Thermal stress fatigue intelligent monitoring method and system based on Green function method

Abstract

The invention provides a thermal stress fatigue intelligent monitoring method and system based on a Green function method, and relates to the technical field of nuclear power equipment or pipeline safety research. Firstly, a Green function of a key position is obtained through a finite element model, and a thermal stress rapid calculation model is established based on the Green function; and generating sample databases in batches by using the rapid calculation model to train the thermal stress prediction agent model. In the online monitoring stage, the collected real-time operation data is subjected to preprocessing and transient recognition and input into the thermal stress prediction agent model, the thermal stress time history is rapidly predicted, and finally fatigue life evaluation and early warning are conducted in combination with mechanical stress. According to the method, a physical mechanism and a data driving technology are fused, the limitation that traditional finite element calculation is low in efficiency and depends on a conservative design transient state is overcome, real-time and accurate online monitoring of thermal stress and residual life of nuclear power plant equipment or pipelines is achieved, and the economy and safety of aging management are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of nuclear power equipment or pipeline safety research technology, specifically to a method and system for intelligent monitoring of thermal stress fatigue based on the Green's function method. Background Technology

[0002] Nuclear power plants contain numerous critical equipment and pipelines, and the accurate calculation and safety assessment of their thermal fatigue damage is one of the core issues in power plant aging and life management. Currently, such analyses typically rely on commercial finite element software to solve for thermal stress and fatigue damage through numerical simulation to estimate remaining service life. However, due to the complexity of equipment structures and long temperature transient processes, such high-fidelity numerical simulations are resource-intensive and time-consuming, resulting in low efficiency. Therefore, existing methods are mainly suitable for the design phase with ample time, and are difficult to use for rapid assessment and response to real-time operational monitoring data of nuclear power plants. Against this backdrop, developing a rapid method for predicting the thermal fatigue life of nuclear power equipment and pipelines has become an urgent industry need.

[0003] Existing patent CN107341322A proposes an online monitoring method for fatigue damage in nuclear-grade equipment and pipelines. This method includes: generating Green's function and structural response under unit load at the analysis location based on finite element software; monitoring fluid temperature and pressure data, using these as inputs, and calculating thermal stress using the Green's function method; further fusing pressure and mechanical load stresses to obtain the total stress field; extracting membrane and bending stress intensities through stress linearization; using the rainflow method to count stress cycles and calculate the fatigue service factor, ultimately achieving real-time output and visualization of cumulative damage.

[0004] Existing patent CN118675773A provides a tiered thermal fatigue monitoring strategy. This method first performs a simplified fatigue assessment based on power plant load data to screen out high-load components; then, it performs a deconservative assessment on the high-load components to identify target components that require detailed analysis; finally, it performs a precise fatigue assessment on the target components to generate tiered monitoring results.

[0005] Although the aforementioned existing technologies have proposed a fast stress solution approach based on the Green's function method, which has improved computational efficiency to some extent, they still have important limitations: First, they have failed to systematically establish a response sample library or database for typical equipment and pipelines, which limits the extended application of this method in plant-wide system-level fatigue monitoring; Second, they lack an intelligent preprocessing mechanism for real-time operating data of nuclear power plants, which affects the accuracy and reliability of stress reconstruction and damage assessment. Summary of the Invention

[0006] Based on the current state of technology, this invention proposes a method and system for intelligent monitoring of thermal stress fatigue based on the Green's function method. This addresses the challenge in the nuclear power industry where low computational efficiency makes online and rapid prediction of thermal stress and assessment of remaining life for equipment and pipelines difficult. The specific technical solution is as follows:

[0007] A method for intelligent monitoring of thermal stress fatigue based on the Green's function method includes the following steps:

[0008] S1. Establish a finite element model based on the design parameters of the target equipment and obtain the Green's function at key locations under unit temperature load.

[0009] S2. Establish a fast calculation model for thermal stress based on the Green's function method;

[0010] S3. Design experiments to generate input samples, apply a fast calculation model to calculate thermal stress time histories in batches, and build a thermal stress sample database.

[0011] S4. Use the sample database to train the surrogate model to quickly predict the thermal stress time history based on transient data at any temperature;

[0012] S5. Collect real-time operating data of the target equipment or pipeline and perform preprocessing;

[0013] S6. Input the preprocessed data into the trained surrogate model to predict the thermal stress time history at key locations.

[0014] S7. Combine the predicted thermal stress with mechanical stress, obtain the remaining service life of the target equipment through fatigue calculation, and integrate it into the online monitoring system for early warning.

[0015] Furthermore, in step S1, a step temperature load is applied to the finite element model for solution, and the time-varying Green's function corresponding to different stress intensities is obtained by linearizing the stress at key locations of the target equipment.

[0016] Furthermore, in step S2, the rapid thermal stress calculation model is based on the principle of linear superposition and is established by combining the Green's function obtained in step S1 with any input temperature time history through convolution integral.

[0017] Furthermore, in step S2, the rapid calculation formula for thermal stress is:

[0018]

[0019] in, Let P be the thermal stress at time t. T represents the equipment boundary temperature value. ref The reference temperature is the zero-stress state. The value of the Green's function at point P in a steady state. The term represents the thermal stress value corresponding to the steady-state temperature; The normalized Green's function that varies with time. Let t be the Green's function of the thermal stress field under step load. d The deadline is γ, and the integral variable is γ. It is the temperature difference between the integration time steps. The term refers to the thermal stress caused by temperature fluctuations.

[0020] Furthermore, in step S3, the Latin hypercube sampling method is used to generate an input parameter sample set, and the thermal stress time history values ​​are calculated in batches using the thermal stress fast calculation model to construct a sample database.

[0021] Furthermore, in the process of constructing the sample database, a sampling verification mechanism is introduced to compare the rapid calculation results of some samples with the high-fidelity simulation results, and to adjust the model or sample generation strategy based on the comparison results.

[0022] Furthermore, in step S4,

[0023] The sample database is divided into a training set, a validation set, and a test set, and the input and output data are preprocessed to standardize them and the standardized parameters are saved.

[0024] A thermal stress prediction proxy model that can map temperature time history to thermal stress time history is trained based on the training set and validation set.

[0025] The model performance was evaluated on the test set and the model was serialized into a standard format file for integration and deployment.

[0026] Furthermore, in step S5, real-time operating data of the target device is collected and preprocessed. The preprocessing includes data cleaning, multi-source data synchronization, feature extraction and transient operation identification, and outputs a structured transient data sequence with operating condition labels.

[0027] Furthermore, the data cleaning includes establishing a statistical anomaly detection model and identifying abnormal data in conjunction with the physical constraints of equipment operation. For the identified abnormal values, interpolation correction or direct removal is used depending on the degree of abnormality.

[0028] Furthermore, the feature extraction integrates features based on physical laws and features based on data; the features based on physical laws are physical parameters related to the transient process of the equipment; the features based on data are statistical and signal features extracted from the operating data.

[0029] Furthermore, the transient operation identification involves intelligently identifying and classifying transient operation conditions into various typical operating conditions, including at least temperature surge / deceleration, pressure alternation, and combined load conditions.

[0030] Furthermore, in step S6:

[0031] The input real-time data is standardized using the standardized parameters saved in step S4.

[0032] The standardized data is input into the trained thermal stress prediction proxy model to predict the thermal stress output.

[0033] Based on the standardized parameters, the thermal stress output is restored to a thermal stress time history with real physical dimensions.

[0034] Furthermore, in step S7:

[0035] The thermal stress time history predicted in step S6 is superimposed with the mechanical stress intensity time history to obtain the total stress time history.

[0036] Based on the total stress time history, and combined with the actual number of cycles corresponding to each stress amplitude and the allowable number of cycles for the material, the cumulative fatigue damage factor is calculated.

[0037] The predicted remaining life of the equipment is assessed based on the cumulative fatigue damage factor, taking into account the equipment's design life.

[0038] Furthermore, the calculation of the cumulative fatigue damage factor adopts Miner's linear cumulative damage rule, and its calculation formula is as follows:

[0039] D=Σ(n i / N i ),

[0040] Where D represents the cumulative fatigue damage factor, n i N represents the number of cycles for each stress amplitude. i This indicates the allowable number of cycles for each stress amplitude.

[0041] This invention also provides a thermal stress fatigue intelligent monitoring system based on the Green's function method, used to implement the above-mentioned thermal stress fatigue intelligent monitoring method based on the Green's function method, comprising:

[0042] The finite element model building module is used to build a finite element model based on the design parameters of the target equipment and obtain the Green's function at key locations by applying a step temperature load.

[0043] The thermal stress calculation engine module is used to establish a fast thermal stress calculation model based on the Green's function, so as to quickly calculate the thermal stress corresponding to any temperature through convolution integral.

[0044] The sample database generation and management module is used to generate input samples through experimental design and call the thermal stress rapid calculation model to generate thermal stress time histories in batches to build a sample database.

[0045] The model training module is used to train a thermal stress prediction proxy model using the sample database.

[0046] The data acquisition and processing module is used to collect real-time operating data of the target device, and to perform data cleaning, multi-source data synchronization, feature extraction, and transient operation identification.

[0047] The online prediction module is used to input the preprocessed real-time data into the thermal stress prediction proxy model to predict the thermal stress time history at key locations.

[0048] The life assessment module is used to superimpose the predicted thermal stress time history and mechanical stress time history, calculate the cumulative fatigue damage factor and remaining life based on fatigue theory, and provide early warning.

[0049] Based on the above technical solution, the present invention has at least the following beneficial effects:

[0050] 1. This invention proposes a method and system for intelligent monitoring of thermal stress fatigue based on the Green's function method. By preprocessing the collected multi-source operational data, sensor anomalies and transmission noise are effectively eliminated, and high-precision data synchronization is achieved based on a unified time source. By introducing unsupervised machine learning algorithms, various typical operational transients can be automatically and intelligently identified and classified from long-term continuous data, and their occurrence frequency can be counted. This provides a high-quality structured data foundation for subsequent stress prediction and life assessment, fundamentally ensuring the accuracy of monitoring input.

[0051] 2. This invention proposes a method and system for intelligent monitoring of thermal stress fatigue based on the Green's function method. The Green's function method transforms complex transient thermo-mechanical coupled finite element simulation into efficient convolution integral operations. By establishing a rapid thermal stress calculation model, its calculation speed is significantly improved compared to traditional high-fidelity finite element analysis. This improvement makes it possible to train high-precision models based on massive samples and enables rapid response to complex, unpredictable transient conditions in actual operation. It overcomes the limitations of traditional design phases that only use finite, conservative transient data for evaluation, enhancing the practicality of this method in online monitoring scenarios.

[0052] 3. This invention proposes a method and system for intelligent monitoring of thermal stress fatigue based on the Green's function method, constructing a dual prediction architecture of "physical model + surrogate model". Using the Green's function method, based on physical laws, as the foundation, a training sample library with both physical realism and large-scale characteristics is generated. Subsequently, data-driven methods such as deep learning are used to train a surrogate model capable of capturing complex nonlinear relationships. This model inherits the physical rationality of the Green's function method while improving computational speed, ultimately achieving real-time, high-precision prediction of the thermal stress time history at key locations.

[0053] 4. This invention proposes a method and system for intelligent monitoring of thermal stress fatigue based on the Green's function method. This method superimposes predicted thermal stress and mechanical stress, and then calculates cumulative fatigue damage. Because this assessment is based on real, full-cycle operational history data, rather than conservative design assumptions, it can more accurately reflect the actual aging state of the equipment and calculate a more realistic remaining lifespan. This provides a direct and scientific basis for life extension assessments, in-service inspection optimization, and preventative maintenance decisions in nuclear power plants, effectively avoiding unnecessary shutdowns and replacement costs caused by excessive conservatism, thereby significantly improving the precision and economy of equipment aging management. Attached Figure Description

[0054] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0055] Figure 1 This is a flowchart illustrating an intelligent monitoring method for thermal stress fatigue based on the Green's function method proposed in this invention.

[0056] Figure 2 This is a schematic diagram of the framework of an intelligent monitoring system for thermal stress fatigue based on the Green's function method proposed in this invention.

[0057] Figure 3 This invention relates to an electronic device. Detailed Implementation

[0058] The present invention will be further described in detail below with reference to specific embodiments, which should not be construed as limiting the scope of protection claimed by the present invention.

[0059] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0060] To address the challenge of low computational efficiency in the nuclear power industry, which hinders online and rapid prediction of thermal stress and assessment of remaining life for equipment and pipelines, this invention proposes a method and system for intelligent monitoring of thermal stress fatigue based on the Green's function method.

[0061] Example 1

[0062] This embodiment proposes a smart monitoring method for thermal stress fatigue based on the Green's function method. (See reference...) Figure 1 As shown, the method includes the following steps:

[0063] S1. Establish a finite element model based on the design parameters of the target equipment and obtain the Green's function at key locations under unit temperature load.

[0064] S2. Establish a fast calculation model for thermal stress based on the Green's function method;

[0065] S3. Design experiments to generate input samples, apply a fast calculation model to calculate thermal stress time histories in batches, and build a thermal stress sample database.

[0066] S4. Use the sample database to train the surrogate model to quickly predict the thermal stress time history based on transient data at any temperature;

[0067] S5. Collect real-time operating data of the target equipment or pipeline and perform preprocessing;

[0068] S6. Input the preprocessed data into the trained surrogate model to predict the thermal stress time history at key locations.

[0069] S7. Combine the predicted thermal stress with mechanical stress, obtain the remaining service life of the target equipment through fatigue calculation, and integrate it into the online monitoring system for early warning.

[0070] The above step S1 specifically includes:

[0071] Specifically, firstly, based on the design drawings and material parameters of the target equipment or pipeline, a high-fidelity three-dimensional finite element model is established using ANSYS finite element software. A temperature step load is then applied to the finite element model for transient thermo-mechanical coupling solution. To ensure the engineering representativeness and safety of the Green's function, the highest temperature value of the applied temperature step load should not be lower than the highest temperature specified in the system design specifications, and the load duration must be long enough until the model reaches steady state to capture the complete stress relaxation process. After the solution is completed, stress linearization is performed on the cross-sectional paths at predefined critical locations to obtain the membrane stress intensity, membrane plus bending stress intensity, and total stress intensity. Critical locations are typically selected based on design experience, stress concentration, and potential fatigue risk areas, such as structural discontinuities, geometric abrupt changes, or high stress gradient regions. The strength curve obtained through the above linearization process represents the stress intensity obtained under the temperature step load. To construct a universal response function applicable to arbitrary temperature loads, the stress intensity time history curve obtained by the above linearization process is normalized to obtain the stress response under a unit temperature step load, i.e., the normalized Green's function, which provides the basis for the stress response under a unit temperature load in subsequent steps.

[0072] In a specific implementation process, for an RCV regenerative heat exchanger in a nuclear power plant, a three-dimensional finite element model was established using ANSYS software based on design parameters provided by upstream suppliers. To obtain the Green's function, a temperature step load was applied to the finite element model, and a transient thermo-mechanical coupling solution was performed to ensure that the calculation duration was sufficient to reach thermal steady state and capture the complete stress relaxation process. After the solution was completed, stress linearization was performed on the critical sections with potential high fatigue risk identified based on design experience and stress analysis, resulting in curves showing the changes in membrane stress, membrane plus bending stress, and total stress intensity over time. After normalization, the Green's function corresponding to the path of the critical section was obtained.

[0073] The above step S2 specifically includes:

[0074] Specifically, a rapid mathematical model for calculating thermal stress based on the Green's function method is established. The theoretical foundation of this model is the superposition principle of linear systems, which decomposes any complex temperature history into a series of linear superpositions of unit steps. Through convolution integration, the Green's function obtained in step S1 is combined with any input temperature history, thereby establishing an efficient and general method for calculating thermal stress. The core mathematical expression of this rapid calculation model is:

[0075]

[0076] in, Let P be the thermal stress at time t. T represents the equipment boundary temperature value. refThe reference temperature is the zero-stress state. The value of the Green's function at point P in a steady state. The term represents the thermal stress value corresponding to the steady-state temperature; The normalized Green's function that varies with time. Let t be the Green's function of the thermal stress field under step load. d The deadline is γ, and the integral variable is γ. It is the temperature difference between the integration time steps. The term refers to the thermal stress caused by temperature fluctuations.

[0077] In practical numerical calculations, the aforementioned continuous convolution integral model is discretized into a summation over finite time steps, thus becoming an algorithm that can run efficiently on a computer. This step establishes a fast computational model driven by the Green's function, which can be used for batch processing of arbitrary temperature histories, providing the core algorithmic tool for the automated construction of a large-scale sample database in the subsequent step S3.

[0078] In a specific implementation, for an RCV regenerative heat exchanger in a nuclear power plant, the normalized Green's function of each key cross-section obtained in step S1 is used as the core parameter and embedded into the aforementioned discrete convolution summation model, thereby completing the construction of a rapid thermal stress calculation model for this specific device. This model takes arbitrary temperature time histories as input and directly outputs the thermal stress time histories at key locations.

[0079] The above step S3 specifically includes:

[0080] Specifically, a thermal stress time history sample database covering all operating conditions is constructed through systematic experimental design. First, based on the historical operating data, design transients, and procedures of the target equipment, key descriptive parameters characterizing the spatiotemporal variation of its temperature load are abstracted, such as transient type, onset / end temperature and time, temperature change rate, and amplitude parameters. Experimental design methods include, but are not limited to, Latin hypercube sampling, full factorial design, and partial factorial design. Based on these methods, a sample set of input parameters with good spatial distribution and representativeness is efficiently generated in the aforementioned multidimensional parameter space. Subsequently, corresponding temperature time history curves are constructed based on these parameters as input to a rapid calculation model.

[0081] The rapid thermal stress calculation model established in step S2 is used to quickly calculate the corresponding thermal stress time history for each sample. To ensure data quality, a sampling verification and closed-loop feedback mechanism is introduced: a certain proportion of samples are randomly selected for comparison through full transient finite element simulation. If the error norm of the comparison results exceeds the preset tolerance, the source of error is analyzed, and the experimental design range is readjusted or the rapid model is calibrated, and the data for that part is regenerated until verification is successful.

[0082] Finally, all the validated input and output data are structured and stored to form a high-quality "temperature-thermal stress" mapping sample database, laying the foundation for the training of subsequent surrogate models.

[0083] In this embodiment, based on the design specifications and historical operating data of an RCV regenerative heat exchanger in a nuclear power plant, its temperature boundary range is determined to be 10℃-293℃, and its temperature change rate boundary range is 1℃ / min-1000℃ / min. A Latin hypercube sampling method is used to generate 3000 sets of temperature load samples within this parameter space. For each set of temperature load samples, the thermal stress fast calculation model established in step S2 is used to calculate the corresponding thermal stress time histories for all samples in batches. To ensure database quality, 70% of the samples are randomly selected for full transient finite element comparison verification to ensure that the calculation error is less than 10%. Finally, all input temperature time histories and output stress time histories are structured and stored to construct a "temperature-thermal stress" mapping sample database, providing a data foundation for subsequent machine learning training.

[0084] The above step S4 specifically includes:

[0085] Specifically, based on the "temperature-thermal stress" mapping sample database constructed in step S3, a surrogate model capable of quickly and accurately predicting thermal stress time histories is trained using machine learning methods. The input temperature time histories and output thermal stress time histories are standardized, and the standardized parameters are saved. The database is divided into training, validation, and test sets according to a certain ratio. Considering the nonlinear characteristics of temperature-thermal stress data, various applicable machine learning model architectures are selected for comparative experiments to effectively capture the relationship between thermal stress and temperature changes. These machine learning models include, but are not limited to, gradient boosting decision trees (such as LightGBM and XGBoost), random forests, temporal convolutional networks, and long short-term memory networks (LSTM). During model training, the performance is monitored using a validation set, and the optimal model architecture and parameter configuration are determined through hyperparameter optimization techniques. Early stopping is employed to prevent overfitting. After model training is complete, it is evaluated on a test set that has never participated in training or tuning. The best-performing surrogate model is serialized and saved in an industry-standard format, forming a computational unit capable of high-precision and high-efficiency thermal stress prediction for any temperature transient data, providing a core engine for subsequent integration into online monitoring systems.

[0086] In this embodiment, the sample database of RCV regenerative heat exchangers, containing 3000 samples, is divided into training, validation, and test sets in a 70%:15%:15% ratio. Comparative experiments were conducted, and the best-performing machine learning model was selected as the surrogate model architecture. During model training, to effectively prevent overfitting, the following strategies were adopted: early stopping driven by validation set performance monitoring, and automatic searching for the optimal hyperparameter combination within a Bayesian optimization framework. After model training, a final evaluation was performed on the test set. The surrogate model achieved a mean absolute percentage error (MASE) of less than 10% and a coefficient of determination (COD) greater than 0.9. Finally, the trained model was serialized and exported in ONNX format. This model has a single prediction time in the millisecond range, excellent cross-platform deployment capabilities, and fully meets the core requirements of online monitoring systems for high accuracy and real-time performance.

[0087] The above step S5 specifically includes:

[0088] Specifically, by deploying sensor systems on target equipment or pipelines in nuclear power plants, key operating parameters required for thermal fatigue calculations are collected in real time, mainly including time-series data such as temperature, pressure, and flow rate, and these raw field data are preprocessed.

[0089] First, data cleaning is performed to establish a statistical anomaly detection model that incorporates physical constraints of equipment operation, such as sensor range and reasonable temperature rise rate range. This identifies and eliminates outliers caused by sensor malfunctions or transmission interference. Minor anomalies are corrected using interpolation, while severe anomalies are directly removed, significantly improving data reliability and consistency. Based on a unified time source, the data streams from all sensors are timestamped to ensure strict synchronization of different physical quantities in time, eliminating errors caused by asynchronous acquisition delays. Building on this, characteristic parameters representing transient loads are extracted, followed by transient identification. Specifically, rule-based or unsupervised machine learning methods are used to segment and intelligently identify operating conditions in continuous operational data. This process aims to automatically identify independent transient events of engineering significance from long-term operational data and classify them into several typical operating conditions. Furthermore, the frequency of occurrence and characteristic parameters of each identified transient type are automatically statistically analyzed. The preprocessing process ultimately outputs a cleaned, synchronized, structured transient data sequence with clear operating condition labels and rich features. This data is then provided to subsequent steps via message queues or databases, laying the foundation for accurate stress prediction and fatigue life assessment using the surrogate model.

[0090] In this embodiment, a nuclear power plant's RCV regenerative heat exchanger is used as an example for data acquisition and preprocessing. Data acquisition is performed at a sampling frequency of 1Hz, continuously acquiring time-series data of temperature, pressure, and flow rate for the four inlet and outlet pipes of the heat exchanger over a continuous 24-month operating cycle. During the data cleaning stage, the raw data is cleaned to remove abnormally high values ​​significantly exceeding the sensor's range, zero or constant values ​​caused by signal interruption, and abnormal temperature rise rates that violate thermophysical laws, ensuring the high quality and reliability of subsequent analysis data. In the feature extraction and transient identification stage, key feature parameters are first extracted from the cleaned data to characterize transient loads, including temperature and mechanical characteristics. Subsequently, clustering algorithms were used to intelligently identify and classify the transients, automatically forming several typical operating condition categories. These included rapid temperature rise / fall categories, characterized by a temperature rise rate greater than 25℃ / min and relatively small pressure fluctuations; alternating pressure categories, characterized by pressure fluctuation amplitudes exceeding 2MPa and temperature rises less than 10℃; and composite load categories, characterized by temperature rises greater than 15℃ and pressure fluctuations greater than 1.5MPa. Finally, the system automatically statistically analyzed the occurrence frequency and core characteristic parameters of each type of transient, forming a structured transient data sequence to provide input for subsequent proxy models.

[0091] The above step S6 specifically includes:

[0092] Specifically, the standardized parameters saved in step S4 are used to standardize the real-time temperature time history data, which is then input into the thermal stress prediction surrogate model trained in step S4. Based on the standardized parameters, the model output is converted into thermal stress with physical dimensions. This process utilizes the powerful nonlinear mapping capability of the surrogate model to transform complex physical simulation into efficient data calculation, achieving real-time and accurate prediction of thermal stress response under any actual operating conditions, providing key input for subsequent fatigue life assessment.

[0093] In this embodiment, for the critical location of the RCV regenerative heat exchanger inlet nozzle, various typical transient temperature data identified through clustering, such as time series data of rapid temperature rises and falls and pressure alternations, are first standardized online using parameters determined during the training phase, and then input into the deployed thermal stress prediction proxy model. After the model output undergoes destandardization, the detailed thermal stress time history curve at this location under the corresponding operating conditions is finally obtained, including the thin film, the thin film plus bending, and the total stress intensity.

[0094] The above step S7 specifically includes:

[0095] Specifically, the thermal stress predicted in step S6 is superimposed with the mechanical stress calculated by the finite element model considering internal pressure, pipe loads, and other mechanical loads. Then, according to the industry standards and specifications required by the target equipment, including but not limited to ASME and RCC-M standards, fatigue calculations are performed on the superimposed total stress time history. The actual number of cycles n corresponding to each stress amplitude is obtained using the rainflow counting method. i And query the allowable number of cycles N corresponding to the material fatigue design curve in the specification. i The cumulative fatigue damage factor D = Σ(n) is calculated using Miner's linear cumulative damage method. i / N i The above calculations, combined with the statistical results of the actual occurrence of various transients in step S5, ensure that the damage assessment is based on real operating history. Finally, based on the cumulative fatigue damage factor D and in conjunction with the equipment's design life, the remaining service life of the equipment or pipeline is predicted. This result needs to be cross-checked with the conclusions of periodic in-service inspections.

[0096] Example 2

[0097] This embodiment provides a thermal stress fatigue intelligent monitoring system based on the Green's function method, used to implement the thermal stress fatigue intelligent monitoring method based on the Green's function method in Embodiment 1. (See reference...) Figure 2 As shown, the system includes a finite element model construction module, a thermal stress calculation engine module, a sample library generation and management module, a model training module, a data acquisition and processing module, an online prediction module, and a life assessment module.

[0098] Among them, the finite element model construction module is the core of the physical foundation of the system. It is responsible for establishing a high-fidelity three-dimensional finite element model based on the equipment design drawings and material parameters. By applying a step temperature load, transient thermo-mechanical coupling simulation is performed, and the Green's function under unit load at key locations is obtained after normalization.

[0099] The thermal stress calculation engine module provides the core computation based on the Green's function method. It establishes a rapid thermal stress calculation model by implementing convolution integrals using the Green's function provided by the finite element construction module. This engine can accept time histories at arbitrary temperatures as input and quickly output the corresponding thermal stress time histories at key locations.

[0100] The sample library generation and management module is responsible for building a sample database for training the thermal stress prediction surrogate model. It automatically generates a large number of representative temperature load samples using experimental design methods and calls the fast calculation module to calculate the corresponding thermal stress in batches. The module also has a built-in quality assurance mechanism that can randomly select samples and automatically compare them with high-fidelity simulation results to ensure data accuracy. Finally, all "temperature-thermal stress" samples are structured, stored, and managed.

[0101] The model training module utilizes a sample database to train, validate, and export a thermal stress prediction surrogate model. Its functions include automated data normalization, model training, hyperparameter optimization, performance evaluation, and model serialization. The final output is an optimized prediction engine file that can be directly integrated into online systems.

[0102] The data acquisition and processing module interfaces with the field sensor system, responsible for the access and preprocessing of real-time data streams. Preprocessing includes data cleaning of raw data and synchronization of multi-source data; using unsupervised machine learning algorithms, it automatically analyzes continuous operating data, extracts characteristic parameters representing the operating state, and then identifies and classifies different typical operating transients, providing structured, high-quality input for stress prediction and life assessment.

[0103] The online prediction module receives real-time temperature data from the data acquisition and processing module, calls the parameters saved during the training phase for online standardization, and then inputs them into the deployed thermal stress prediction proxy model to efficiently complete the prediction calculation and obtain the thermal stress time history at key locations. The output is then converted into stress values ​​with physical dimensions, and the results are passed to the life assessment module.

[0104] The life assessment module superimposes the predicted thermal stress with the mechanical stress calculated through the finite element model to obtain the total stress time history. Based on specifications, the cumulative fatigue damage factor is calculated using Miner's linear cumulative damage rule. Combining historical operating data with the design life, the remaining lifespan of the equipment is assessed in real time, and pre-set warning thresholds are established. When the lifespan is consumed too quickly or approaches the design value, the system automatically triggers alarms of different levels and visualizes the results on the monitoring interface, providing direct evidence for operation and maintenance decisions.

[0105] Example 3

[0106] This embodiment provides an electronic device, which is used as a terminal device for illustration. (See reference...) Figure 3 As shown, the electronic device includes a memory 302 and a processor 304. The memory 302 stores a computer program, and the processor 304 is configured to execute the steps of the intelligent monitoring method for thermal stress fatigue based on the Green's function method in Embodiment 1 above through the computer program.

[0107] Optionally, in this embodiment, the aforementioned electronic device may be located in at least one of a plurality of network devices in a computer network.

[0108] Alternatively, as those skilled in the art will understand, Figure 3 The structure shown is for illustrative purposes only. Figure 3 This does not limit the structure of the aforementioned electronic devices. For example, the electronic device may also include components that are more... Figure 3The more or fewer components shown (such as network interfaces, etc.), or having the same Figure 3 The different configurations shown.

[0109] The memory 302 can be used to store software programs and modules, such as the program instructions / modules corresponding to the intelligent monitoring method for thermal stress fatigue based on the Green's function method in Embodiment 1 of this application. The processor 304 executes various functional applications and data processing by running the software programs and modules stored in the memory 302, thereby realizing the intelligent monitoring method for thermal stress fatigue based on the Green's function method in Embodiment 1. The memory 302 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage systems, flash memory, or other non-volatile solid-state memory. In some instances, the memory 302 may further include memory remotely located relative to the processor 304, and these remote memories can be connected to the terminal via a network. Examples of the above-mentioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. Specifically, the memory 302 may be used for storing data information. In addition, it may include, but is not limited to, other module units in the above system, which will not be described in detail in this example.

[0110] Optionally, the aforementioned electronic device further includes a transmission system 306 for receiving or sending data via a network. Specific examples of the network described above may include wired and wireless networks. In one example, the transmission system 306 includes a Network Interface Controller (NIC), which can be connected to other network devices and a router via a network cable to communicate with the Internet or a local area network. In another example, the transmission system 306 is a radio frequency (RF) module used for wireless communication with the Internet.

[0111] In addition, the above-mentioned electronic device also includes a display 308 and a connection bus 310, which is used to connect the various module components in the above-mentioned electronic device.

[0112] Example 4

[0113] This embodiment provides a computer program product, which includes a computer program. When the computer program is run by an electronic device, it executes the steps of the intelligent monitoring method for thermal stress fatigue based on the Green's function method in Embodiment 1.

[0114] Example 5

[0115] This embodiment provides a computer-readable storage medium. The processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the steps of the intelligent monitoring method for thermal stress fatigue based on the Green's function method provided in Embodiment 1 above.

[0116] Optionally, in this embodiment, the computer-readable storage medium may be configured to store the intelligent monitoring method for thermal stress fatigue based on the Green's function method in Embodiment 1 of this application.

[0117] Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0118] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0119] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0120] It should be noted that, in the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

Claims

1. A method for intelligent monitoring of thermal stress fatigue based on the Green's function method, characterized in that, Includes the following steps: S1. Establish a finite element model based on the design parameters of the target equipment and obtain the Green's function at key locations under unit temperature load. S2. Establish a fast calculation model for thermal stress based on the Green's function method; S3. Design experiments to generate input samples, apply a fast calculation model to calculate thermal stress time histories in batches, and build a thermal stress sample database. S4. Use the sample database to train the surrogate model to quickly predict the thermal stress time history based on transient data at any temperature; S5. Collect real-time operating data of the target equipment or pipeline and perform preprocessing; S6. Input the preprocessed data into the trained surrogate model to predict the thermal stress time history at key locations. S7. Combine the predicted thermal stress with mechanical stress, obtain the remaining service life of the target equipment through fatigue calculation, and integrate it into the online monitoring system for early warning.

2. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 1, characterized in that: In step S1, a step temperature load is applied to the finite element model for solution. By linearizing the stress at key locations of the target equipment, time-varying Green's functions corresponding to different stress intensities are obtained.

3. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 2, characterized in that: In step S2, the rapid thermal stress calculation model is based on the principle of linear superposition and is established by combining the Green's function obtained in step S1 with any input temperature time history through convolution integral.

4. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 3, characterized in that: In step S2, the rapid calculation formula for thermal stress is: in, Let P be the thermal stress at the critical location at time t. T represents the equipment boundary temperature value. ref The reference temperature is the zero-stress state. The value of the Green's function at point P in a steady state. The term represents the thermal stress value corresponding to the steady-state temperature; The normalized Green's function that varies with time. Let t be the Green's function of the thermal stress field under step load. d The deadline is γ, and the integral variable is γ. It is the temperature difference between the integration time steps. The term refers to the thermal stress caused by temperature fluctuations.

5. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 1, characterized in that: In step S3, an input parameter sample set is generated using an experimental design method, and the thermal stress fast calculation model is applied to calculate the thermal stress time history corresponding to each sample; based on the input parameter sample set and its corresponding thermal stress time history, a sample database is constructed.

6. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 5, characterized in that: In the process of building the sample database, a sampling verification mechanism is introduced to compare the rapid calculation results of some samples with the high-fidelity simulation results, and to adjust the model or sample generation strategy based on the comparison results.

7. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 1, characterized in that: In step S4, The sample database is divided into a training set, a validation set, and a test set, and the input and output data are preprocessed to standardize them and the standardized parameters are saved. A thermal stress prediction proxy model that can map temperature time history to thermal stress time history is trained based on the training set and validation set. The model performance was evaluated on the test set and the model was serialized into a standard format file for integration and deployment.

8. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 1, characterized in that: In step S5, real-time operating data of the target device is collected and preprocessed. The preprocessing includes data cleaning, multi-source data synchronization, feature extraction and transient operation identification, and outputs a structured transient data sequence with operating condition labels.

9. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 8, characterized in that: The data cleaning process includes establishing a statistical anomaly detection model and identifying abnormal data by combining it with the physical constraints of equipment operation. For the identified abnormal values, interpolation correction or direct removal is used depending on the degree of abnormality.

10. The intelligent monitoring method for thermal stress fatigue based on the Green's function method according to claim 8, characterized in that: The feature extraction integrates features based on physical laws and features based on data; the features based on physical laws are physical parameters related to the transient process of the equipment; the features based on data are statistical and signal features extracted from the operating data.

11. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 8 or 10, characterized in that: The transient operation identification involves intelligently identifying and classifying transient operation conditions into various typical operating conditions, including at least temperature surge / deceleration, pressure alternation, and combined load conditions.

12. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 1, characterized in that, In step S6: The input real-time data is standardized using the standardized parameters saved in step S4. The standardized data is input into the trained thermal stress prediction proxy model to predict the thermal stress output. Based on the standardized parameters, the thermal stress output is restored to a thermal stress time history with real physical dimensions.

13. The intelligent monitoring method for thermal stress fatigue based on the Green's function method according to claim 12, characterized in that, In step S7: The thermal stress time history predicted in step S6 is superimposed with the mechanical stress intensity time history to obtain the total stress time history. Based on the total stress time history, and combined with the actual number of cycles corresponding to each stress amplitude and the allowable number of cycles for the material, the cumulative fatigue damage factor is calculated. The predicted remaining life of the equipment is assessed based on the cumulative fatigue damage factor, taking into account the equipment's design life.

14. The intelligent monitoring method for thermal stress fatigue based on Green's function method according to claim 13, characterized in that: The cumulative fatigue damage factor is calculated using Miner's linear cumulative damage rule, and its calculation formula is as follows: D=Σ(n i / N i ), Where D represents the cumulative fatigue damage factor, n i N represents the number of cycles for each stress amplitude. i This indicates the allowable number of cycles for each stress amplitude.

15. A thermal stress fatigue intelligent monitoring system based on Green's function method, used to implement the thermal stress fatigue intelligent monitoring method based on Green's function method according to any one of claims 1-14, characterized in that, include: The finite element model building module is used to build a finite element model based on the design parameters of the target equipment and obtain the Green's function at key locations by applying a step temperature load. The thermal stress calculation engine module is used to establish a fast thermal stress calculation model based on the Green's function, so as to quickly calculate the thermal stress corresponding to any temperature through convolution integral. The sample database generation and management module is used to generate input samples through experimental design and call the thermal stress rapid calculation model to generate thermal stress time histories in batches to build a sample database. The model training module is used to train a thermal stress prediction proxy model using the sample database. The data acquisition and processing module is used to collect real-time operating data of the target device, and to perform data cleaning, multi-source data synchronization, feature extraction, and transient operation identification. The online prediction module is used to input the preprocessed real-time data into the thermal stress prediction proxy model to predict the thermal stress time history at key locations. The life assessment module is used to superimpose the predicted thermal stress time history and mechanical stress time history, calculate the cumulative fatigue damage factor and remaining life based on fatigue theory, and provide early warning.

16. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the steps of the intelligent monitoring method for thermal stress fatigue based on the Green's function method according to any one of claims 1 to 14 through the computer program.

17. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the intelligent monitoring method for thermal stress fatigue based on the Green's function method as described in any one of claims 1 to 14.

18. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein the computer program can be executed by an electronic device to perform the steps of the intelligent monitoring method for thermal stress fatigue based on the Green's function method according to any one of claims 1 to 14.

Images