Non-intrusive intelligent boiler thick wall stress monitoring method and system

Abstract

The invention discloses a non-intrusive intelligent boiler thick wall stress monitoring method and system, and belongs to the technical field of boiler stress monitoring, and the method comprises the steps: collecting and preprocessing multi-source data of a boiler thick wall; after the stress distribution of the thick wall of the boiler is independently calculated through each data source, a data verification rule is established to carry out comparison verification on the stress distribution of different data sources. If the stress distribution of the data sources is matched with each other, comprehensive stress calculation is carried out based on multi-source data, comprehensive stress distribution of the boiler thick wall is obtained, whether the boiler thick wall is abnormal or not is determined according to the comprehensive stress distribution, and early warning is carried out when abnormity occurs; and if not, further analyzing based on a matching result to determine whether abnormity exists and performing corresponding early warning. According to the method, multi-source data are fused, the accuracy and reliability of stress monitoring are improved through a data verification rule, potential safety hazards of the thick wall of the boiler are effectively found, a scientific basis is provided for safe operation of the boiler, and the method has important practical application value.

Description

Technical Field

[0001] This invention relates to the field of boiler stress monitoring technology, and more specifically to a non-invasive intelligent boiler thick-wall stress monitoring method and system. Background Technology

[0002] As core equipment in industrial production and energy supply, the safe and stable operation of boilers is crucial for ensuring production efficiency and personnel safety. During long-term operation, thick-walled components of boilers are subjected to high temperatures, high pressures, and complex mechanical stresses. The stress state of these thick-walled components directly affects the structural integrity and service life of the boiler. Therefore, real-time and accurate monitoring of the stress in the boiler's thick walls, and timely detection of potential stress concentrations and structural damage, is a key step in preventing boiler malfunctions and accidents.

[0003] Currently, traditional methods for monitoring stress in boiler thick walls mostly employ invasive techniques, such as installing strain gauges and stress sensors on the boiler wall. While these methods can obtain some stress data, they have significant limitations. First, invasive installation requires damaging the boiler structure, increasing installation difficulty and cost, and potentially causing further damage during installation. Second, invasive sensors have a limited lifespan; they are prone to failure under harsh conditions such as high temperature, high pressure, and corrosion, leading to reduced reliability of the monitoring data. Furthermore, a single type of sensor can only acquire stress information from one aspect, making it difficult to comprehensively reflect the actual stress state of the boiler's thick walls and failing to meet the need for accurate assessment of boiler safety operation.

[0004] With the development of information technology, multi-source data fusion technology has been gradually applied in the field of structural health monitoring. Image processing technology can intuitively reflect the deformation and crack information on the boiler surface; acoustic emission signals can capture the generation and propagation of micro-cracks inside the material; vibration signals contain dynamic information about the equipment's operating status; and temperature data is closely related to thermal stress. However, most existing studies focus on the analysis of single or a few types of data, and a non-invasive intelligent monitoring method that effectively integrates image processing, acoustic emission signals, vibration signals, and temperature data has not yet been developed. This makes it impossible to fully leverage the synergistic advantages of multi-source data and achieve comprehensive and accurate monitoring of boiler thick-wall stress.

[0005] Therefore, how to provide a method and system that can combine various data information to achieve real-time, comprehensive, and accurate monitoring of boiler thick-wall stress without damaging the boiler structure is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a non-invasive intelligent boiler thick-wall stress monitoring method and system. Through the collaborative analysis of image processing, acoustic emission signals, vibration signals, and temperature data, real-time, comprehensive, and accurate monitoring of boiler thick-wall stress is achieved.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: On one hand, the present invention provides a non-invasive intelligent boiler thick-wall stress monitoring method, comprising the following steps: Multi-source data on the thick walls of the boiler were collected separately and preprocessed. Based on the multi-source data, the stress values ​​of the boiler wall thickness under different data sources are calculated to obtain the stress distribution under different data sources; Establish data verification rules, compare the stress distributions calculated from different data sources. If the stress distributions calculated from each data source match each other, perform comprehensive stress calculation based on the multi-source data to obtain the comprehensive stress distribution of the boiler wall. Based on the comprehensive stress distribution, determine whether there is an anomaly in the boiler wall and issue an early warning when an anomaly occurs. If the stress distributions calculated from each data source do not match, further analyze the matching results to determine whether there is an anomaly in the boiler wall and issue an early warning when an anomaly occurs.

[0008] Preferably, the multi-source data includes: image data, acoustic emission signal data, vibration signal data, and temperature data under different operating conditions.

[0009] Preferably, based on the multi-source data, a comprehensive stress calculation is performed to obtain the comprehensive stress distribution of the boiler's thick wall, including: Extract stress characteristics from each data source; The false detection rate of different data sources is calculated based on historical multi-source data. The false detection rate is used as the fusion weight to integrate the stress features extracted from different data sources and generate a multi-dimensional fusion feature vector. The multidimensional fusion feature vector is trained using the support vector machine algorithm. The known stress states and corresponding multidimensional fusion feature vectors in historical multi-source data are used as training samples to establish a mapping relationship model between the boiler thick-wall stress state and the features of multi-source data. By using a trained mapping model to analyze the real-time collected and fused multi-source data of the boiler thick wall, the comprehensive stress distribution of the boiler thick wall is obtained.

[0010] Preferably, the process of calculating the false detection rate of different data sources based on historical multi-source data includes: N historical monitoring data points, including monitoring data number, monitoring data, whether an anomaly was detected, and the actual anomaly; The historical false positive rate is calculated using the following formula: FPR=FA / N In the formula, FA represents the number of error detections, and N represents the number of detections.

[0011] Preferably, based on the matching results, further analysis is performed to determine whether there are any abnormalities in the boiler's thick wall, and an early warning is issued when an abnormality is detected, including: The presence of anomalies in the boiler wall is determined based on the matching stress distributions. An early warning is issued when an anomaly is detected. If no anomaly is detected, an anomaly is determined for the mismatched stress distributions. If there are anomalies in the mismatched stress distribution, time series analysis is used to analyze the matching stress distribution to predict the stress change trend and obtain the predicted stress distribution; if there are no anomalies, no warning is issued. An early warning is issued when the predicted stress distribution matches the mismatched stress distribution; otherwise, no early warning is issued.

[0012] Preferably, the established data verification rules include: geometric data consistency verification and physical data consistency verification.

[0013] Preferably, the geometric data consistency verification process includes: The spatial coverage of stress distribution from one data source in a multi-source dataset is obtained, and the spatial coverage is partitioned to obtain spatial grid data corresponding to the stress distribution of the data source within each spatial unit. Statistical processing is performed on the stress distribution and spatial grid data of the data source to obtain the temporal and spatial distribution characteristic parameters of the stress distribution and the spatial grid data of the data source. Using the temporal and spatial distribution characteristic parameters of the stress distribution from the aforementioned data source, as well as the temporal and spatial distribution characteristic parameters of the spatial grid data, as standard parameters, a temporal and spatial distribution consistency test is performed on the stress distribution from the remaining data sources.

[0014] Preferably, the physical data consistency verification includes: The stress distribution of one data source from the multi-source data was selected as the standard value; Based on historical data, determine the consistency error range between stress values ​​from various data sources; Calculate the difference between the stress value and the standard value for each data point in the stress distribution; Use the aforementioned consistency error range to determine whether each data point meets the consistency criteria.

[0015] Preferably, the stress distribution is matched using time series analysis to predict the stress change trend and obtain the predicted stress distribution, including: The stress time series data of the boiler thick wall is divided into multiple patches along the time dimension; a learnable pseudo-timestamp is introduced for each data patch, and the initial value of the pseudo-timestamp is determined according to the attention score corresponding to each patch; the pseudo-timestamp is dynamically updated through a patch attention mechanism. The outputs of each patch are connected according to the pre-set gating rules in the pseudo-timestamp cyclic connection to obtain the connected feature sequence; The feature sequence is input into a multi-layer patch attention structure stacked in a triangular manner, and time series processing including self-attention mechanism and feedforward neural network is iteratively applied at multiple time scales with different time granularities. The outputs of each layer in the multi-layer patch attention structure, as well as the outputs at different time scales, are aggregated and input into the fully connected neural network predictor. An error loss function is used to iteratively train the fully connected neural network predictor, continuously optimizing the model parameters to obtain a well-trained stress prediction model. By inputting the matched stress distributions into the trained stress prediction model, the predicted stress distribution is obtained.

[0016] On the other hand, the present invention provides a non-invasive intelligent boiler thick-wall stress monitoring system, comprising: The data acquisition module is used to collect multi-source data on the thick walls of the boiler and perform preprocessing. The stress distribution module is used to calculate the stress distribution under different data sources based on the multi-source data. The data matching module is used to compare stress distributions calculated from different data sources based on data validation rules; The data fusion module is used to perform comprehensive stress calculation based on the multi-source data when the stress distributions calculated from various data sources match each other, thereby obtaining the comprehensive stress distribution of the boiler's thick wall. The decision analysis module is used to perform further analysis based on the matching results when the stress distributions calculated from different data sources do not match. The early warning module is used to determine whether there is an abnormality in the boiler wall based on the comprehensive stress distribution and further analysis results, and to issue an early warning when an abnormality occurs.

[0017] Preferably, the data matching module includes: The geometric data verification unit is used to verify the spatiotemporal consistency of stress distribution from various data sources. The physical data verification unit is used to verify the consistency of stress distribution data from various data sources.

[0018] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a non-invasive intelligent boiler thick-wall stress monitoring method and system, which has the following beneficial effects: (1) By collecting multi-source data of the boiler thick wall and performing comprehensive processing and analysis, compared with monitoring from a single data source, it can more comprehensively and accurately reflect the actual stress state of the boiler thick wall, effectively improving the accuracy of stress monitoring.

[0019] (2) The stress distributions calculated from different data sources are compared and verified. Only when the stress distributions of each data source match each other or when further analysis confirms that there are no abnormalities, can the monitoring results be considered reliable. This reduces the risk of misjudgment and improves the credibility of the monitoring results.

[0020] (3) Based on the comprehensive stress distribution and further analysis results, it is possible to determine in a timely and accurate manner whether there are any abnormalities in the boiler wall and to issue an early warning when abnormalities occur, which provides a strong guarantee for the safe operation of the boiler and helps to take measures in advance to avoid accidents and reduce equipment damage and maintenance costs.

[0021] (4) Extract the stress characteristics of each data source, calculate the false detection rate of different data sources based on historical multi-source data, use the false detection rate as the fusion weight to perform multi-source data fusion, generate a multi-dimensional fusion feature vector, give full play to the advantages of each data source, and further improve the accuracy of stress state assessment.

[0022] (5) The support vector machine algorithm is used to train the multi-dimensional fusion feature vector to establish a mapping relationship model between the boiler thick wall stress state and multi-source data features, realizing real-time and accurate analysis of boiler thick wall stress. At the same time, advanced sequence analysis methods are used to predict the matched stress distribution, which can predict the stress change trend in advance and provide more time for taking measures in advance. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the monitoring method provided by the present invention.

[0025] Figure 2 The present invention provides a structural framework diagram of a point monitoring system. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] This invention discloses a non-invasive intelligent boiler thick-wall stress monitoring method, such as... Figure 1 As shown, it includes the following steps: Multi-source data on the thick walls of the boiler were collected separately and preprocessed. The stress values ​​of the thick wall of the boiler were calculated based on multi-source data from different data sources to obtain the stress distribution under different data sources. Establish data verification rules to compare the stress distributions calculated from different data sources. If the stress distributions calculated from each data source match each other, perform comprehensive stress calculation based on multi-source data to obtain the comprehensive stress distribution of the boiler wall. Based on the comprehensive stress distribution, determine whether there are any anomalies in the boiler wall and issue an early warning when an anomaly is found. If the stress distributions calculated from each data source do not match, further analyze the matching results to determine whether there are any anomalies in the boiler wall and issue an early warning when an anomaly is found.

[0028] Specifically, the multi-source data includes: image data, acoustic emission signal data, vibration signal data, and temperature data under different operating conditions. High-resolution cameras or infrared thermal imagers are used to capture the temperature distribution and deformation of the boiler's thick-walled surface, generating high-precision image data. Acoustic emission sensors are used to monitor elastic wave signals generated during boiler operation, capturing real-time information on microcrack propagation and stress changes within the material. Vibration sensors are installed to measure the vibration frequency and amplitude of the boiler's thick-walled structure under different operating conditions, analyzing the structural dynamic response. Thermocouples or fiber optic temperature sensors are used to monitor the temperature distribution of the boiler's thick-walled structure in real time, understanding changes in thermal stress.

[0029] Preprocessing of the collected data includes: noise reduction of the raw data, using filtering algorithms to remove high-frequency noise and interference signals, and improving data quality.

[0030] For image data, grayscale and binarization processing are performed, and the contour and feature points of the boiler's thick wall are extracted using an edge detection algorithm for subsequent analysis.

[0031] For acoustic emission signal data, feature extraction is performed to calculate key parameters such as signal amplitude, energy, and duration, reducing the amount of data and highlighting important information.

[0032] For vibration signal data, frequency domain analysis is performed, and frequency features of the vibration signal are extracted by Fast Fourier Transform (FFT) to identify potential abnormal vibration modes.

[0033] For temperature data, temporal and spatial smoothing is performed to eliminate the impact of local temperature fluctuations and generate a more accurate temperature field distribution map.

[0034] Furthermore, based on multi-source data, comprehensive stress calculations are performed to obtain the comprehensive stress distribution of the boiler's thick wall, including: Stress characteristics are extracted from various data sources. Specifically, from image data, image processing algorithms are used to extract features such as surface deformation, displacement vector, and temperature distribution gradient of the boiler's thick wall. From acoustic emission signal data, parameters such as signal amplitude, rise time, duration, and energy are calculated to reflect changes in internal material stress. For vibration signal data, time-domain and frequency-domain analysis is used to extract vibration acceleration, velocity, frequency components, and amplitude modulation features, which are then correlated with the dynamic stress response of the structure. From temperature data, temperature gradient, heat flux, and local temperature anomaly rate of change are obtained to quantify the impact of thermal stress. These characteristic parameters characterize the stress state of the boiler's thick wall from different dimensions.

[0035] False detection rates from different data sources are calculated based on historical multi-source data. These false detection rates are then used as fusion weights to integrate stress features extracted from different data sources, generating a multi-dimensional fusion feature vector. Specifically, the reciprocal of the false detection rate is taken and normalized to determine the weight of each data source in the fusion process; data sources with lower false detection rates are considered more reliable and thus have larger weights. Based on these weights, the stress features extracted from different data sources are weighted and integrated to generate a multi-dimensional fusion feature vector, comprehensively integrating boiler thick-wall stress-related information.

[0036] The support vector machine algorithm is used to train the multidimensional fusion feature vector. The known stress state and the corresponding multidimensional fusion feature vector in the historical multi-source data are used as training samples to establish a mapping relationship model between the stress state of the boiler thick wall and the features of multi-source data. By using a trained mapping model to analyze the real-time collected and fused multi-source data of the boiler thick wall, the comprehensive stress distribution of the boiler thick wall is obtained.

[0037] Furthermore, the process of calculating the false detection rate of different data sources based on historical multi-source data includes: N historical monitoring data points, including monitoring data number, monitoring data, whether an anomaly was detected, and the actual anomaly; The historical false positive rate is calculated using the following formula: FPR=FA / N In the formula, FA represents the number of error detections, and N represents the number of detections.

[0038] In another embodiment, based on the matching results, further analysis is performed to determine whether there are any anomalies in the boiler's thick wall. When an anomaly is detected, an early warning is issued, including: The presence of anomalies in the boiler wall is determined based on the matching stress distributions. An early warning is issued when an anomaly is detected. If no anomaly is detected, an anomaly is determined for the mismatched stress distributions. If there are anomalies in the mismatched stress distribution, time series analysis is used to analyze the matching stress distribution to predict the stress change trend and obtain the predicted stress distribution; if there are no anomalies, no warning is issued. Time series analysis is a powerful tool that can reveal trends, periodicity, and regularity in data over time. By performing time series analysis on matched stress distributions, we can predict stress change trends and generate predicted stress distribution maps.

[0039] An early warning is issued when the predicted stress distribution matches the mismatched stress distribution; otherwise, no warning is issued. If the predicted stress distribution matches the mismatched stress distribution, it indicates that the mismatched stress distribution has a stronger anomaly detection capability than the matched stress distribution, suggesting an anomaly exists or is about to occur in the boiler's thick wall structure, requiring timely warning to facilitate necessary maintenance measures. If the predicted stress distribution does not match the mismatched stress distribution, it indicates an error in the monitoring data, and the current operating status of the boiler's thick wall structure is normal, reducing false alarms.

[0040] Specifically, the established data verification rules include: geometric data consistency verification and physical data consistency verification.

[0041] Furthermore, the geometric data consistency verification process includes: The process involves acquiring the spatial coverage of stress distribution from one data source within a multi-source dataset. This spatial coverage is then partitioned to obtain spatial mesh data corresponding to the stress distribution from the data source within each spatial cell. The purpose of partitioning is to divide the entire spatial coverage into multiple smaller, manageable spatial cells. The spatial mesh data corresponding to the stress distribution from the data source within each spatial cell is extracted. This spatial mesh data can be pixel-level image data or more complex 3D mesh data, depending on the type of data being processed and the required precision.

[0042] Statistical processing is performed on the stress distribution and spatial grid data from the data source to obtain the temporal and spatial distribution characteristic parameters of both the stress distribution and the spatial grid data. The purpose of this step is to obtain the temporal and spatial distribution characteristic parameters of both the stress distribution and the spatial grid data. Temporal distribution characteristic parameters may include the rate of change, frequency, and magnitude of stress over time, while spatial distribution characteristic parameters may include the stress gradient, concentration, and uniformity in space. These parameters help us comprehensively understand the characteristics of stress distribution from both temporal and spatial dimensions.

[0043] Using the temporal and spatial distribution characteristic parameters of the stress distribution from the data source, as well as the temporal and spatial distribution characteristic parameters of the spatial grid data, as standard parameters, the spatiotemporal distribution consistency of the stress distribution from the remaining data sources is checked.

[0044] Furthermore, physical data consistency verification includes: The stress distribution of one data source from the multi-source data was selected as the standard value; Based on historical data, determine the consistency error range between stress values ​​from various data sources; Calculate the difference between the stress value and the standard value for each data point in the stress distribution; Use the consistency error range to determine whether each data point meets the consistency criteria.

[0045] Furthermore, by using time series analysis to match the stress distribution, the stress change trend is predicted, resulting in a predicted stress distribution, including: For the stress time series data of thick boiler walls, the data is divided into multiple patches along the time dimension; a learnable pseudo-timestamp is introduced for each data patch, and the initial value of the pseudo-timestamp is determined according to the attention score corresponding to each patch; the pseudo-timestamp is dynamically updated through a patch attention mechanism. The outputs of each patch are connected according to the pre-defined gating rules in the pseudo-timestamp circular connection to obtain the connected feature sequence; The feature sequence is input into a multi-layered patch attention structure stacked in a triangular manner, and time series processing including self-attention mechanism and feedforward neural network is iteratively applied at multiple time scales with different time granularities. The outputs of each layer in the multi-layer patch attention structure, as well as the outputs at different time scales, are aggregated and input into the fully connected neural network predictor. An error loss function is used to iteratively train the fully connected neural network predictor, continuously optimizing the model parameters to obtain a well-trained stress prediction model. By inputting the matched stress distributions into the trained stress prediction model, the predicted stress distribution is obtained.

[0046] On the other hand, the present invention provides a non-invasive intelligent boiler thick-wall stress monitoring system, such as Figure 2 As shown, it includes: The data acquisition module is used to collect multi-source data on the thick walls of the boiler and perform preprocessing. The stress distribution module is used to calculate the stress distribution under different data sources based on multi-source data. The data matching module is used to compare stress distributions calculated from different data sources based on data validation rules; The data fusion module is used to perform comprehensive stress calculation based on multi-source data when the stress distributions calculated from various data sources match each other, thereby obtaining the comprehensive stress distribution of the boiler's thick wall. The decision analysis module is used to perform further analysis based on the matching results when the stress distributions calculated from different data sources do not match. The early warning module is used to determine whether there are any abnormalities in the boiler's thick wall based on the comprehensive stress distribution and further analysis results, and to issue an early warning when an abnormality occurs.

[0047] Furthermore, the data matching module includes: The geometric data verification unit is used to verify the spatiotemporal consistency of stress distribution from various data sources. The physical data verification unit is used to verify the consistency of stress distribution data from various data sources.

[0048] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0049] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A non-invasive intelligent boiler thick-wall stress monitoring method, characterized in that, Includes the following steps: Multi-source data on the thick walls of the boiler were collected separately and preprocessed. Based on the multi-source data, the stress values ​​of the boiler wall thickness under different data sources are calculated to obtain the stress distribution under different data sources; Establish data verification rules, compare the stress distributions calculated from different data sources, and if the stress distributions calculated from each data source match each other, then perform comprehensive stress calculation based on the multi-source data to obtain the comprehensive stress distribution of the boiler wall. Based on the comprehensive stress distribution, determine whether there is an anomaly in the boiler wall and issue an early warning when an anomaly occurs. If the stress distributions calculated from different data sources do not match, further analysis is conducted based on the matching results to determine whether there are any abnormalities in the boiler wall thickness, and an early warning is issued when an abnormality occurs.

2. The non-invasive intelligent boiler thick-wall stress monitoring method according to claim 1, characterized in that, The multi-source data includes: image data, acoustic emission signal data, vibration signal data, and temperature data under different operating conditions.

3. The non-invasive intelligent boiler thick-wall stress monitoring method according to claim 2, characterized in that, Based on the multi-source data, a comprehensive stress calculation is performed to obtain the comprehensive stress distribution of the boiler's thick wall, including: Extract stress characteristics from each data source; The false detection rate of different data sources is calculated based on historical multi-source data. The false detection rate is used as the fusion weight to integrate the stress features extracted from different data sources and generate a multi-dimensional fusion feature vector. The multidimensional fusion feature vector is trained using the support vector machine algorithm. The known stress states and corresponding multidimensional fusion feature vectors in historical multi-source data are used as training samples to establish a mapping relationship model between the boiler thick-wall stress state and the features of multi-source data. By using a trained mapping model to analyze the real-time collected and fused multi-source data of the boiler thick wall, the comprehensive stress distribution of the boiler thick wall is obtained.

4. The non-invasive intelligent boiler thick-wall stress monitoring method according to claim 3, characterized in that, The process of calculating the false detection rate of different data sources based on historical multi-source data includes: N historical monitoring data points, including monitoring data number, monitoring data, whether an anomaly was detected, and the actual anomaly; The historical false positive rate is calculated using the following formula: FPR=FA / N In the formula, FA represents the number of error detections, and N represents the number of detections.

5. The non-invasive intelligent boiler thick-wall stress monitoring method according to claim 1, characterized in that, Further analysis based on the matching results determines whether there are any abnormalities in the boiler's thick wall. When an abnormality is detected, an early warning is issued, including: The presence of anomalies in the boiler wall is determined based on the matching stress distributions. An early warning is issued when an anomaly is detected. If no anomaly is detected, an anomaly is determined for the mismatched stress distributions. If there are anomalies in the mismatched stress distribution, time series analysis is used to analyze the matching stress distribution to predict the stress change trend and obtain the predicted stress distribution; if there are no anomalies, no warning is issued. An early warning is issued when the predicted stress distribution matches the mismatched stress distribution; otherwise, no early warning is issued.

6. The non-invasive intelligent boiler thick-wall stress monitoring method according to claim 1, characterized in that, The established data validation rules include: geometric data consistency verification and physical data consistency verification.

7. The non-invasive intelligent boiler thick-wall stress monitoring method according to claim 6, characterized in that, The geometric data consistency verification process includes: The spatial coverage of stress distribution from one data source in a multi-source dataset is obtained, and the spatial coverage is partitioned to obtain spatial grid data corresponding to the stress distribution of the data source within each spatial unit. Statistical processing is performed on the stress distribution and spatial grid data of the data source to obtain the temporal and spatial distribution characteristic parameters of the stress distribution and the spatial grid data of the data source. Using the temporal and spatial distribution characteristic parameters of the stress distribution from the aforementioned data source, as well as the temporal and spatial distribution characteristic parameters of the spatial grid data, as standard parameters, a temporal and spatial distribution consistency test is performed on the stress distribution from the remaining data sources.

8. The non-invasive intelligent boiler thick-wall stress monitoring method according to claim 6, characterized in that, The physical data consistency verification includes: The stress distribution of one data source from the multi-source data was selected as the standard value; Based on historical data, determine the consistency error range between stress values ​​from various data sources; Calculate the difference between the stress value and the standard value for each data point in the stress distribution; Use the aforementioned consistency error range to determine whether each data point meets the consistency criteria.

9. A non-invasive intelligent boiler thick-wall stress monitoring method according to claim 5, characterized in that, By using time series analysis to match the stress distribution, the stress change trend is predicted, and the predicted stress distribution is obtained, including: The stress time series data of the boiler thick wall is divided into multiple patches along the time dimension; a learnable pseudo-timestamp is introduced for each data patch, and the initial value of the pseudo-timestamp is determined according to the attention score corresponding to each patch; the pseudo-timestamp is dynamically updated through a patch attention mechanism. The outputs of each patch are connected according to the pre-set gating rules in the pseudo-timestamp cyclic connection to obtain the connected feature sequence; The feature sequence is input into a multi-layer patch attention structure stacked in a triangular manner, and time series processing including self-attention mechanism and feedforward neural network is iteratively applied at multiple time scales with different time granularities. The outputs of each layer in the multi-layer patch attention structure, as well as the outputs at different time scales, are aggregated and input into the fully connected neural network predictor. An error loss function is used to iteratively train the fully connected neural network predictor, continuously optimizing the model parameters to obtain a well-trained stress prediction model. By inputting the matched stress distributions into the trained stress prediction model, the predicted stress distribution is obtained.

10. A non-invasive intelligent boiler thick-wall stress monitoring system, characterized in that, include: The data acquisition module is used to collect multi-source data on the thick walls of the boiler and perform preprocessing. The stress distribution module is used to calculate the stress distribution under different data sources based on the multi-source data. The data matching module is used to compare stress distributions calculated from different data sources based on data validation rules; The data fusion module is used to perform comprehensive stress calculation based on the multi-source data when the stress distributions calculated from various data sources match each other, thereby obtaining the comprehensive stress distribution of the boiler's thick wall. The decision analysis module is used to perform further analysis based on the matching results when the stress distributions calculated from different data sources do not match. The early warning module is used to determine whether there is an abnormality in the boiler wall based on the comprehensive stress distribution and further analysis results, and to issue an early warning when an abnormality occurs.

11. A non-invasive intelligent boiler thick-wall stress monitoring system according to claim 10, characterized in that, The data matching module includes: The geometric data verification unit is used to verify the spatiotemporal consistency of stress distribution from various data sources. The physical data verification unit is used to verify the consistency of stress distribution data from various data sources.

Images