A method and system for valve vibration and shock testing based on digital twins

Abstract

A method and system for valve vibration and shock testing based on digital twins belong to the field of industrial equipment testing and evaluation technology. The method includes: First, deploying a multi-source sensor network to construct the physical entity layer for valve vibration and shock testing, acquiring and preprocessing various physical monitoring data, and constructing a digital twin layer. Second, based on the various physical monitoring data and the digital twin layer, real-time data assimilation with the physical test bench is achieved through parameter calibration and updating of the digital twin model. Finally, virtual vibration and shock testing is performed, and the global response and key component information of the physical entity are predicted and evaluated based on the test results, ultimately achieving interaction and performance prediction. The method for valve vibration and shock testing is implemented through a valve vibration and shock testing system. This invention can significantly improve fault diagnosis capabilities, early damage identification capabilities, and valve condition prediction and evaluation capabilities, overcoming the shortcomings of insufficient on-site monitoring information.

Description

Technical Field

[0001] This invention belongs to the field of industrial equipment testing and evaluation technology, and relates to a method and system for valve vibration and impact testing based on digital twins. Background Technology

[0002] As a core component of fluid control, the health of valves directly affects the safe, stable, and efficient operation of the entire system. Vibration and shock are among the main causes of valve failure (such as stem breakage, seat damage, seal leakage, and actuator malfunction). Therefore, conducting vibration and shock tests on valves helps ensure their reliability and safety, and provides important data support for valve design and optimization.

[0003] Valve vibration and impact testing has been extensively studied. Yu Muquan proposed a valve vibration testing device and method (Chinese invention patent CN202111147023.1). However, his method deviates from real-world operating conditions, ignores the fluid-structure interaction effect in pipelines, lacks internal state sensing capabilities, and cannot monitor key parameters such as sealing surface contact stress and valve stem root stress. Furthermore, destructive testing is costly and requires destroying a physical prototype. Based on existing research, traditional valve vibration and impact testing suffers from the following significant drawbacks:

[0004] The model and data are severely disconnected. Traditional valve design models (such as CAD and FEA static analysis) cannot reflect their real-time behavior under dynamic vibration and impact loads. They lack closed-loop verification that integrates virtual and real data, and cannot use high-fidelity virtual model prediction results to guide physical testing or optimize operation. They cannot perform status early warning and performance evaluation. Physical testing is costly and can easily damage test prototypes.

[0005] The rise of digital twin technology offers an opportunity to address these issues. Digital twins aim to construct a dynamic virtual mapping of the entire lifecycle of a physical entity, achieving deep interaction and integration between the physical and information worlds. This enables performance prediction, status warning, lifespan assessment, and virtual experiment optimization, while reducing the number of physical experiments and lowering testing costs and risks. However, existing research has not yet effectively solved this technical problem. Fu Yumin et al. proposed a vibration testing method, system, equipment, and medium based on digital twins (Chinese Invention Patent CN202210124770.1). However, their invention lacks a model update mechanism, heavily relies on preset parameters, and ignores multi-physics coupling effects, focusing only on mechanical vibration. Therefore, to overcome the shortcomings of existing technologies and achieve more efficient, comprehensive, and intelligent valve performance testing, evaluation, and safety assurance, it is necessary to invent a valve vibration and impact testing method and system based on digital twins. Summary of the Invention

[0006] To address the problems and shortcomings of existing technologies, this invention provides a method and system for valve vibration and shock testing based on digital twins. By constructing a physical entity layer and a digital twin layer for valve vibration and shock testing, and dynamically fusing them in a data-model fusion driving engine, virtual vibration and shock testing and evaluation are performed, realizing virtual-real interaction and performance prediction of valves.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A method for valve vibration and shock testing based on digital twins includes the following steps:

[0009] The first step is to construct the physical entity layer for valve vibration and impact testing, as follows:

[0010] Step 1.1: Install the target valve on the vibration and impact test bench and connect it to the external pipeline to build a physical test bench.

[0011] Step 1.2: Based on the physical test bench built in Step 1.1, deploy a multi-source sensor network at key locations on the target valve, external piping, and support structure. These key locations primarily include the valve body, valve cover, valve stem, valve core, actuator, and upstream and downstream pipe flanges. Specifically:

[0012] High-frequency accelerometers are installed around the valve body, at the top of the valve stem, and at the base of the valve seat to capture small-amplitude high-frequency vibration and impact signals.

[0013] Speed ​​sensors are installed around the valve body, on the actuator housing, and on the upstream and downstream pipe flanges to monitor low-frequency vibration signals and forced vibration signals.

[0014] Dynamic strain gauges are installed at stress concentration points on the valve stem, at the root of the valve seat, and at the valve cover-valve body connection to measure vibration stress and strain.

[0015] Acoustic emission sensors are installed at stress concentration points on the valve stem, at the root of the valve seat, at the valve body-external pipeline connection, and at the valve cover-bolt connection to monitor abnormal acoustic signals.

[0016] Temperature sensors are installed on the valve seat sealing surface contact area and valve stem to monitor the temperature rise caused by friction.

[0017] A pressure sensor is installed at the valve core to monitor the fluid acceleration impact force.

[0018] Displacement sensors are installed on the valve stem, valve seat, and actuator push rod to measure actual displacement and minute deformation.

[0019] The flow meter is installed at the inlet and outlet of the external pipeline to monitor the actual flow rate.

[0020] A thrust sensor is installed at the actuator push rod to monitor the actuator's status.

[0021] The second step involves collecting and preprocessing various physical monitoring data based on the multi-source sensor network deployed in the first step, as detailed below:

[0022] Step 2.1: Real-time, synchronous, and high-precision acquisition of various physical monitoring data based on a multi-source sensor network.

[0023] Step 2.2: Based on the various physical monitoring data obtained in Step 2.1, perform in-depth cleaning to remove outliers, reduce noise, and remove missing interpolated data, resulting in cleaned data.

[0024] Step 2.3: Based on the cleaned data obtained in Step 2.2, time synchronization alignment is achieved by compensating for time shift.

[0025] Step 2.4: Based on the aligned data obtained in Step 2.3, feature extraction is performed, including a time-domain feature set, a frequency-domain feature set, and a time-frequency-domain feature set. The time-domain feature set includes RMS values, peak values, and kurtosis. The frequency-domain feature set includes the spectrum, envelope spectrum, and cepstrum. The time-frequency-domain feature set is a wavelet transform.

[0026] The third step involves constructing a digital twin layer for valve vibration and shock testing. Based on the physical entity layer for valve vibration and shock testing constructed in the first step and the various physical monitoring data preprocessed in the second step, a digital twin model is constructed and its order is reduced to achieve a dynamic virtual mapping of the physical entity layer for valve vibration and shock testing. The details are as follows:

[0027] Step 3.1, Construct the geometric model. Based on the physical entity layer of the valve vibration and impact test constructed in the first step, a precise CAD model of the valve, including key details, is constructed. These key details mainly include the valve disc, valve stem, and flow channel.

[0028] Step 3.2: Construct a multiphysics property model. Based on the physical entity layer for valve vibration and impact testing constructed in Step 1, this includes establishing a library of material properties, flow properties, and boundary conditions. Specifically:

[0029] The establishment of material properties involves accurately inputting the mechanical properties of each component material at the service temperature, including elastic modulus, Poisson's ratio, density, yield strength, tensile strength, and thermophysical properties, including thermal conductivity, specific heat capacity, and coefficient of thermal expansion.

[0030] The establishment of fluid properties: defining the physical properties of the medium, including density, viscosity, and compressibility.

[0031] The establishment of the boundary condition library involves creating boundary condition templates for typical operating conditions, including steady-state, opening and closing transient, water hammer, seismic testing, and safety valve discharge.

[0032] Step 3.3: Construct the structural dynamics simulation model. Based on the geometric model constructed in Step 3.1, a dynamic simulation model of the valve structure is established using the finite element method to ensure support for transient dynamic analysis, while also considering nonlinear simulation calculations, as shown in Equation (1):

[0033] (1)

[0034] in, For displacement; For speed; For acceleration; For time; For fluid pressure; This is the quality matrix; Here is the damping matrix; Here is the stiffness matrix; External incentives; Contact force; It is a fluid force.

[0035] Step 3.4: Construct a computational fluid dynamics (CFD) simulation model. Based on the geometric model constructed in Step 3.1, the complex flow field, pressure pulsation, transient hydraulic or aerodynamic forces inside the valve and along the leakage path are calculated using transient turbulence simulation to ensure accurate simulation of the flow field inside the valve, as shown in Equations (2) and (3):

[0036] (2)

[0037] (3)

[0038] in, For flow rate; For time; For fluid pressure; For fluid density; Kinematic viscosity; External force; This is the Del operator.

[0039] Step 3.5: Construct a fluid-structure interaction (FSI) simulation model. Based on the structural dynamics simulation model constructed in Step 3.3 and the computational fluid dynamics (CFD) simulation model constructed in Step 3.4, a two-way coupled data transfer is established between the two through fluid-structure interaction simulation. This involves transferring displacement to fluid mesh deformation and fluid pressure to structural loads, thereby realizing the influence of fluid loads on structural deformation or vibration, as well as the reaction of structural deformation to the flow field, achieving force balance and motion coordination at the coupling interface.

[0040] Step 3.6, construct the friction and wear model. Based on the physical entity layer of the valve vibration and impact test constructed in the first step, the Arcard wear model is used to predict the fretting wear of key friction pairs, including the sealing pair, valve stem and packing, under vibration and impact, as shown in Equation (4):

[0041] (4)

[0042] in, This refers to the wear volume; The wear coefficient is a dimensionless coefficient. For normal loads; This is the sliding distance; It is Brinell hardness.

[0043] Step 3.7, construct the fatigue damage model. Based on the various physical monitoring data after preprocessing in the second step, the fatigue crack initiation life of key parts is predicted using the SWT model in the critical plane method, based on the dynamic stress or strain results, as shown in formula (5):

[0044] (5)

[0045] in, For SWT parameters, For maximum stress, For strain amplitude.

[0046] Then, steps 3.1 to 3.7 above complete the construction of the digital twin model.

[0047] Step 3.8, Reduction of the Digital Twin Model. Based on the digital twin model constructed in steps 3.1 to 3.7, which is a high-dimensional full-order model, by using intrinsic orthogonal decomposition or dynamic modal analysis, while ensuring the accuracy of key dynamic characteristics (including main modes and stress concentration areas), the high-dimensional full-order model is transformed into a low-order surrogate model suitable for real-time or near-real-time computation, and a reduction model library is established.

[0048] The fourth step is to construct a data-model fusion driving engine. Based on the various physical monitoring data preprocessed in the second step and the valve vibration and impact test digital twin layer constructed in the third step, real-time data assimilation with the physical test bench is achieved through parameter calibration and updating of the digital twin model, as detailed below:

[0049] Step 4.1, Digital Twin Model Parameter Calibration. Based on the various physical monitoring data preprocessed in Step 2, a genetic algorithm is used to calibrate the key uncertain parameters in the digital twin model constructed in Step 3, including contact stiffness, damping coefficient, and boundary conditions.

[0050] Step 4.2, Digital Twin Model Update. Based on the various physical monitoring data preprocessed in Step 2 and the calibrated digital twin model from Step 4.1, the real-time acquired physical monitoring data (especially vibration and shock response) is dynamically fused with the predicted output of the digital twin model using the Kalman filter algorithm. This achieves digital twin model updates, including updates to the digital twin model's state, parameters, and structure. Specifically:

[0051] The digital twin model state update estimates the current true "hidden state" of the target valve, including internal stress, microcrack propagation, and contact state.

[0052] The digital twin model parameters are updated online, including local stiffness degradation, damping changes, and friction coefficients, so that the prediction results of the digital twin model can approximate the actual response of the physical entity as closely as possible.

[0053] The digital twin model results are updated as follows: when a significant performance drift or impairment is detected, model reconstruction is triggered, and the model is replaced with a suitable reduced-order model.

[0054] The fifth step is to conduct virtual vibration and shock testing and evaluation. Based on the valve vibration and shock testing digital twin layer processed in the fourth step, various virtual vibration and shock testing scenarios are defined on the platform of the digital twin model, virtual vibration and shock tests are executed, and the global response and key component information of the physical entity are predicted and evaluated based on the test results.

[0055] Step 6: Based on the virtual vibration and shock test results obtained in Step 5, implement interactive performance prediction, as detailed below:

[0056] Step 6.1, Visualization and Interaction. The state of the digital twin model, such as real-time response, stress cloud map, and prediction results, is overlaid onto the target valve using a 3D visualization platform, providing an intuitive and interactive experience.

[0057] Step 6.2, Anomaly Diagnosis and Early Warning. Through the fusion analysis of physical monitoring data characteristics and predicted characteristics after virtual vibration and impact testing, a machine learning-driven intelligent diagnostic algorithm is used to achieve early fault detection, type identification, severity assessment, and location. When an anomaly is detected or predicted damage approaches a set threshold, multi-level early warnings are automatically triggered.

[0058] Step 6.3, Condition Prediction and Assessment. The processed digital twin model is used to predict conditions inside the valve that are difficult to measure directly, including the contact stress distribution on the sealing surface, stress concentration at the valve stem root, and microcrack propagation trends. Simultaneously, fatigue damage accumulation analysis and prediction are performed based on the fatigue damage model constructed in step three to assess the remaining service life of key valve components.

[0059] A valve vibration and shock testing system based on digital twins is disclosed. The system modularizes each step of the valve vibration and shock testing method based on digital twins. The system includes a physical sensing module, a digital twin model management module, a data-driven engine module, a virtual testing module, and an intelligent application and interaction module. Specifically:

[0060] The physical sensing module mainly consists of a target valve physical test bench, a multi-source sensor network, and a data acquisition and processing unit. By deploying a multi-source sensor network on the physical test bench, the data acquisition unit collects physical monitoring data with high precision, multiple channels, and synchronously. The data is preprocessed in the data processing unit to ensure data integrity and accuracy, while providing high-quality input data for subsequent modules.

[0061] The digital twin model management module is connected to the physical sensing module. It is used to integrate and call the solver of the digital twin model to solve the model, and to implement the order reduction algorithm using the model order reduction toolkit. At the same time, it completes the storage, loading and scheduling of the full-order model and the order reduction proxy model, and realizes the dynamic virtual mapping of the physical entity layer of the valve vibration and impact test.

[0062] The data-driven engine module connects to the digital twin model management module and the physical sensing module, and achieves real-time data assimilation with the physical test bench through model calibration and updates.

[0063] The virtual testing module is connected to the data-driven engine module, the digital twin model management module, and the physical sensing module. It is used to manage the definition of virtual testing scenarios, load application, submission and monitoring of virtual vibration and impact test tasks, and reading of test results, so as to predict and evaluate the global response and key part information of physical entities.

[0064] The intelligent application and interaction module connects to the virtual testing module, data-driven engine module, digital twin model management module, and physical perception module. It mainly consists of a visualization platform, intelligent diagnosis and prediction module, and analysis and evaluation module, which displays physical data, digital twin status, analysis results, and early warning information. At the same time, by integrating a machine learning algorithm library, it realizes valve fault diagnosis and life prediction.

[0065] The beneficial effects of this invention are as follows:

[0066] This invention significantly improves fault diagnosis capabilities, early damage identification capabilities, and valve condition prediction and assessment capabilities by constructing a physical entity layer and a digital twin layer for valve vibration and shock testing, and dynamically fusing them in a data-model fusion engine, overcoming the shortcomings of insufficient on-site monitoring information. Simultaneously, by relying on virtual vibration and shock testing to replace most high-risk, high-cost physical destructive tests, it reduces the number of prototype tests and the testing cycle, greatly reducing costs and risks. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of a valve vibration and impact testing method based on digital twins according to the present invention.

[0068] Figure 2 This is a schematic diagram of the system flow for valve vibration and impact testing based on digital twins according to the present invention. Detailed Implementation

[0069] The present invention will be further illustrated below with an example and comparative example of vibration and shock testing applied to a high-pressure large-diameter ball valve at an LNG receiving station.

[0070] Example 1

[0071] A method for valve vibration and shock testing based on digital twins includes the following steps:

[0072] The first step is to construct the physical entity layer for valve vibration and impact testing, as follows:

[0073] Step 1.1: Install the target valve on the vibration and impact test bench and connect it to the external pipeline to build a physical test bench.

[0074] Step 1.2: Based on the physical test bench built in Step 1.1, deploy a multi-source sensor network at key locations on the target valve, external piping, and support structure. These key locations primarily include the valve body, valve cover, valve stem, valve core, actuator, and upstream and downstream pipe flanges. Details are as follows:

[0075] High-frequency accelerometers are installed around the valve body, at the top of the valve stem, and at the base of the valve seat to capture small-amplitude high-frequency vibration and impact signals.

[0076] Speed ​​sensors are installed around the valve body, on the actuator housing, and on the upstream and downstream pipe flanges to monitor low-frequency vibration signals and forced vibration signals.

[0077] Dynamic strain gauges are installed at stress concentration points on the valve stem, at the root of the valve seat, and at the valve cover-valve body connection to measure vibration stress and strain.

[0078] Acoustic emission sensors are installed at stress concentration points on the valve stem, at the root of the valve seat, at the valve body-external pipeline connection, and at the valve cover-bolt connection to monitor abnormal acoustic signals.

[0079] Temperature sensors are installed on the valve seat sealing surface contact area and valve stem to monitor the temperature rise caused by friction.

[0080] A pressure sensor is installed at the valve core to monitor the fluid acceleration impact force.

[0081] Displacement sensors are installed on the valve stem, valve seat, and actuator push rod to measure actual displacement and minute deformation.

[0082] The flow meter is installed at the inlet and outlet of the external pipeline to monitor the actual flow rate.

[0083] A thrust sensor is installed at the actuator push rod to monitor the actuator's status.

[0084] The second step involves collecting and preprocessing various physical monitoring data based on the multi-source sensor network deployed in the first step, as detailed below:

[0085] Step 2.1: Based on the multi-source sensor network deployed in the first step, perform real-time, synchronous, and high-precision acquisition of various physical monitoring data.

[0086] Step 2.2 involves deep cleaning of the physical monitoring data obtained in Step 2.1, removing outliers, reducing noise, and interpolating missing data.

[0087] Step 2.3: Based on the cleaned data obtained in Step 2.2, time synchronization alignment is achieved by compensating for time shift.

[0088] Step 2.4: Based on the aligned data obtained in Step 2.3, feature extraction is performed, including a time-domain feature set, a frequency-domain feature set, and a time-frequency-domain feature set. The time-domain feature set includes RMS values, peak values, and kurtosis. The frequency-domain feature set includes the spectrum, envelope spectrum, and cepstrum. The time-frequency-domain feature set is a wavelet transform.

[0089] The third step involves constructing a digital twin layer for valve vibration and shock testing. Based on the physical entity layer for valve vibration and shock testing constructed in the first step and the various physical monitoring data preprocessed in the second step, a digital twin model is constructed and its order is reduced to achieve a dynamic virtual mapping of the physical entity layer for valve vibration and shock testing. The details are as follows:

[0090] Step 3.1, Construct the geometric model. Based on the physical entity layer of the valve vibration and impact test constructed in the first step, a precise CAD model of the valve, including key details, is constructed. These key details mainly include the valve disc, valve stem, and flow channel.

[0091] Step 3.2: Construct a multiphysics property model. Based on the physical entity layer for valve vibration and impact testing constructed in Step 1, this includes establishing a library of material properties, flow properties, and boundary conditions. Specifically:

[0092] The establishment of material properties involves accurately inputting the mechanical properties of each component material at the service temperature, including elastic modulus, Poisson's ratio, density, yield strength, tensile strength, and thermophysical properties, including thermal conductivity, specific heat capacity, and coefficient of thermal expansion.

[0093] The establishment of fluid properties: defining the physical properties of the medium, including density, viscosity, and compressibility.

[0094] The establishment of the boundary condition library involves creating boundary condition templates for typical operating conditions, including steady-state, opening and closing transient, water hammer, seismic testing, and safety valve discharge.

[0095] In this embodiment, the valve body and valve core are made of... Stainless steel has a modulus of 214 GPa, a Poisson's ratio of 0.29, and a density of [missing information] at an operating temperature of -162℃. The yield strength is 550 MPa, the tensile strength is 850 MPa, and the thermal conductivity is... Specific heat capacity is The coefficient of thermal expansion is The fluid is liquefied natural gas (LNG), with a density of [insert density here] at an operating temperature of -162°C. The dynamic viscosity is The kinematic viscosity is Compression rate For various typical operating conditions, a corresponding boundary condition database should be established with reference to national or international standards.

[0096] Step 3.3, construct the structural dynamics simulation model. Based on the geometric model constructed in Step 3.1, a dynamic simulation model of the valve structure is established using the finite element method to ensure support for transient dynamic analysis, while also considering nonlinear simulation calculations, as shown in Equation (1).

[0097] In this embodiment, transient dynamics simulation is performed in Ansys-Workbench to analyze the dynamic response of the valve structure.

[0098] Step 3.4: Construct a computational fluid dynamics (CFD) simulation model. Based on the geometric model constructed in step 3.1, the complex flow field, pressure pulsation, transient hydraulic or aerodynamic forces inside the valve and the leakage path are calculated through transient turbulence simulation to ensure accurate simulation of the flow field inside the valve, as shown in formulas (2) and (3).

[0099] In this embodiment, a fluid simulation model of the internal flow channel of the ball valve is constructed in Ansys-Fluent to simulate the flow field under different opening degrees, including pressure distribution, flow velocity, turbulence, and pay special attention to the flow near the ball.

[0100] Step 3.5: Construct a fluid-structure interaction (FSI) simulation model. Based on the structural dynamics simulation model constructed in Step 3.3 and the computational fluid dynamics (CFD) simulation model constructed in Step 3.4, a two-way coupled data transfer is established between the two through fluid-structure interaction simulation. This involves transferring displacement to fluid mesh deformation and fluid pressure to structural loads, thereby realizing the influence of fluid loads on structural deformation or vibration, as well as the reaction of structural deformation to the flow field, achieving force balance and motion coordination at the coupling interface.

[0101] In this embodiment, Ansys-System-Coupling is used to implement bidirectional fluid-structure interaction simulation, which transmits fluid pressure to the structure and feeds displacement back to the fluid to update the mesh.

[0102] Step 3.6, construct the friction and wear model. Based on the physical entity layer of the valve vibration and impact test constructed in the first step, the Arcard wear model is used to predict the fretting wear of key friction pairs including the sealing pair, valve stem and packing under vibration and impact, as shown in Equation (4).

[0103] Step 3.7, construct the fatigue damage model. Based on the various physical monitoring data after the preprocessing in the second step, the fatigue crack initiation life of key parts is predicted by using the SWT model in the critical plane method and the dynamic stress or strain results, as shown in formula (5).

[0104] Then, steps 3.1 to 3.7 above complete the construction of the digital twin model.

[0105] Step 3.8, Reduction of the Digital Twin Model. Based on the digital twin model constructed in steps 3.1 to 3.7, which is a high-dimensional full-order model, by using intrinsic orthogonal decomposition or dynamic modal analysis, while ensuring the accuracy of key dynamic characteristics (including main modes and stress concentration areas), the high-dimensional full-order model is transformed into a low-order surrogate model suitable for real-time or near-real-time computation, and a reduction model library is established.

[0106] In this embodiment, modal analysis is performed on the full-order finite element model to extract the first 50 modes, covering 0-1500Hz. The intrinsic orthogonal decomposition method is applied to establish a reduced-order model library, covering combined working conditions with different opening degrees and different pressure levels, while verifying the accuracy of the reduced-order model.

[0107] The fourth step is to construct a data-model fusion driving engine. Based on the various physical monitoring data preprocessed in the second step and the valve vibration and impact test digital twin layer constructed in the third step, real-time data assimilation with the physical test bench is achieved through model parameter calibration and model updates of the digital twin model, as detailed below:

[0108] Step 4.1, Digital Twin Model Parameter Calibration. Based on the various physical monitoring data preprocessed in Step 2, a genetic algorithm is used to calibrate the key uncertain parameters in the digital twin model constructed in Step 3, including contact stiffness, damping coefficient, and boundary conditions.

[0109] Step 4.2, Digital Twin Model Update. Based on the various physical monitoring data preprocessed in Step 2 and the calibrated digital twin model from Step 4.1, the real-time acquired physical monitoring data (especially vibration and shock response) is dynamically fused with the predicted output of the digital twin model using the Kalman filter algorithm. This achieves the digital twin model update, including updates to the digital twin model's state, parameters, and structure.

[0110] Digital twin model state update: Estimate the current true "hidden state" of the target valve, including internal stress, microcrack propagation, and contact state.

[0111] Digital twin model parameter updates: Online adjustment of model parameters, including local stiffness degradation, damping changes, and friction coefficient, to make the prediction results of the digital twin model approximate the actual response of the physical entity as closely as possible.

[0112] Digital twin model result update: When significant performance drift or impairment is detected, model reconstruction is triggered and replaced with a suitable reduced-order model.

[0113] The fifth step is to conduct virtual vibration and shock testing and evaluation. Based on the valve vibration and shock testing digital twin layer processed in the fourth step, various virtual vibration and shock testing scenarios are defined on the platform of the digital twin model, virtual vibration and shock tests are executed, and the global response and key component information of the physical entity are predicted and evaluated based on the test results.

[0114] Step 6: Based on the virtual vibration and shock test results obtained in Step 5, implement interactive performance prediction, as detailed below:

[0115] Step 6.1, Visualization and Interaction. The state of the digital twin model, such as real-time response, stress cloud map, and prediction results, is overlaid onto the target valve using a 3D visualization platform, providing an intuitive and interactive experience.

[0116] In this embodiment, physical sensor data and images are displayed in real time using Unity-3D, and the status information obtained from the real-time virtual test is mapped and displayed, marking the potential damage location and severity estimated by Kalman filtering.

[0117] Step 6.2, Anomaly Diagnosis and Early Warning. Through the fusion analysis of physical monitoring data characteristics and predicted characteristics after virtual vibration and impact testing, a machine learning-driven intelligent diagnostic algorithm is used to achieve early fault detection, type identification, severity assessment, and location. When an anomaly is detected or predicted damage approaches a set threshold, multi-level early warnings are automatically triggered.

[0118] Step 6.3, Condition Prediction and Assessment. The processed digital twin model is used to predict conditions inside the valve that are difficult to measure directly, including the contact stress distribution on the sealing surface, stress concentration at the valve stem root, and microcrack propagation trends. Simultaneously, fatigue damage accumulation analysis and prediction are performed based on the fatigue damage model constructed in step three to assess the remaining service life of key valve components.

[0119] A valve vibration and shock testing system based on digital twins is disclosed. The system modularizes each step of the valve vibration and shock testing method based on digital twins. The system includes a physical sensing module, a digital twin model management module, a data-driven engine module, a virtual testing module, and an intelligent application and interaction module. Specifically:

[0120] The physical sensing module mainly consists of a target valve physical test bench, a multi-source sensor network, and a data acquisition and processing unit. By deploying a multi-source sensor network on the physical test bench, the data acquisition unit collects physical monitoring data with high precision, multiple channels, and synchronously. The data is preprocessed in the data processing unit to ensure data integrity and accuracy, while providing high-quality input data for subsequent modules.

[0121] The digital twin model management module is connected to the physical sensing module. It is responsible for integrating and calling the solver of the digital twin model to solve the model, and using the model reduction toolkit to implement the reduction algorithm. At the same time, it completes the storage, loading and scheduling of the full-order model and the reduced-order proxy model, realizing the dynamic virtual mapping of the physical entity layer of the valve vibration and impact test.

[0122] The data-driven engine module connects to the digital twin model management module and the physical sensing module, and achieves real-time data assimilation with the physical test bench through model calibration and updates.

[0123] The virtual testing module connects to the data-driven engine module, the digital twin model management module, and the physical sensing module. It is responsible for managing the definition of virtual testing scenarios, load application, submission and monitoring of virtual vibration and impact test tasks, and reading of test results, thereby enabling the prediction and evaluation of the global response and key component information of physical entities.

[0124] The intelligent application and interaction module connects to the virtual testing module, data-driven engine module, digital twin model management module, and physical perception module. It mainly consists of a visualization platform, intelligent diagnosis and prediction module, and analysis and evaluation module, which displays physical data, digital twin status, analysis results, and early warning information. At the same time, by integrating a machine learning algorithm library, it realizes valve fault diagnosis and life prediction.

[0125] Comparative Example 1

[0126] To demonstrate the superiority of this invention, a seismic test of a high-pressure, large-diameter ball valve at an LNG receiving station is used as an example. As shown in Table 1, compared with traditional seismic tests based on GB / T-2424.25-2024 "Environmental Testing Part 3: Test Guidelines - Seismic Test Methods", the seismic testing cost of this invention is reduced by more than 99%, the testing efficiency is increased by more than 100 times, and the testing coverage is wider; fault warnings are provided more than 6 months in advance, achieving effective lifespan prediction and avoiding significant losses caused by unplanned downtime; valve reliability is improved. Table 1 shows that the valve vibration and shock testing method and system designed in this invention, based on digital twins, transforms valve management from "passive response" to "active prediction and optimization" under the drive of digital twins, creating significant economic and safety value.

[0127] Table 1. Specific Implementation Examples: Improvement Effect of the Invention

[0128]

[0129] The above embodiments are merely illustrative of the implementation methods of the present invention, but should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the protection scope of the present invention.

Claims

1. A method for valve vibration and shock testing based on digital twins, characterized in that, The method includes the following steps: The first step is to construct the physical entity layer for valve vibration and impact testing; The second step involves collecting and preprocessing various physical monitoring data based on the multi-source sensor network deployed in the first step. The third step involves constructing a digital twin layer for valve vibration and impact testing. Based on the physical entity layer for valve vibration and impact testing constructed in the first step and the various physical monitoring data preprocessed in the second step, a digital twin model is constructed and its order is reduced to achieve dynamic virtual mapping of the physical entity layer for valve vibration and impact testing. Specifically: Step 3.1, Construct the geometric model; Based on the physical entity layer of the valve vibration and impact test constructed in the first step, construct an accurate CAD model of the valve including key details, which mainly include the valve disc, valve stem and flow channel; Step 3.2, construct a multiphysics property model; based on the physical entity layer of valve vibration and impact testing constructed in the first step, including establishing a library of material properties, flow properties and boundary conditions; Step 3.3: Construct a structural dynamics simulation model; Based on the geometric model constructed in step 3.1, a dynamic simulation model of the valve structure is established using the finite element method to ensure support for transient dynamic analysis, while also considering nonlinear simulation calculations, as shown in formula (1): （1）； in, For displacement; For speed; For acceleration; For time; For fluid pressure; This is the quality matrix; Here is the damping matrix; Here is the stiffness matrix; External incentives; For contact force; Forces acting on fluids; Step 3.4: Construct a computational fluid dynamics (CFD) simulation model; based on the geometric model constructed in step 3.1, perform transient turbulence simulation to calculate the complex flow field, pressure pulsation, transient hydraulic or aerodynamic forces inside the valve and along the leakage path, ensuring accurate simulation of the flow field inside the valve, as shown in formulas (2) and (3): （2）； （3）； in, For flow rate; For time; For fluid pressure; For fluid density; Kinematic viscosity; External force; For the Del operator; Step 3.5: Construct a fluid-structure interaction (FSI) simulation model. Based on the structural dynamics simulation model constructed in Step 3.3 and the computational fluid dynamics (CFD) simulation model constructed in Step 3.4, establish bidirectional coupling data transfer between the two through fluid-structure interaction simulation, namely, displacement to fluid mesh deformation and fluid pressure to structural load, to realize the influence of fluid load on structural deformation or vibration, as well as the reaction of structural deformation on the flow field, and achieve force balance and motion coordination at the coupling interface. Step 3.6, construct the friction and wear model; based on the physical entity layer of the valve vibration and impact test constructed in the first step, the Arcard wear model is used to predict the fretting wear of key friction pairs including the sealing pair, valve stem and packing under vibration and impact, as shown in formula (4): （4）； in, This refers to the wear volume; The wear coefficient is a dimensionless coefficient. For normal loads; This is the sliding distance; Brinell hardness; Step 3.7, construct a fatigue damage model; based on the various physical monitoring data after preprocessing in the second step, the fatigue crack initiation life of key parts is predicted using the SWT model in the critical plane method and the dynamic stress or strain results, as shown in formula (5): （5）； in, For SWT parameters, For maximum stress, For strain amplitude; Then, steps 3.1 to 3.7 above complete the construction of the digital twin model; Step 3.8, Reduction of the order of the digital twin model; Based on the digital twin model constructed in steps 3.1 to 3.7, which is a high-dimensional full-order model, the high-dimensional full-order model is transformed into a low-order proxy model suitable for real-time or near-real-time computing by using intrinsic orthogonal decomposition or dynamic modal analysis, and a reduction model library is established. The fourth step is to construct a data-model fusion driving engine. Based on the various physical monitoring data preprocessed in the second step and the valve vibration and impact test digital twin layer constructed in the third step, real-time data assimilation with the physical test bench is achieved through parameter calibration and updating of the digital twin model. Specifically: Step 4.1, Digital Twin Model Parameter Calibration; Based on the various physical monitoring data preprocessed in the second step, a genetic algorithm is used to calibrate the key uncertain parameters in the digital twin model constructed in the third step, including contact stiffness, damping coefficient and boundary conditions. Step 4.2, Digital Twin Model Update: Based on the various physical monitoring data preprocessed in the second step and the digital twin model calibrated in step 4.1, the real-time collected physical monitoring data and the prediction output of the digital twin model are dynamically fused using the Kalman filter algorithm to realize the digital twin model update, including updating the digital twin model state, digital twin model parameters and digital twin model structure. The fifth step is to conduct virtual vibration and shock testing and evaluation. Based on the valve vibration and shock test digital twin layer processed in the fourth step, various virtual vibration and shock test scenarios are defined on the platform of the digital twin model, virtual vibration and shock tests are executed, and the global response and key part information of the physical entity are predicted and evaluated based on the test results. The sixth step involves using the virtual vibration and shock test results obtained in the fifth step to achieve interaction and performance prediction.

2. The method for valve vibration and impact testing based on digital twins according to claim 1, characterized in that, The first step is specifically as follows: Step 1.1: Install the target valve on the vibration and shock test bench and connect it to the external pipeline to build a physical test bench; Step 1.2: Based on the physical test bench built in Step 1.1, deploy a multi-source sensor network at key locations of the target valve, external piping, and support structure. These key locations primarily include the valve body, valve cover, valve stem, valve core, actuator, and upstream and downstream pipe flanges. The deployment of the multi-source sensor network specifically includes: High-frequency accelerometers are installed around the valve body, at the top of the valve stem, and at the base of the valve seat to capture small-amplitude high-frequency vibration and impact signals. Speed ​​sensors are installed around the valve body, on the actuator housing, and on the upstream and downstream pipe flanges to monitor low-frequency vibration signals and forced vibration signals. Dynamic strain gauges are installed at stress concentration points on the valve stem, at the root of the valve seat, and at the valve cover-valve body connection to measure vibration stress and strain. Acoustic emission sensors are installed at stress concentration points on the valve stem, at the root of the valve seat, at the valve body-external pipeline connection, and at the valve cover-bolt connection to monitor abnormal acoustic signals. Temperature sensors are installed on the valve seat sealing surface contact area and valve stem to monitor the temperature rise caused by friction; A pressure sensor is installed at the valve core to monitor the fluid acceleration impact force; Displacement sensors are installed on the valve stem, valve seat, and actuator push rod to measure actual displacement and minute deformation; The flow meter is installed at the inlet and outlet of the external pipeline to monitor the actual flow rate; A thrust sensor is installed at the actuator push rod to monitor the actuator's status.

3. The method for valve vibration and impact testing based on digital twins according to claim 2, characterized in that, The second step is as follows: Step 2.1: Real-time, synchronous, and high-precision acquisition of various physical monitoring data based on a multi-source sensor network; Step 2.2: Based on the various physical monitoring data obtained in Step 2.1, perform deep cleaning to remove outliers, reduce noise, and remove missing interpolated data from the physical monitoring data to obtain the cleaned data; Step 2.3: Based on the cleaned data obtained in Step 2.2, time synchronization and alignment are achieved by compensating for time shift. Step 2.4: Based on the aligned data obtained in Step 2.3, feature extraction is performed, including time-domain feature set, frequency-domain feature set, and time-frequency-domain feature set; the time-domain feature set includes RMS value, peak value, and kurtosis; the frequency-domain feature set includes spectrum, envelope spectrum, and cepstrum; the time-frequency-domain feature set is wavelet transform.

4. The method for valve vibration and impact testing based on digital twins according to claim 3, characterized in that, Step 3.2 specifically involves: The establishment of material properties involves accurately inputting the mechanical properties of each component material at the service temperature, including elastic modulus, Poisson's ratio, density, yield strength, tensile strength, and thermophysical properties, including thermal conductivity, specific heat capacity, and coefficient of thermal expansion. The establishment of fluid properties: defining the physical properties of the medium, including density, viscosity, and compressibility; The establishment of the boundary condition library involves creating boundary condition templates for typical operating conditions, including steady-state, opening and closing transient, water hammer, seismic testing, and safety valve discharge.

5. The method for valve vibration and impact testing based on digital twins according to claim 4, characterized in that, In step 4.2: The digital twin model state update estimates the current true "hidden state" of the target valve, including internal stress, microcrack propagation, and contact state. The digital twin model parameters are updated online, including local stiffness degradation, damping changes, and friction coefficient, so that the prediction results of the digital twin model can approximate the actual response of the physical entity as closely as possible. The digital twin model results are updated as follows: when a significant performance drift or impairment is detected, model reconstruction is triggered, and the model is replaced with a suitable reduced-order model.

6. The method for valve vibration and impact testing based on digital twins according to claim 4, characterized in that, The sixth step is as follows: Step 6.1, Visualization and Interaction: The state of the digital twin model, such as real-time response, stress cloud map, and prediction results, is superimposed onto the target valve through a 3D visualization platform; Step 6.2, Anomaly Diagnosis and Early Warning: By fusing and analyzing the characteristics of physical monitoring data and the predicted characteristics after virtual vibration and impact testing, and using machine learning-driven intelligent diagnostic algorithms, early detection, type identification, severity assessment, and location of faults are achieved; when an anomaly is detected or the predicted damage approaches a set threshold, multi-level early warnings are automatically triggered. Step 6.3, Condition Prediction and Assessment: The processed digital twin model is used to predict the internal conditions of the valve that are difficult to measure directly, including the contact stress distribution on the sealing surface, stress concentration at the valve stem root, and the microcrack propagation trend. At the same time, fatigue damage accumulation analysis and prediction are performed based on the fatigue damage model constructed in step 3 to assess the remaining service life of the valve's key components.

7. A valve vibration and impact testing system based on digital twins, characterized in that, The valve vibration and impact testing system described in the above-mentioned valve vibration and impact testing system implements the valve vibration and impact testing method based on digital twin as described in any one of claims 1-6, including a physical sensing module, a digital twin model management module, a data-driven engine module, a virtual testing module, and an intelligent application and interaction module.

8. A valve vibration and impact testing system based on digital twin according to claim 7, characterized in that, The valve vibration and impact testing system is specifically as follows: The physical sensing module includes a target valve physical test bench, a multi-source sensor network, and a data acquisition and processing unit. By deploying a multi-source sensor network on the physical test bench, the data acquisition unit collects physical monitoring data with high precision, multiple channels, and synchronously, and the data is preprocessed in the data processing unit. The digital twin model management module is connected to the physical sensing module. It is used to integrate and call the solver of the digital twin model to solve the model, and to implement the order reduction algorithm using the model order reduction toolkit. At the same time, it completes the storage, loading and scheduling of the full-order model and the order reduction proxy model, and realizes the dynamic virtual mapping of the physical entity layer of valve vibration and impact testing. The data-driven engine module connects to the digital twin model management module and the physical sensing module, and achieves real-time data assimilation with the physical test bench by calibrating and updating the digital twin model. The virtual testing module is connected to the data-driven engine module, the digital twin model management module, and the physical sensing module. It is used to manage the definition of virtual testing scenarios, load application, submission and monitoring of virtual vibration and shock test tasks, and reading of test results, so as to realize the prediction and evaluation of the global response and key part information of physical entities. The intelligent application and interaction module is connected to the virtual testing module, data-driven engine module, digital twin model management module, and physical perception module to display physical data, digital twin status, analysis results, and early warning information, thereby enabling valve fault diagnosis and life prediction.

Images