0052] It should be noted that the modules in the system have clear division of labor and work together. The real-time synchronization and data preprocessing module provides high-quality input data for the system, the multi-modal weighted fusion synchronous error calculation module accurately feeds back the synchronization deviation, the intelligent self-calibration trigger module ensures that the calibration is started in time, and the parameter automatic calibration and safety update module realizes parameter optimization and safety update. The entire system architecture is complete and the logic is clear, which can efficiently realize the locomotive engine state synchronization and self-calibration function, improve the state consistency of the digital twin and the physical engine, and provide reliable support for the operation and management of the engine.

[0053] Embodiment 2 The embodiment provides a locomotive engine state synchronization and self-calibration method, which takes multi-modal weighted fusion synchronous error as the core index, automatically starts the optimization algorithm to adjust the key parameters of the digital twin in reverse when the error is out of limit through a multi-condition intelligent trigger mechanism, and ensures the reliability of the calibration process through a safety update and rollback mechanism, and finally realizes the self-calibration and long-term high-fidelity of the digital twin.

[0054] As shown in Figure 2 The locomotive engine state synchronization and self-calibration method of the embodiment includes the following steps: S1. Real-time synchronization and data preprocessing Receive multi-modal raw data collected by physical sensors at a fixed period T; preprocess the data, including noise filtering, invalid value elimination, and unit unification; and use the preprocessed physical data as boundary conditions or input excitation to drive the multi-domain coupled mechanism model in the digital twin to run for one simulation step.

[0055] S2. Multi-modal weighted fusion synchronous error calculation The synchronous error defined in the embodiment Not a single indicator of error, but a comprehensive, dimensionless health status index, the calculation method is as follows: Sub-modal error calculation: Vibration error : Calculate the root mean square error (RMSE) of the measured and simulated vibration signals at the same measuring point and time period, and perform normalization processing (divide by the maximum amplitude of the measured signal) to eliminate the dimension effect.

[0056]

[0057] Thermodynamic error : Calculate the average absolute percentage error (MAPE) of the temperature values of multiple thermocouple measuring points and the simulated temperature values.

[0058]

[0059] Weighted fusion into comprehensive synchronous error :

[0060] Among them, the weight , can be dynamically configured according to the focus of fault diagnosis, and the sum of the two is 1.

[0061] S3. Intelligent self-calibration trigger Set up multi-condition intelligent trigger logic, including: Condition one (threshold trigger): ; Condition two (continuous over-limit trigger): and duration > , this condition is used to capture slow performance degradation; Condition three (change rate of synchronous error trigger): , this condition is used to quickly respond to sudden failures.

[0062] If any of the conditions are met, the parameter self-calibration process is automatically triggered.

[0063] S4. Parameter automatic calibration and safe update After receiving the trigger instruction, the parameter self-calibration module starts the optimization engine.

[0064] Optimization model: construct the objective function: , where is the parameter vector to be calibrated (such as damping coefficient, heat convection coefficient).

[0065] Optimization algorithm: adopt particle swarm optimization algorithm, because it is suitable for multi-peak, nonlinear problems, and does not require gradient information.

[0066] Algorithm configuration: the number of particles can be set to 20-50; the maximum number of iterations can be set to 50-100; for each parameter Set a reasonable search range based on physical meaning.

[0067] Security update mechanism: Optimization to get the optimal parameter set After that, it is not immediately forced to write into the running real-time model.

[0068] The system first verifies the effectiveness of the historical data in the simulation environment , that is, to verify whether it is significantly reduced.

[0069] After verification, select the most stable time of engine working condition (such as idle state), through write-ahead backup, atomic operation, seamless update of new parameters to online running digital twin model, ensure continuous operation of model without interruption.

[0070] Verification after update: after the update is completed, the system automatically monitors the synchronization error in the subsequent period of time . If the intersection of the error before calibration does not exceed the preset value or the error increases, it is determined that this calibration fails, and automatically rolls back to the last set of valid parameters, and sends an alarm for manual intervention.

[0071] In summary, the embodiment proposes a multi-modal weighted fusion comprehensive synchronization error as the core health indicator, which integrates multi-dimensional information such as vibration and temperature, and can dynamically adjust the diagnostic tendency through weight configuration. The embodiment designs a set of multi-condition intelligent triggering logic, which integrates threshold, duration and change rate three judgment conditions, to ensure the sensitivity, reliability and efficiency of the self-calibration system triggering mechanism. The embodiment establishes a parameter update process including safety verification and rollback mechanism, to ensure the safety and robustness of the calibration process, and to ensure the stability of the continuous operation of the digital twin.

[0072] Embodiment 3 Based on embodiment 2: Figure 3 As shown in the figure, the embodiment provides a locomotive engine state synchronization and self-calibration system, which includes a multi-modal sensor network, a synchronous data acquisition system and a hierarchical computing and communication platform, which can ensure the timeliness, synchronization and processing capacity of data, as follows.

[0073] (1) Multi-modal sensor network deployment and signal conditioning 1) Vibration monitoring: Piezoelectric acceleration sensors are placed at key locations on the engine body. Specifically, two sensors are installed on the side of each of the 3rd and 7th cylinders in both the left and right cylinder heads to monitor the cylinder head vibration caused by combustion excitation; one sensor is installed on the main bearing seat and the free end of the engine output shaft to monitor the torsional and lateral vibrations of the entire machine. All sensors are rigidly fixed to the measurement points through magnetic bases or threaded connections to ensure measurement frequency response characteristics. The sensor signals are connected to a charge amplifier through low-noise shielded cables for signal conditioning.

[0074] 2) Temperature monitoring: High-temperature armored thermocouples are used. Measurement points include: 16 exhaust manifold outlets, engine cylinder sleeve cooling water outlet manifold, exhaust manifold after turbocharger, intercooler outlet air pipe, main lubricating oil channel. The thermocouples are fixed to the measurement point wall surface through threaded installation or welding to ensure good thermal contact. The millivolt-level signals generated by the thermocouples are connected to a temperature signal conditioning module for cold end compensation and amplification.

[0075] 3) Process parameter monitoring: Use existing engine sensors or add piezoresistive pressure sensors to monitor boost pressure and oil pressure; use Hall effect or magneto electric speed sensors to monitor crankshaft speed and accurately capture top dead center (TDC) phase signals. This TDC signal will serve as the global synchronization trigger reference for the entire data acquisition system.

[0076] (2) Synchronous data acquisition system 1) Use a multi-channel dynamic signal acquisition system composed of a main control box and several functional modules. The voltage output of the vibration signal conditioning is connected to a high-precision dynamic signal acquisition module; the thermocouple signal is connected to a multi-channel thermocouple dedicated acquisition module; the pressure, speed, etc. signals are connected to high-level analog and digital input modules.

[0077] 2) Key synchronization technology: All acquisition modules share the same hardware clock source provided by the main control box. Acquisition is controlled by a unified hardware trigger, which is triggered by the rising edge of the crankshaft TDC signal. This design ensures that the data acquisition of all physical channels is strictly synchronized within microseconds, ensuring the time consistency of the data from the source.

[0078] (3) Layered computing and communication platform: Real-time simulation layer: Use a real-time simulation computer running a deterministic real-time operating system. This layer is responsible for deploying and running the high-fidelity digital twin model of the engine and executing hard real-time simulation tasks with a fixed simulation step size of 1ms. This layer directly interacts with the data acquisition system through a high-speed I / O board to obtain synchronized sensor data.

[0079] Monitoring and optimization layer: an industrial workstation is used to run general operating system and corresponding scientific computing environment. This layer is responsible for running human-computer interaction interface, synchronous error calculation, intelligent decision logic, optimization algorithm and other computing-intensive but non-hard real-time tasks.

[0080] Interlayer communication: the real-time layer and the monitoring layer are connected through high-speed industrial Ethernet, and the data exchange is carried out through the special communication interface based on TCP / IP protocol. The real-time layer uploads the simulation results (such as simulation vibration, temperature) to the monitoring layer; the monitoring layer issues calibration instructions and optimized parameter set to the real-time layer.

[0081] Correspondingly, the embodiment also provides a locomotive engine state synchronization and self-calibration method, comprising the following steps: 1) Digital twin model modeling and deployment Using multi-physics domain modeling and simulation software, based on the physical mechanism of the engine, a multi-domain coupled high-fidelity model is constructed, which includes a crankshaft system dynamics model, an in-cylinder working process model (covering intake, compression, combustion, expansion, exhaust), a turbocharger system model, a cooling system model and a lubrication system model.

[0082] The following four key physical parameters sensitive to performance degradation in the model are set as online adjustable input variables: Equivalent damping coefficient of crankshaft system (unit: N·m·s / rad); Thermal convection coefficient of cylinder liner-cooling water (unit: W / m²·K); Turbosupercharger isentropic efficiency correction multiplier (dimensionless); Combustion heat release rate shape parameter (dimensionless).

[0083] Using the code generation function of the simulation software, the mechanism model is compiled into C code that can run on the real-time operating system, and is deployed on the real-time simulation computer to form a real-time digital twin that runs in parallel with the physical engine.

[0084] 2) Real-time calculation of synchronization error On the monitoring layer workstation, a timing cycle task is created, with a period of 100 ms; Data alignment and acquisition: at the beginning of each task period, through the communication interface, all sensor measured data and their corresponding simulation data in the digital twin within the last 100 ms period are synchronously read.

[0085] Itemized error calculation: Vibration error: calculate the root mean square error (RMSE) of the measured and simulated vibration signals, and normalize it (i.e. divide by the maximum amplitude value of the measured vibration signal).

[0086] Thermodynamic error: Calculate the mean absolute percentage error (MAPE) of the simulated temperature values ​​for all thermocouple measuring points.

[0087] Assign dynamic weights to each component error: Different physical quantities have varying degrees of importance in characterizing engine health. Therefore, it is necessary to perform weighted fusion of the aforementioned individual errors.

[0088] Weighting coefficients: assign a weighting coefficient to each of the vibration error and thermodynamic error.

[0089] Configuration method: Weight configuration strategies can be preset through the system's human-computer interaction interface.

[0090] Calculate the overall synchronization error: Add the weighted individual errors together to obtain the final overall synchronization error value. .

[0091] 3) Self-calibration triggering and execution The system monitors the comprehensive synchronization error value in real time. It also checks the following three conditions simultaneously. If any one of the conditions is met, the parameter self-calibration process is automatically triggered.

[0092] Threshold trigger (used to respond to significant deviations): when the overall synchronization error is... Exceeding a set master threshold At that time, that is It triggers immediately. This condition is used to provide a rapid response when there is a significant drop in model accuracy.

[0093] Continuous over-limit triggering (used to capture slow degradation): When the overall synchronization error is... Exceeding a lower sub-threshold And this state continues for more than a preset period of time. When this occurs, calibration is triggered.

[0094] Rate of change trigger (used to respond to sudden failures): When the overall synchronization error is... The instantaneous rate of change exceeds a set rate of change threshold. At that time, that is It will be triggered immediately.

[0095] Self-calibration execution: Once the triggering conditions are met, the system will automatically execute the following optimization process: Define the optimization problem: Find a set of optimal model parameters such that the overall synchronization error generated by running a digital twin model under these parameters is minimized. Minimum, that is ,in These are the key parameters of the model to be calibrated, such as damping coefficient, thermal convection coefficient, and friction coefficient.

[0096] Optimization algorithm selection: Particle Swarm Optimization (PSO) algorithm is adopted.

[0097] Algorithm configuration: Number of particles: Set to 20-50 to strike a balance between search efficiency and computational overhead.

[0098] Maximum number of iterations: Set to 50-100 times to control optimization time.

[0099] Parameter range: for each parameter to be optimized Set a reasonable search range based on its physical meaning (e.g., limit the search range of the friction coefficient to between 0.08 and 0.15) to ensure that the optimization results are physically feasible.

[0100] 4) Model update and validation The new parameter set obtained through optimization Before updating to the live online model, its effectiveness can be verified in the simulation environment using a period of historical data, and errors can be observed. Whether it is significantly reduced.

[0101] Once the verification is successful, the system will choose to perform the update when the engine is in a stable operating condition (such as idling).

[0102] By using atomic operations or a double-buffering mechanism, new parameters are replaced in one go in the running digital twin model during the intervals between simulation steps, ensuring continuous and uninterrupted model operation and achieving seamless hot updates.

[0103] After the update is complete, the system automatically monitors and calculates the average synchronization error over the subsequent period. Only when Compared to the error before calibration The calibration is considered successful only when the decrease exceeds the preset value.

[0104] Failure rollback: If the success criteria are not met, or even if the error increases, the system determines that the calibration has failed and automatically rolls back to the previous set of valid parameters, while issuing an alarm requiring manual intervention. This mechanism effectively prevents the risk of model divergence caused by incorrect parameters.

[0105] Example 4 This embodiment is based on embodiment 1: This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the locomotive engine state synchronization and self-calibration method of Embodiment 1 or 2. The computer program can be in the form of source code, object code, executable file, or some intermediate form.

[0106] Example 5 This embodiment is based on embodiment 1: This embodiment provides a computer readable storage medium storing a computer program, which, when executed by a processor, implements the locomotive engine state synchronization and self-calibration method of embodiments 1 or 2. The computer program can be in the form of source code, object code, an executable file, or some intermediate form, etc. The storage medium includes any entity or device capable of carrying the computer program code, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium, etc.

[0107] The above only describes the preferred embodiments of the present application, and it should be understood that the present application is not limited to the forms disclosed herein, and should not be considered as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concepts described herein by the above teachings or related art or knowledge. Any modification and change made by those skilled in the art without departing from the spirit and scope of the present application shall be within the protection scope of the claims of the present application.

[0108] It should be noted that, for the foregoing method embodiments, in order to facilitate description, they are expressed as a combination of a series of actions, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

Claims

1. A method of locomotive engine state synchronization and self-calibration, the method comprising: The method comprises the following steps: Real-time synchronization and data preprocessing: receiving multi-modal raw data of a locomotive engine collected by physical sensors, and pre-processing the multi-modal raw data as boundary conditions or input excitations to drive the multi-domain coupled mechanism model in the digital twin to run one simulation step in real time; Multi-modal weighted fusion and synchronization error calculation: respectively calculating vibration error and thermodynamic error, and then obtaining comprehensive synchronization error through weighted fusion; Intelligent self-calibration triggering: setting threshold triggering conditions, continuous over-limit triggering conditions and synchronization error change rate triggering conditions, and triggering parameter self-calibration instructions when any triggering condition is met; Parameter automatic calibration and safe update: after receiving the parameter self-calibration instruction, starting the optimization engine to build an optimization model with the objective of minimizing the comprehensive synchronization error, obtaining the optimal parameter set through the optimization algorithm, and updating the multi-domain coupled mechanism model in the digital twin when the engine operating condition is stable after validity verification.

2. The locomotive engine state synchronization and self-calibration method of claim 1, wherein, The preprocessing of the multi-modal raw data of the locomotive engine comprises denoising filtering, invalid value elimination and unit unification.

3. The locomotive engine state synchronization and self-calibration method of claim 1, wherein, The method for calculating the vibration error and the thermodynamic error comprises: calculating the root mean square error of the measured vibration signal and the simulated vibration signal in the same time period at the same measuring point, and then normalizing the root mean square error by dividing the maximum amplitude of the measured vibration signal, thereby obtaining the vibration error; calculating the average absolute percentage error of the measured temperature and the simulated temperature of the plurality of thermocouple measuring points, the average absolute percentage error being the average value of the ratio of the absolute value of the temperature difference of each measuring point to the corresponding measured temperature, thereby obtaining the thermodynamic error.

4. The locomotive engine state synchronization and self-calibration method of claim 1, wherein, The threshold triggering condition comprises that the comprehensive synchronization error is greater than a first preset threshold; the continuous over-limit triggering condition comprises that the comprehensive synchronization error is greater than a second preset threshold and the duration exceeds a preset holding time; and the synchronization error change rate triggering condition comprises that the change rate of the comprehensive synchronization error is greater than a preset change rate threshold.

5. The locomotive engine state synchronization and self-calibration method of claim 1, wherein, The optimization algorithm comprises a particle swarm optimization algorithm, and the parameter types in the optimal parameter set comprise damping coefficients and heat convection coefficients.

6. The locomotive engine state synchronization and self-calibration method of claim 1, wherein, After the validity verification, the multi-domain coupled mechanism model in the digital twin is updated at the engine operating condition stable time, which comprises: verifying whether the optimal parameter set can significantly reduce the comprehensive synchronization error in the simulation environment using historical data, and if so, updating the multi-domain coupled mechanism model in the digital twin according to the optimal parameter set under the idle state of the engine by using the write-ahead backup or atomic operation method.

7. The locomotive engine state synchronization and self-calibration method of claim 1, wherein, The post-update verification comprises: monitoring the comprehensive synchronization error in a period of time after the update, and if the comprehensive synchronization error decreases by less than a preset value or increases compared with the comprehensive synchronization error before calibration, determining that the calibration fails and automatically rolling back to the last set of valid parameters.

8. A locomotive engine state synchronization and self-calibration system, characterized by, The method comprises the following steps: A real-time synchronization and data preprocessing module is configured to receive multi-modal raw data of a locomotive engine collected by physical sensors, and pre-process the multi-modal raw data as boundary conditions or input excitations to drive the multi-domain coupled mechanism model in the digital twin to run one simulation step in real time; The multi-modal weighted fusion synchronous error calculation module is configured to calculate vibration errors and thermodynamic errors, and then obtain a comprehensive synchronous error through weighted fusion; The intelligent self-calibration trigger module is configured to set threshold trigger conditions, continuous over-limit trigger conditions, and synchronous error change rate trigger conditions, and trigger a parameter self-calibration instruction when any trigger condition is met; The parameter automatic calibration and safety update module is configured to, after receiving the parameter self-calibration instruction, start an optimization engine to build an optimization model with the objective of minimizing the comprehensive synchronous error, obtain an optimal parameter set through an optimization algorithm, and update the multi-domain coupled mechanism model in the digital twin to the engine stable working condition time after effectiveness verification. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the locomotive engine state synchronization and self-calibration method of any one of claims 1-7.

10. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the locomotive engine state synchronization and self-calibration method of any one of claims 1-7.

Images