Intelligent decision-making method and system for pressure vessel inspection

Abstract

The application discloses a pressure container inspection intelligent decision method and system, relates to the technical field of equipment inspection, and realizes a fundamental change from static risk assessment to dynamic mechanism driven intelligent decision, ensures that a called damage evolution model has a solid physical basis and is verified in multiple dimensions, can accurately depict the degradation behavior of equipment under complex working conditions, effectively solves bottleneck problems that traditional technologies cannot dynamically respond to equipment states, models lack interpretability, and inspection strategies are rigid, and significantly improves the scientificity and safety of inspection decisions, avoids missed inspection and excessive inspection, and provides reliable support for intelligent and accurate safety management of the whole life cycle of pressure containers.

Description

Technical Field

[0001] This invention relates to the field of equipment inspection technology, and in particular to an intelligent decision-making method and system for pressure vessel inspection. Background Technology

[0002] In practical applications, many pressure vessels have exceeded their design service life or have defects requiring monitoring, significantly increasing the risks they face. Time-related damage mechanisms such as corrosion, fatigue, creep, and material degradation accumulate continuously, making traditional inspection strategies ineffective. Currently, most mainstream pressure vessel inspection methods both domestically and internationally employ risk assessment-based inspection techniques. While this technology has played a significant role in optimizing inspection plans and extending operating cycles, it is essentially a static analysis method. It primarily relies on initial equipment design parameters, media characteristics, and historical experience to classify risk levels, failing to fully consider the dynamic risk evolution caused by factors such as operating condition fluctuations, parameter drift, and environmental changes during actual operation. This static assessment model lacks the ability to perceive and respond to the real-time status of the equipment, easily leading to problems such as unreasonable inspection cycle settings, missed inspections of key risk points, or over-inspection. This not only increases the company's operation and maintenance costs but also makes it difficult to fundamentally guarantee the safe operation of the equipment. In recent years, although some studies have attempted to introduce online monitoring data or life prediction models to enhance the dynamism of testing decisions, there are still obvious limitations: on the one hand, most models are not deeply coupled with specific failure mechanisms, resulting in a lack of physical interpretability in the prediction results; on the other hand, existing methods generally lack a quantitative verification mechanism for the reliability of the damage evolution model itself, making it impossible to determine whether the model used is truly suitable for the actual degradation behavior of the current equipment; in addition, even when multi-source data is introduced, there is often a lack of a systematic intelligent criterion system, making it difficult to comprehensively evaluate and classify the model matching degree in multiple dimensions such as accuracy, trend consistency, and operating condition stability. Therefore, there is an urgent need for technical solutions for intelligent decision-making methods and systems for pressure vessel inspection. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides an intelligent decision-making method for pressure vessel inspection, specifically comprising the following steps: S1. Obtain several failure mechanisms of the pressure vessel, and construct a damage evolution model for each failure mechanism. Perform quantization-based logical verification on the damage evolution model of each failure mechanism, and construct a failure mechanism database containing several failure mechanisms and damage evolution models of each failure mechanism based on the verification results. S1a1. Determine a failure mechanism as the modeling object; S1a2. Based on the current failure mechanism, collect historical inspection data, real-time operation monitoring data, and basic equipment information generated during the actual use of several pressure vessels already in operation. The historical inspection data refers to the damage amount related to the current failure mechanism measured in each pressure vessel during each inspection. The real-time operation monitoring data refers to the operating parameters continuously recorded between two inspections for each pressure vessel. The basic equipment information includes material type, design temperature, design pressure, medium composition, and structural form. S1a3. Based on the physical or chemical action principle of the current failure mechanism, establish a mathematical relationship between the damage amount and time. The mathematical relationship takes time as the independent variable and the damage amount as the dependent variable, and takes the operating parameters in the real-time operation monitoring data as the input variables affecting the damage evolution rate. S1a4. Based on the material type, medium composition, design temperature, design pressure and structural form in the basic information of the equipment, obtain the initial material constants, initial environmental sensitivity coefficients and initial dynamic rate coefficients related to the current failure mechanism from the material performance database, and use these initial parameters as adjustable parameters in the mathematical relationship; S1a5. For each pressure vessel, select the damage amount of two historical detection data obtained from any two adjacent inspections, take the damage amount at the time of the previous inspection as the initial damage amount, and take the real-time operation monitoring data recorded between the two inspections as the operation parameter input. Substitute it into the mathematical relationship established in S1a3, and calculate the damage evolution model of the current failure mechanism by adjusting the adjustable parameters in the mathematical relationship. S1b1. For any failure mechanism, for the damage evolution model, obtain historical detection data, real-time operation monitoring data and basic equipment information of several pressure vessels that have been put into operation to form a verification dataset. S1b2. Input the historical inspection data, real-time operation monitoring data and basic equipment information of each pressure vessel in the verification dataset into the damage evolution model of the current failure mechanism to generate a predicted damage sequence for each pressure vessel at one or more subsequent inspection times. S1b3. Compare the predicted damage sequence with the historical detection data of the same pressure vessel at the same inspection time in the validation dataset, and calculate the first quantitative index characterizing the prediction accuracy, the second quantitative index characterizing the consistency of the damage evolution trend, and the third quantitative index characterizing the stability of the model under different operating conditions. S1b31. For each pressure vessel in the verification dataset, at each same inspection time, extract the predicted damage amount from the predicted damage amount sequence, calculate the absolute error between the predicted damage amount and the damage amount in the historical inspection data, and perform statistical processing on the absolute error of all pressure vessels at all the same inspection times to generate the first quantitative index characterizing the prediction accuracy. S1b32. Based on the predicted damage sequence and the damage sequence in the historical detection data of each pressure vessel, fit their respective damage evolution trend curves, calculate the similarity between the two trend curves, and then aggregate the similarity of all pressure vessels to generate a second quantitative index characterizing the consistency of the damage evolution trend. S1b33. Based on the operating condition category of each pressure vessel, the validation dataset is divided into several operating condition subsets. S1b31 and S1b32 are repeated in each operating condition subset to obtain the first quantitative index and the second quantitative index of each operating condition subset. Then, the first variance of the first quantitative index and the second variance of the second quantitative index of all operating condition subsets are calculated. After normalizing the first variance and the second variance, the mean is taken to obtain the third quantitative index that characterizes the stability of the model under different operating conditions. S1b4. Based on preset weights, the first quantitative indicator, the second quantitative indicator, and the third quantitative indicator are integrated to generate a comprehensive matching score. S1b41. Based on the physical characteristics and engineering safety sensitivity of the current failure mechanism, the initial weights of the first quantitative index, the second quantitative index, and the third quantitative index are set respectively. S1b42. Obtain the historical verification case library, which contains several verification datasets of pressure vessels. Each verification dataset is explicitly assigned to a known failure mechanism during its construction. S1b43. Input each verification case in the historical verification case library into the damage evolution model of the current failure mechanism, and execute steps S1b2 to S1b3 to obtain the first quantitative index, the second quantitative index and the third quantitative index of the current verification case. S1b44. Based on the consistency between the known failure mechanism to which each verification case belongs and the current failure mechanism, an objective function is constructed. The objective function takes the comprehensive matching score as the output, the first quantitative index, the second quantitative index, and the third quantitative index as the input, and the initial weight as an adjustable parameter. When the known failure mechanism of the verification case is the same as the current failure mechanism, the objective function expects the comprehensive matching score to approach the preset high score threshold; when the two are different, the objective function expects the comprehensive matching score to approach the preset low score threshold. S1b45. The objective function is solved iteratively using an optimization algorithm, and the initial weights are adjusted to make the overall discrimination performance of the comprehensive matching score on the historical verification case library optimal, so as to obtain the optimized preset weights. S1b46. Using preset weights, the first, second, and third quantitative indicators of the damage evolution model of the current failure mechanism are weighted and fused to generate a comprehensive matching score for the damage evolution model of the current failure mechanism. S1b5. Based on the first quantitative index, the second quantitative index, the third quantitative index and the comprehensive matching score, the degree of matching between the damage evolution model of the current failure mechanism and the current failure mechanism is judged, and the damage evolution model of the current failure mechanism is selected to be retained or reconstructed based on the judgment result. S1b51. Set the primary allowable threshold and the strict allowable threshold for the first quantitative indicator, wherein the primary allowable threshold is greater than the strict allowable threshold; set the trend qualified threshold and the trend excellent threshold for the second quantitative indicator, wherein the trend excellent threshold is greater than the trend qualified threshold; set the stability lower limit threshold and the stability upper limit threshold for the third quantitative indicator, wherein the stability upper limit threshold is greater than the stability lower limit threshold; set the entry threshold value and the preferred certification value for the comprehensive matching score, wherein the preferred certification value is greater than the entry threshold value. S1b52. Determine whether the comprehensive matching score of the damage evolution model of the current failure mechanism is less than the entry threshold. If yes, determine that the damage evolution model of the current failure mechanism does not match the current failure mechanism and perform reconstruction. If no, proceed to S1b53. S1b53. Determine whether the first quantitative index of the damage evolution model of the current failure mechanism is less than or equal to the primary tolerance threshold, whether the second quantitative index is greater than or equal to the trend qualification threshold, and whether the third quantitative index is between the lower stability threshold and the upper stability threshold. If all three conditions are met, proceed to S1b54; if any one condition is not met, it is determined that the damage evolution model of the current failure mechanism does not match the current failure mechanism, and reconstruction is performed. S1b54. Determine whether the comprehensive matching score is greater than or equal to the preferred certification value. If yes, directly retain the damage evolution model of the current failure mechanism; otherwise, proceed to S1b55. S1b55. When the overall matching score is greater than or equal to the entry threshold but less than the preferred certification score, perform multi-dimensional indicator collaborative reinforcement verification: determine whether the following two conditions are met simultaneously: First, the first quantitative indicator is less than or equal to the strict allowable threshold; Second, the second quantitative indicator is greater than or equal to the excellent trend threshold; If both conditions are met, the damage evolution model of the current failure mechanism is retained; if either condition is not met, reconstruction is performed until the damage evolution model of the current failure mechanism is retained; and all successfully retained failure mechanisms and damage evolution models are stored in the database to obtain a failure mechanism database containing several failure mechanisms and the damage evolution model of each failure mechanism. S2. Obtain real-time operational monitoring data of the pressure vessel to be inspected from the last inspection to the present, historical test data obtained from the last inspection, and basic equipment information; S3. Based on the real-time operation monitoring data and basic equipment information of the pressure vessel under inspection since the last inspection, the failure mechanism that is compatible with the pressure vessel under inspection under the current operating conditions is matched from the failure mechanism database, and the current damage evolution model is called according to the current failure mechanism. S4. Input the historical test data obtained from the last inspection and the real-time operation monitoring data from the last inspection to the current damage evolution model to calculate the remaining life prediction value and damage status quantitative index of the pressure vessel under inspection at the current moment. S5. Based on the predicted remaining life and quantitative indicators of damage status of the pressure vessel to be inspected at the current moment, generate an inspection plan for the pressure vessel to be inspected.

[0004] This embodiment also discloses an intelligent decision-making system for pressure vessel inspection, including the following modules: Database construction module: used to obtain several failure mechanisms of pressure vessels, build a damage evolution model for each failure mechanism, perform quantization-based logical verification on the damage evolution model of each failure mechanism, and build a failure mechanism database containing several failure mechanisms and damage evolution models of each failure mechanism based on the verification results; Data acquisition module: used to acquire real-time operational monitoring data of the pressure vessel under inspection from the last inspection to the present, historical test data obtained from the last inspection, and basic equipment information; Data matching module: Connected to the database construction module and data acquisition module, it is used to match the failure mechanism that is suitable for the current pressure vessel under inspection under the current operating conditions from the failure mechanism database based on the real-time operation monitoring data and basic equipment information of the pressure vessel under inspection since the last inspection, and call the current damage evolution model according to the current failure mechanism. Data calculation module: connected to the data acquisition module and the data matching module, it is used to input the historical detection data obtained from the last inspection and the real-time operation monitoring data from the last inspection to the current damage evolution model, and calculate the remaining life prediction value and damage status quantitative index of the pressure vessel under inspection at the current moment. Solution generation module: Connected to the data calculation module, it is used to generate an inspection plan for the pressure vessel under inspection based on the predicted remaining life and quantitative indicators of damage status at the current moment.

[0005] The embodiments of the present invention have the following technical effects: This invention effectively overcomes the core shortcomings of existing risk assessment methods, such as static nature, lack of real-time response capability, and insufficient physical interpretability of models, by constructing an intelligent decision-making system that deeply couples failure mechanisms and damage evolution models. Addressing the difficulty of traditional inspection strategies in perceiving the dynamic risk evolution during equipment operation, this invention integrates real-time operational monitoring data, historical inspection data, and basic equipment information to dynamically identify the currently dominant failure mechanism and invoke a highly matched damage evolution model, enabling the inspection plan to adaptively adjust to changes in operating conditions. Crucially, when constructing the failure mechanism database, this invention establishes a dedicated damage evolution model based on the physical or chemical principles of each failure mechanism and introduces… A rigorous multi-dimensional quantitative verification mechanism is implemented, comprehensively evaluating the degree of fit between the model and actual failure behavior, from prediction accuracy and consistency of damage evolution trends to stability under different operating conditions. Based on this, a comprehensive matching score is generated through weighted fusion, and graded criteria are set to achieve model retention or reconstruction, thereby ensuring that every damage evolution model stored in the failure mechanism database has high fidelity and strong generalization ability. Therefore, the models called from it not only have clear physical mechanisms embedded in their structure, but also undergo rigorous screening of multiple scenarios and indicators in terms of performance, which can truly reflect the damage evolution law of specific equipment under specific operating conditions, fundamentally ensuring the reliability and engineering applicability of remaining life prediction and damage status assessment. In summary, this invention achieves a fundamental shift from static risk assessment to dynamic mechanism-driven intelligent decision-making. By constructing a high-precision, high-reliability failure mechanism database, it ensures that the damage evolution model invoked has both a solid physical foundation and has undergone multi-dimensional quantitative verification, accurately depicting the degradation behavior of equipment under complex operating conditions. This method effectively solves the bottleneck problems of traditional technologies, such as the inability to dynamically respond to equipment status, the lack of interpretability of models, and rigid inspection strategies. It significantly improves the scientificity and safety of inspection decisions, and provides reliable support for intelligent and precise safety management of pressure vessels throughout their entire life cycle while avoiding missed or over-inspections. Attached Figure Description

[0006] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0007] Figure 1 This is a flowchart of the intelligent decision-making method for pressure vessel inspection provided in an embodiment of the present invention; Figure 2This is a schematic diagram of the statistical results of the quantitative indicators for the verification of the damage evolution model under different subsets of working conditions provided in the embodiments of the present invention; Figure 3 This is a schematic diagram of the comparison curve between predicted damage and measured damage during the training and verification process of the damage evolution model provided in this embodiment of the invention. Figure 4 This is a framework diagram of the intelligent decision-making system for pressure vessel inspection provided in an embodiment of the present invention. Detailed Implementation

[0008] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0009] Example 1: As Figure 1 As shown, the intelligent decision-making method for pressure vessel inspection provided by this invention includes the following steps: S1. Obtain several failure mechanisms of the pressure vessel, and construct a damage evolution model for each failure mechanism. Perform quantization-based logical verification on the damage evolution model of each failure mechanism, and construct a failure mechanism database containing several failure mechanisms and damage evolution models of each failure mechanism based on the verification results. It is worth noting that in this embodiment, several failure mechanisms of pressure vessels are obtained primarily through the systematic integration of industry standards and specifications, such as TSG 21 and ASME BPVC, historical inspection and testing reports, accident case databases, and research results on material service behavior. These are then summarized and refined in conjunction with common operating conditions and media environments of the equipment. Specifically, these include, but are not limited to, the following typical failure mechanisms: uniform corrosion, localized corrosion such as pitting, high-temperature hydrogen corrosion, crevice corrosion, stress corrosion cracking, hydrogen-induced cracking, high-temperature creep, low-cycle / high-cycle fatigue, brittle fracture, and material aging and deterioration. For different material types such as carbon steel, stainless steel, and low-alloy steel, and for process conditions such as high temperature and high pressure, and sulfur / chlorine-containing media, the failure modes are further refined to ensure that the obtained failure mechanisms cover the main damage forms that pressure vessels may encounter in actual service, laying the foundation for the subsequent construction of targeted damage evolution models.

[0010] S1a1. Determine a failure mechanism as the modeling object; S1a2. Based on the current failure mechanism, collect historical inspection data, real-time operation monitoring data, and basic equipment information generated during the actual use of several pressure vessels already in operation. The historical inspection data refers to the damage amount related to the current failure mechanism measured in each pressure vessel during each inspection. The real-time operation monitoring data refers to the operating parameters continuously recorded between two inspections for each pressure vessel. The basic equipment information includes material type, design temperature, design pressure, medium composition, and structural form. S1a3. Based on the physical or chemical action principle of the current failure mechanism, establish a mathematical relationship between the damage amount and time. The mathematical relationship takes time as the independent variable and the damage amount as the dependent variable, and takes the operating parameters in the real-time operation monitoring data as the input variables affecting the damage evolution rate. S1a4. Based on the material type, medium composition, design temperature, design pressure and structural form in the basic information of the equipment, obtain the initial material constants, initial environmental sensitivity coefficients and initial dynamic rate coefficients related to the current failure mechanism from the material performance database, and use these initial parameters as adjustable parameters in the mathematical relationship; S1a5. For each pressure vessel, select the damage amount of two historical detection data obtained from any two adjacent inspections, take the damage amount at the time of the previous inspection as the initial damage amount, and take the real-time operation monitoring data recorded between the two inspections as the operation parameter input. Substitute it into the mathematical relationship established in S1a3, and calculate the damage evolution model of the current failure mechanism by adjusting the adjustable parameters in the mathematical relationship. It is worth noting that in this embodiment, firstly, one of the several failure mechanisms already acquired is selected as the modeling object, for example, high-temperature hydrogen corrosion is selected as the current modeling target. Subsequently, for this failure mechanism, relevant data accumulated during the actual service life of multiple pressure vessels already in operation are collected: including damage quantities directly corresponding to high-temperature hydrogen corrosion measured during each periodic inspection, such as methane bubble number density, microcrack length, or effective wall thickness reduction; real-time operational monitoring data continuously collected between two inspections, such as equipment wall temperature, system total pressure, and hydrogen partial pressure; and basic equipment information, covering material types such as chromium-molybdenum steel, design temperature, design pressure, medium composition such as high-temperature and high-pressure hydrogen-containing gas flow, and structural forms such as reactors and other key components. Next, based on the chemical-metallurgical coupling mechanism of hydrogen diffusion into the metal matrix and reaction with carbides to generate methane, leading to the formation of grain boundary microcracks in high-temperature hydrogen corrosion, a mathematical relationship of damage quantity evolution over time is established. This relationship is expressed in definite integral form, with time as the independent variable and damage quantity as the dependent variable. Its core formula is: ; In this model structure, the aforementioned operating parameters are used as key input variables for regulating the rate of damage evolution, and are embedded into the corresponding function terms, whereby... Characterizing the exponential effect of temperature on hydrogen diffusion rate and chemical reaction kinetics, The effect of hydrogen partial pressure on surface adsorption equilibrium was characterized, and This characterizes the material's microstructure in terms of its modification of damage resistance. This indicates the initial damage amount; subsequently, based on the specific material type, medium composition, design temperature, and other conditions in the equipment's basic information, initial parameters related to high-temperature hydrogen corrosion are extracted from authoritative material performance databases such as NACEMR0175 and APIRP941, including... The baseline kinetic rate coefficient reflects the basic damage sensitivity of a specific steel grade in a high-temperature hydrogen environment; the temperature sensitivity function coefficient is used to construct... The accelerating effect of temperature changes on creep-oxidation coupled damage was quantified; hydrogen pressure sensitivity function coefficients were used to construct... The description of the driving effect of increased hydrogen partial pressure on crack nucleation rate; material and environmental correction factors, used to construct... The differences caused by alloy element content and stress state are corrected, and these parameters are set as adjustable parameters in the integral model. Finally, for each pressure vessel, damage data obtained from any two adjacent inspections are selected, with the damage value at the time of the previous inspection used as the initial damage value. The real-time operational monitoring data recorded between the two inspection intervals are discretized into time series segments and substituted into the aforementioned integral mathematical relationship. The cumulative damage increment is calculated using numerical integration methods such as Simpson's rule. Simultaneously, the adjustable parameters are adjusted using nonlinear least squares methods. The coefficients of each function term are used to determine the damage state quantity at the current time t, which is the final damage quantity, as output by the model. By minimizing the error between the measured value and the subsequent test, a high-precision damage evolution model applicable to this specific high-temperature hydrogen corrosion failure mechanism can be obtained.

[0011] S1b1. For any failure mechanism, for the damage evolution model, obtain historical detection data, real-time operation monitoring data and basic equipment information of several pressure vessels that have been put into operation to form a verification dataset. S1b2. Input the historical inspection data, real-time operation monitoring data and basic equipment information of each pressure vessel in the verification dataset into the damage evolution model of the current failure mechanism to generate a predicted damage sequence for each pressure vessel at one or more subsequent inspection times. It is worth noting that in this embodiment, when constructing the input data for verification, it is essential to ensure that the selected pressure vessel, during actual service, truly exhibits the damage mode dominated by the failure mechanism corresponding to the current model to be verified. This is because the damage evolution model is a mathematical expression based on specific physical or chemical mechanisms such as stress corrosion cracking, high-temperature creep, and hydrogen-induced cracking. Its structural form, variable coupling method, and the physical meaning of adjustable parameters are all closely dependent on the intrinsic evolution law of this mechanism. Therefore, the model can only accurately describe and predict damage behavior dominated by this mechanism. If equipment data dominated by other failure mechanisms is input into the model, even if its material type, design parameters, and operating conditions appear to match on the surface, the damage essence reflected by its historical detection data will still be inconsistent with the model's assumptions. For example, if a vessel mainly develops cracks due to fatigue load, and its wall thickness reduction or crack length data is input into the high-temperature creep model for prediction, the results will not reflect the true creep evolution process. Instead, the mismatch of mechanisms will introduce systematic biases, leading to distorted verification conclusions or even misleading subsequent engineering decisions. Therefore, before formally collecting historical testing data, real-time operational monitoring data, and basic equipment information for verification, a rigorous technical screening of the failure modes of candidate pressure vessels is necessary. Specifically, this involves a comprehensive analysis of damage morphology characteristics recorded in previous periodic inspection reports, such as whether cracks propagate intergranularly, the presence of oxide scale accumulation, and whether the fracture surface exhibits dimples or cleavage features; environmental conditions such as chloride ion concentration, pH value, hydrogen sulfide partial pressure, and operating temperature range; actual load spectrum characteristics such as the presence of frequent start-ups and shutdowns, pressure fluctuation amplitude, and duration of constant load; and the material's sensitivity to such environments, such as the stress corrosion sensitivity of austenitic stainless steel in chloride-containing media and the hydrogen embrittlement tendency of low-alloy steel in wet hydrogen sulfide environments. This multi-dimensional evidence is crucial for determining the failure mode of the pressure vessel. The dominant damage mechanism must be consistent with the failure mechanism targeted by the model to be validated. Based on this, historical monitoring data directly related to the mechanism can be extracted, including quantitative damage indicators clearly attributable to the target failure mechanism, such as crack depth corresponding to stress corrosion cracking, cylinder outer diameter expansion corresponding to high-temperature creep, and local bulging height corresponding to hydrogen bubbling. Simultaneously, it must be ensured that the collected real-time operational monitoring data covers key operational parameters that drive the evolution of the mechanism, such as temperature-time integral, corrosive component concentration, and static stress level. Complete basic equipment information must also be retained, including material grade, heat treatment status, welding process, design temperature and pressure, medium composition, and structural geometry, to ensure the integrity and consistency of the model input conditions.

[0012] Only when all the above prerequisites are met can the constructed verification dataset truly and objectively reflect the actual law of damage evolution over time under the current failure mechanism, thus making the predicted damage sequence generated after inputting the dataset into the corresponding damage evolution model have engineering credibility.

[0013] S1b3. Compare the predicted damage sequence with the historical detection data of the same pressure vessel at the same inspection time in the validation dataset, and calculate the first quantitative index characterizing the prediction accuracy, the second quantitative index characterizing the consistency of the damage evolution trend, and the third quantitative index characterizing the stability of the model under different operating conditions. S1b31. For each pressure vessel in the verification dataset, at each same inspection time, extract the predicted damage amount from the predicted damage amount sequence, calculate the absolute error between the predicted damage amount and the damage amount in the historical inspection data, and perform statistical processing on the absolute error of all pressure vessels at all the same inspection times to generate the first quantitative index characterizing the prediction accuracy. S1b32. Based on the predicted damage sequence and the damage sequence in the historical detection data of each pressure vessel, fit their respective damage evolution trend curves, calculate the similarity between the two trend curves, and then aggregate the similarity of all pressure vessels to generate a second quantitative index characterizing the consistency of the damage evolution trend. S1b33. Based on the operating condition category of each pressure vessel, the validation dataset is divided into several operating condition subsets. S1b31 and S1b32 are repeated in each operating condition subset to obtain the first quantitative index and the second quantitative index of each operating condition subset. Then, the first variance of the first quantitative index and the second variance of the second quantitative index of all operating condition subsets are calculated. After normalizing the first variance and the second variance, the mean is taken to obtain the third quantitative index that characterizes the stability of the model under different operating conditions. It is worth noting that, in this embodiment, as Figure 2 As shown, firstly, for each pressure vessel in the validation dataset, at each actual inspection time (i.e., the time corresponding to the historical inspection data), the predicted value at the corresponding time point is extracted from the predicted damage sequence output by the model, and compared point by point with the historical damage measured at that time to calculate the absolute error between the two; then, the absolute errors of all equipment at all common inspection times are collected and summarized using statistical methods such as mean absolute error to form a first quantitative indicator that can comprehensively reflect the overall prediction deviation level of the model. Secondly, such as Figure 3As shown, for each pressure vessel, based on the damage quantity sequence in its historical inspection data and the predicted damage quantity sequence generated by the model, the damage evolution trend curves are fitted using the same functional form, such as polynomial, exponential decay, or power law function. On this basis, the trend consistency score of a single device is obtained by calculating the similarity measure of the two curves in the domain, such as the Pearson correlation coefficient. Then, the scores of all devices are aggregated, such as by taking the average, to obtain a second quantitative index characterizing whether the model can accurately reproduce the damage evolution pattern as a whole. Specifically, the formula for calculating the trend consistency score of a single device through similarity measurement is as follows: First, let the damage quantity sequence in the historical inspection data of a single pressure vessel, i.e., the actual detected damage quantity sequence, be: The corresponding predicted damage sequence is: ; ; In the formula, Indicates the consistency score of the current damage evolution trend of the equipment; Indicates the total number of data samples; This represents the measured damage value at the i-th time point; This represents the predicted damage value at the i-th point; This represents the average value of the measured damage. This represents the average value of the predicted damage.

[0014] Furthermore, based on the actual operating conditions of each pressure vessel during its service life, such as high temperature and high pressure, medium temperature and chlorine content, and low temperature and high humidity, the validation dataset is divided into several engineering-significant subsets of operating conditions. The calculation process of the above two indicators is independently repeated in each subset to obtain the first and second quantitative indicators under that operating condition. Subsequently, the variance between the corresponding indicator values ​​of all operating condition subsets is calculated, including the first variance of the first quantitative indicator and the second variance of the second quantitative indicator. Since the first and second variances have different numerical dimensions, they cannot be directly compared or superimposed. Therefore, these two variances are normalized, for example, mapped to the [0,1] interval, to obtain the normalized first and second variances. Finally, the arithmetic mean of the two normalized fluctuation terms is calculated to measure the degree of performance fluctuation of the model under different operating conditions, and finally a third quantitative indicator is formed to evaluate the operating condition adaptability and robustness of the model. By constructing the three quantitative indicators mentioned above in a coordinated manner, we can not only comprehensively characterize the verification performance of the damage evolution model from three dimensions: point prediction accuracy, trend evolution consistency, and working condition stability, but also provide objective, comparable, and traceable technical basis for subsequent model optimization, scope of application definition, and engineering deployment decisions.

[0015] S1b4. Based on preset weights, the first quantitative indicator, the second quantitative indicator, and the third quantitative indicator are integrated to generate a comprehensive matching score. S1b41. Based on the physical characteristics and engineering safety sensitivity of the current failure mechanism, the initial weights of the first quantitative index, the second quantitative index, and the third quantitative index are set respectively. S1b42. Obtain the historical verification case library, which contains several verification datasets of pressure vessels. Each verification dataset is explicitly assigned to a known failure mechanism during its construction. S1b43. Input each verification case in the historical verification case library into the damage evolution model of the current failure mechanism, and execute steps S1b2 to S1b3 to obtain the first quantitative index, the second quantitative index and the third quantitative index of the current verification case. S1b44. Based on the consistency between the known failure mechanism to which each verification case belongs and the current failure mechanism, an objective function is constructed. The objective function takes the comprehensive matching score as the output, the first quantitative index, the second quantitative index, and the third quantitative index as the input, and the initial weight as an adjustable parameter. When the known failure mechanism of the verification case is the same as the current failure mechanism, the objective function expects the comprehensive matching score to approach the preset high score threshold; when the two are different, the objective function expects the comprehensive matching score to approach the preset low score threshold. S1b45. The objective function is solved iteratively using an optimization algorithm, and the initial weights are adjusted to make the overall discrimination performance of the comprehensive matching score on the historical verification case library optimal, so as to obtain the optimized preset weights. S1b46. Using preset weights, the first, second, and third quantitative indicators of the damage evolution model of the current failure mechanism are weighted and fused to generate a comprehensive matching score for the damage evolution model of the current failure mechanism. It is worth noting that, in this embodiment, when comprehensively verifying and evaluating the damage evolution model of the current failure mechanism, the three quantitative indicators mentioned above should first be assigned initial weights based on the physical characteristics of the failure mechanism itself and its sensitivity to engineering safety. For example, if the current failure mechanism is stress corrosion cracking, its damage development is sudden and insidious, and it is extremely sensitive to prediction accuracy, then the first quantitative indicator can be assigned a higher initial weight. If it is high-temperature creep, its evolution process is slow, but trend stability is crucial, then the initial weight of the second quantitative indicator can be appropriately increased. If the mechanism is highly sensitive to fluctuations in the operating environment, such as changes in medium composition or temperature significantly affecting the damage rate, then the initial weight of the third quantitative indicator should also be increased accordingly. Subsequently, a historical verification case library was constructed, which contains multiple pressure vessel instances that have been put into operation. Each instance is equipped with a complete verification dataset, including historical test data, real-time monitoring records, and basic equipment information. During the construction process, the known failure mechanisms to which each instance belongs, such as fatigue, hydrogen-induced cracking, and uniform corrosion, were clearly identified through technical identification. These cases cover a variety of materials, structural forms, and service environments, and are representative and diverse, forming the basic data support for model verification and weight optimization. Next, each validation case in the historical validation case library is sequentially input into the current damage evolution model to be validated, and the aforementioned complete process from model prediction to the generation of three quantitative indicators is executed. That is, for each case, the model is run to generate a predicted damage sequence, and its first, second and third quantitative indicators are calculated to obtain the three-dimensional performance of the case under the current model. Based on this, an objective function is constructed to measure the current model's ability to distinguish cases with different failure mechanisms. The objective function outputs a comprehensive matching score, and its input consists of the specific values ​​of the three quantitative indicators mentioned above. The initial weights are used as adjustable parameters in the score calculation. The design logic of the objective function is as follows: when the actual failure mechanism of a validation case is consistent with the failure mechanism targeted by the current model, its comprehensive matching score is expected to be as close as possible to a preset high score threshold, such as above 90 points, indicating that the model has a high degree of adaptability to this type of case. Conversely, when the actual failure mechanism of the case is inconsistent with the current model, its comprehensive matching score is expected to be as close as possible to a preset low score threshold, such as below 30 points, indicating that the model can effectively identify and eliminate interference from non-target mechanisms. Through this guidance of high scores for similar cases and low scores for dissimilar cases, the objective function essentially constructs a discrimination criterion based on mechanism consistency. To optimize this discrimination performance, mature optimization algorithms, such as particle swarm optimization, genetic algorithms, or gradient descent methods, are used to iteratively solve the objective function. In each iteration, the algorithm automatically adjusts the weight combination of the three indicators, recalculates the comprehensive matching score of all historical cases, and evaluates the overall discrimination effect, for example, through evaluation criteria such as classification accuracy, area under the ROC curve, or inter-class separation. After multiple iterations, the algorithm converges to a set of weight values ​​that optimize the overall discrimination performance, which is the optimized preset weight. Finally, the optimized preset weights are applied to the damage evolution model verification results of the current failure mechanism, and the first, second and third quantitative indicators are weighted and fused to generate a unified comprehensive matching score. This score not only integrates information from three dimensions: prediction accuracy, trend consistency and working condition stability, but its weight configuration has also been dynamically calibrated through feedback learning from the historical case library, which can more realistically and sensitively reflect the degree of matching between the model and the target failure mechanism. The weight optimization driven by historical cases ensures that the contribution of each indicator is determined by the data itself and the consistency of the mechanism, avoiding the subjectivity of human setting and significantly improving the objectivity, adaptability and generalization ability of the verification system. Furthermore, the preset high score threshold is set as follows: This threshold represents the ideal level that the model should achieve in terms of overall matching score when facing verification cases consistent with the current failure mechanism. Its setting should be based on empirical analysis of high-quality verification cases. Specifically, several benchmark cases that have been fully verified in engineering practice and highly match the current failure mechanism can be selected from the historical verification case library. These cases typically have a clear chain of failure evidence, complete monitoring data, a stable service environment, and have been confirmed through other independent means, such as destructive testing, long-term tracking, or expert review, that their damage evolution pattern is consistent with the model assumptions. These benchmark cases are then input into the current model, their three quantitative indicators are calculated, and an initial comprehensive score is generated using initial weights. Their natural distribution range is observed. Based on this, combined with engineering safety margin requirements, such as requiring model prediction errors to be controlled within an acceptable range and trend deviations not exceeding a certain proportion, the high-score threshold is set to the score value corresponding to the upper quartile or mean plus one standard deviation in this distribution, typically in the range of 85 to 95 points. This threshold reflects the lower limit of performance that the model should stably achieve under ideal matching conditions, rather than the theoretical maximum value, to retain a reasonable tolerance margin. The preset low score threshold is set as follows: This threshold is used to define the low confidence region where the overall matching score of the model should fall when dealing with cases that are inconsistent with the current failure mechanism. Its setting also depends on empirical data. Several interference cases that clearly belong to other failure mechanisms can be selected from the historical verification case library. For example, when using a creep model to verify fatigue-dominated containers, these cases can be input into the current model, the overall score can be calculated, and their typical distribution can be observed. Since mechanism mismatch usually leads to a serious deviation of the prediction from the actual trend, such scores are often significantly lower. Based on this, combined with the need for misjudgment risk control, such as avoiding misjudging non-target mechanism cases as fit, the low score threshold is set to the lower quartile of the interference case score distribution or the value corresponding to the mean minus one standard deviation, which is usually in the range of 20 to 40 points. This threshold should ensure that the scores of the vast majority of outlier cases are significantly lower than the high score threshold, forming a clear discrimination boundary. In addition, to enhance the robustness of the threshold setting, a cross-validation strategy can be introduced: repeat the above analysis on historical cases in different subsets to test the stability of the threshold; or refer to the requirements for model validation accuracy in industry standards and specifications, such as the allowable range of remaining life prediction error in API and ASME guidelines, to transform engineering compliance requirements into constraints for the scoring threshold; the final high and low score thresholds satisfy two core principles: first, distinguishability, that is, the score distribution of similar and dissimilar cases is significantly separated; second, feasibility, that is, the high score threshold can be stably reached in real matching cases, and the low score threshold is generally lower than this value in mismatch cases.

[0016] S1b5. Based on the first quantitative index, the second quantitative index, the third quantitative index and the comprehensive matching score, the degree of matching between the damage evolution model of the current failure mechanism and the current failure mechanism is judged, and the damage evolution model of the current failure mechanism is selected to be retained or reconstructed based on the judgment result. S1b51. Set the primary allowable threshold and the strict allowable threshold for the first quantitative indicator, wherein the primary allowable threshold is greater than the strict allowable threshold; set the trend qualified threshold and the trend excellent threshold for the second quantitative indicator, wherein the trend excellent threshold is greater than the trend qualified threshold; set the stability lower limit threshold and the stability upper limit threshold for the third quantitative indicator, wherein the stability upper limit threshold is greater than the stability lower limit threshold; set the entry threshold value and the preferred certification value for the comprehensive matching score, wherein the preferred certification value is greater than the entry threshold value. It is worth noting that, in this embodiment, the primary and strict allowable thresholds for the first quantitative indicator are set as follows: The first quantitative indicator reflects the point prediction accuracy of the model at a specific test moment, such as the mean absolute error. Its primary allowable threshold represents the lowest level of accuracy that the model can accept in engineering. It is usually set according to industry standards such as API 581 for the allowable range of remaining lifetime prediction error or internal safety procedures of the enterprise, and is calibrated in combination with the typical error distribution of similar mechanism equipment in historical verification cases. For example, the 80th percentile of the error distribution of similar cases is taken as the primary allowable threshold to ensure that most qualified models can pass this threshold. The strict allowable threshold is used for high reliability scenarios such as critical equipment or high-risk operating conditions. It is usually taken as the 50th percentile or better of the error distribution of similar cases to reflect the higher requirements for prediction accuracy. The thresholds for acceptable and excellent trends in the second quantitative indicator are set as follows: The second quantitative indicator measures the model's ability to reproduce the damage evolution pattern, such as the similarity of the trend curve; the acceptable trend threshold is set as the minimum similarity level that can ensure that the direction and rate of damage evolution are basically consistent. It can be determined by expert evaluation of the visual matching degree of multiple sets of measured and predicted trends, combined with statistical methods such as floating the lower limit of the similarity distribution of historical matching cases upward by a certain percentage; the excellent trend threshold corresponds to the case of high matching, usually referring to the similarity performance of benchmark cases, such as the average of the top 20% of the best cases, to identify models with high-fidelity evolution capabilities; The stability lower and upper thresholds for the third quantitative indicator are set as follows: The third quantitative indicator characterizes the performance fluctuation of the model under different operating conditions, such as the variance of the indicator values ​​of each subset of operating conditions. Since too low variance may mean insufficient operating condition coverage, such as all cases being concentrated in a single operating condition, while too high variance indicates poor model robustness, a reasonable range needs to be set. The stability lower threshold is used to exclude false stability caused by insufficient operating condition diversity, and can be calculated based on the number of effective operating condition categories and distribution entropy in the historical case library. The stability upper threshold is set according to the acceptable performance fluctuation range in the project, for example, referring to the maximum allowable deviation of similar models in cross-operating condition verification, to ensure that the model will not fail due to changes in operating conditions in actual deployment. The entry threshold and preferred certification value for the comprehensive matching score are set as follows: These two thresholds directly determine whether the model can be used in engineering practice; the entry threshold is the minimum credibility requirement for the model to be put into use, and is usually set as a critical value that allows the vast majority of real matching cases to pass while effectively excluding obvious mismatch cases. The optimal classification threshold can be determined on the historical case database through ROC curve analysis; the preferred certification value is used to identify high-performance models, which usually correspond to the top 30% quantile in the comprehensive score distribution and are suitable for key application scenarios with extremely high requirements for security or prediction accuracy.

[0017] S1b52. Determine whether the comprehensive matching score of the damage evolution model of the current failure mechanism is less than the entry threshold. If yes, determine that the damage evolution model of the current failure mechanism does not match the current failure mechanism and perform reconstruction. If no, proceed to S1b53. It is worth noting that the reason for this judgment in this embodiment is that the comprehensive matching score is an overall credibility measure after weighted fusion of three indicators. If it is lower than the entry threshold, it means that the model cannot effectively match the target failure mechanism as a whole, and it does not have engineering usability regardless of the individual indicators. Moreover, since the comprehensive score has been optimized with weights through historical cases and has discriminative power, if even the minimum credibility cannot be reached, there is no need to continue to examine the detailed indicators. This judgment process can quickly screen out obviously mismatched models in the early stage, avoid subsequent redundant calculations, and improve verification efficiency.

[0018] S1b53. Determine whether the first quantitative index of the damage evolution model of the current failure mechanism is less than or equal to the primary tolerance threshold, whether the second quantitative index is greater than or equal to the trend qualification threshold, and whether the third quantitative index is between the lower stability threshold and the upper stability threshold. If all three conditions are met, proceed to S1b54; if any one condition is not met, it is determined that the damage evolution model of the current failure mechanism does not match the current failure mechanism, and reconstruction is performed. It is worth noting that in this embodiment, even if the overall score meets the standard, it is still necessary to ensure that the model does not have any fatal flaws in the three dimensions of accuracy, trend, and stability. For example, a model may have a high overall score due to an extremely excellent performance in one indicator, but if another indicator is seriously inaccurate, such as a completely wrong trend, such a model is unreliable. At the same time, since the engineering value of the damage evolution model depends on the synergy of multiple capabilities: point accuracy ensures the assessment of the current state, trend consistency supports future prediction, and stability ensures generalization ability, the lack of any dimension will lead to application failure. This approach of preventing overgeneralization can effectively ensure that the model meets the standard in a balanced manner across multiple dimensions, and avoid the risk of failure in a single aspect being masked by weight allocation.

[0019] S1b54. Determine whether the comprehensive matching score is greater than or equal to the preferred certification value. If yes, directly retain the damage evolution model of the current failure mechanism; otherwise, proceed to S1b55. It is worth noting that in this embodiment, for high-risk equipment, simply meeting the basic access requirements is insufficient to support decision-making. Therefore, it is necessary to further identify the preferred model. If the model performs well overall, it indicates that it highly matches the target mechanism in multiple verification dimensions and can be directly adopted without further stringent testing, thus improving the efficiency of the verification process. This is because the preferred certification value already represents the ideal level in engineering practice. Reaching this level means that the model has high reliability and generalization ability, thus eliminating the need for additional constraints. This approach of providing a green channel for high-performance models can significantly reduce redundant verification steps, thereby accelerating the deployment of high-quality models.

[0020] S1b55. When the overall matching score is greater than or equal to the entry threshold but less than the preferred certification score, perform multi-dimensional indicator collaborative reinforcement verification: determine whether the following two conditions are met simultaneously: First, the first quantitative indicator is less than or equal to the strict allowable threshold; Second, the second quantitative indicator is greater than or equal to the excellent trend threshold; If both conditions are met, the damage evolution model of the current failure mechanism is retained; if either condition is not met, reconstruction is performed until the damage evolution model of the current failure mechanism is retained; and all successfully retained failure mechanisms and damage evolution models are stored in the database to obtain a failure mechanism database containing several failure mechanisms and the damage evolution model of each failure mechanism. It is worth noting that in this embodiment, although such models are generally well-matched, they are not optimal and may have slight sensitivity to operating conditions or deviations in individual cases. In this case, if they perform well in the two most core capabilities, namely point prediction accuracy (strict tolerance) and evolution trend restoration capability (excellent trend), they can be considered to have sufficient engineering value and can be retained. This is because the model is in a marginal usable state at this stage and must be constrained by more stringent dual core indicators to ensure that it does not mask potential risks due to the weighted average effect of the comprehensive score. Only when the two indicators that best reflect the essence of the mechanism matching are both excellent can it be considered reliable. This approach, which allows for some tolerance for slight fluctuations in non-critical dimensions such as operating condition stability while ensuring core performance, can avoid rejecting a substantially usable model due to secondary factors, thereby improving the utilization rate of model resources. It is worth further elaborating that once a reconstruction is triggered, for example, if a model mismatch is determined in any of the steps S1b52, S1b53, or S1b55, the process will revert to stage S1b1, where sample vessels matching the current failure mechanism label will be re-selected from the candidate pressure vessel set. Subsequently, in S1b2, the same technical verification process will be performed on this batch of candidate vessels to ensure the accuracy of the failure mechanism labeling. Then, in S1b3, the corresponding historical detection data, real-time operational monitoring data, and basic equipment information will be re-collected. Finally, in S1b4, the damage evolution will be refitted or adjusted based on the updated dataset. The structural parameters of the model are modified; finally, in S1b51, the thresholds are reset or the original settings are retained, and the step-by-step discrimination process from S1b52 to S1b55 is entered again until the generated damage evolution model can successfully pass all the discrimination conditions from S1b52 to S1b55. This design ensures both the rigor of model development and the self-consistency and repeatability of the entire verification system. Finally, all successfully retained failure mechanisms and corresponding damage evolution models are solidified into the database, thus obtaining a failure mechanism database containing several failure mechanisms and damage evolution models for each failure mechanism.

[0021] S2. Obtain real-time operational monitoring data of the pressure vessel to be inspected from the last inspection to the present, historical test data obtained from the last inspection, and basic equipment information; It is worth noting that in this embodiment, it is first necessary to comprehensively collect three types of core data related to the pressure vessel under inspection, including real-time operational monitoring data: through a sensor network deployed on the vessel body or related pipelines, such as online monitoring devices for temperature, pressure, vibration, medium composition, and wall temperature difference, continuous or periodic operational parameter records are obtained from the end of the last statutory inspection to the current moment. This data should include timestamps, cumulative operating time, operating status indicators such as start-up, shutdown, load changes, abnormal fluctuations, and environmental conditions such as ambient temperature and humidity to ensure a complete reconstruction of the equipment's service history; historical inspection data: retrieving data from the last inspection report confirmed by an authoritative institution. Damage quantification results include, but are not limited to, wall thickness measurements, crack length and depth, corrosion pit distribution, hardness changes, and metallographic degradation levels. The detection location, method (e.g., ultrasonic thickness measurement, X-ray imaging, eddy current testing), and detection time must be clearly indicated. Basic equipment information includes: extracting the container's design and manufacturing records, covering static attributes such as material grade, equipment name, pressure chamber, design pressure and temperature, structural form, welding process, heat treatment records, service life, past maintenance history, and the process unit to which it belongs. These three types of data must undergo data cleaning, time alignment, and unit standardization to form a structured input data package, providing complete information support for subsequent analysis.

[0022] S3. Based on the real-time operation monitoring data and basic equipment information of the pressure vessel under inspection since the last inspection, the failure mechanism that is compatible with the pressure vessel under inspection under the current operating conditions is matched from the failure mechanism database, and the current damage evolution model is called according to the current failure mechanism. It is worth noting that in this embodiment, identifying the dominant failure mode of the current vessel specifically involves reusing the failure mode identification criteria established in S1b2. This involves comprehensively analyzing the material properties of the pressure vessel under inspection, such as steel type, heat treatment state, service environment characteristics such as medium composition, pH value, sulfur / chloride ion concentration, presence of wet hydrogen, load history including temperature-pressure time series, start-up and shutdown frequency, fluctuation amplitude, and structurally sensitive parts such as welds, openings, and geometric discontinuities. This is combined with real-time operational monitoring data since the last inspection, such as abnormal wall temperature increases, increased local vibration, and sudden changes in medium corrosivity parameters, to conduct a multi-dimensional coupled assessment. Essentially, this process is an engineering judgment on whether the current vessel is under a specific type of failure physical condition. Its criterion system, classification logic, and evidence weight allocation are all consistent with the failure assessment of candidate pressure vessels in the S1b2 stage. The mechanism identification is completely consistent. Through the above-mentioned technical identification, the most likely failure mechanism to dominate the damage evolution is clearly identified, such as the synergistic effect of high-temperature hydrogen corrosion and creep or crack initiation dominated by low-cycle fatigue. It is worth noting that this failure mechanism is not obtained by database matching, but is the result of independent judgment by engineers or intelligent inference engines based on physical rules and engineering experience. Subsequently, the identified failure mechanism is used as a unique index key for precise querying in a pre-built and maintained failure mechanism database. This database uses standardized failure mechanisms as entries, such as SSCC in a wet H2S environment. Each entry is associated with a damage evolution model, and the current damage evolution model that strictly corresponds to the current failure mechanism is retrieved accordingly. This ensures that the physical assumptions, control equations, and input / output interfaces of the model are highly consistent with the actual damage process, providing a reliable basis for subsequent lifetime calculations.

[0023] S4. Input the historical test data obtained from the last inspection and the real-time operation monitoring data from the last inspection to the current damage evolution model to calculate the remaining life prediction value and damage status quantitative index of the pressure vessel under inspection at the current moment. It is worth noting that in this embodiment, the historical detection data compiled in S2 is used as the initial state input of the model, i.e., the starting point of damage evolution. Real-time monitoring data is then input into the model segment by segment according to the time series, driving it to simulate the accumulation process of damage over time. Internally, the model dynamically updates the damage amount based on the physical laws of the current failure mechanism, such as the Paris equation describing the crack propagation rate or the Arrhenius equation characterizing the creep rate, combined with the influence factors of real-time operating conditions on the damage rate, such as temperature acceleration coefficient and stress intensity factor correction. Finally, two key results are output: Remaining life prediction: refers to the estimated time required from the current moment for the container damage to reach the critical failure threshold, such as minimum allowable wall thickness or critical crack size, usually in years or months; Damage status quantification indicators: include the absolute value of the current damage and the safety margin ratio, i.e. the ratio of the current damage to the critical value, etc., used to characterize the precision of the current health status of the equipment.

[0024] S5. Based on the predicted remaining life and quantitative indicators of damage status of the pressure vessel to be inspected at the current moment, generate an inspection plan for the pressure vessel to be inspected. It is worth noting that in this embodiment, if the predicted remaining lifespan is greater than the next planned inspection cycle, such as 3 years, and the quantitative indicators of damage status are in the low-risk range, such as a safety margin > 70%, a routine inspection plan is generated, which only requires the execution of basic inspection items in the standard procedure, such as fixed-point thickness measurement and visual inspection. If the damage status has entered the high-risk range, such as a safety margin < 30% or a remaining lifespan of less than 6 months, an emergency intervention process is triggered, and it is recommended to immediately stop or reduce the load of operation, and arrange comprehensive non-destructive testing and structural integrity reassessment. The final generated inspection plan includes testing methods, testing locations, testing frequency, acceptance standards, and subsequent monitoring recommendations, which can be directly issued to the inspection execution unit to achieve a closed loop from intelligent assessment to action instructions.

[0025] It is worth noting that, in order to verify the actual effect of the damage evolution model and its verification mechanism constructed in this invention, this embodiment takes high-temperature hydrogen corrosion as an example of failure mechanism for experimental training and effect verification, as follows: Construction of the Experimental Training Dataset: This embodiment collects historical inspection data, real-time operational monitoring data, and basic equipment information of multiple 2.25Cr-1Mo chromium-molybdenum steel pressure vessels already in operation in a chemical plant area during their actual service life, forming the experimental training dataset. The historical inspection data consists of damage directly related to high-temperature hydrogen corrosion measured during each periodic inspection of each pressure vessel; the real-time operational monitoring data consists of operating parameters such as equipment wall temperature, total system pressure, hydrogen partial pressure, load factor, and number of start-stop cycles, continuously recorded hourly between two inspections; the basic equipment information includes material type, design temperature, design pressure, medium composition, and structural form. The validation dataset is divided into three subsets based on actual operating conditions: a high-temperature, high-pressure subset; a medium-temperature, hydrogen-containing subset; and a low-temperature, fluctuating subset.

[0026] Damage evolution model parameter calibration: Based on the damage evolution mathematical relationship established in S1a3, the cumulative damage increment between two inspections is calculated using the Simpson numerical integration method, and the adjustable parameters in the mathematical relationship are adjusted using the nonlinear least squares method to minimize the error between the damage output of the model and the measured damage of the pressure vessel in the subsequent inspection, thereby completing the calibration of the high-temperature hydrogen corrosion damage evolution model parameters.

[0027] Weight Optimization Training: Following the process described in S1b4, a case library containing multiple historical verification cases is constructed. These historical verification cases cover various failure mechanisms such as high-temperature hydrogen corrosion, fatigue damage, and hydrogen stress corrosion. Each case has its assigned failure mechanism clearly labeled. Using the first, second, and third quantitative indicators as inputs, and the initial weights as adjustable parameters, an objective function is constructed. A particle swarm optimization algorithm is used for iterative solution to optimize the overall discrimination performance of the comprehensive matching score on the historical verification case library, thus obtaining the optimized preset weights. The weight configurations before and after optimization are shown in Table 1 below. Table 1: Comparison Table Before and After Weight Optimization Verification Results: The above verification dataset was input into the calibrated high-temperature hydrogen corrosion damage evolution model, and steps S1b2 to S1b3 were executed to calculate the first, second, and third quantitative indices for each working condition subset and the full verification dataset. These indices were then weighted and fused according to optimized preset weights to generate a comprehensive matching score. The verification results are shown in Table 2 below: Table 2: Summary of Validation Results of High-Temperature Hydrogen Corrosion Damage Evolution Model Further, following the judgment process described in S1b5, the initial tolerance threshold for the first quantitative indicator is set to 0.02, and the strict tolerance threshold is set to 0.015. The trend qualification threshold for the second quantitative indicator is set to 90%, and the trend excellence threshold is set to 95%. The stability lower limit threshold for the third quantitative indicator is set to 0, and the stability upper limit threshold is set to 0.05. The entry threshold for the comprehensive matching score is 90, and the preferred certification value is 95. The verification results show that: under the full verification dataset, the comprehensive matching score of the model is 96.10, which is not lower than the entry threshold; the first quantitative indicator is 0.0145, which is not greater than the initial tolerance threshold; the second quantitative indicator is 97.65%, which is not lower than the trend qualification threshold; the third quantitative indicator is 0.0000, which is between the stability lower limit threshold and the stability upper limit threshold; and the comprehensive matching score is not lower than the preferred certification value. Through step-by-step judgment from S1b52 to S1b55, this high-temperature hydrogen corrosion damage evolution model is highly matched with the high-temperature hydrogen corrosion failure mechanism, and is therefore retained and solidified into the database.

[0028] Engineering Application Results: The validated high-temperature hydrogen corrosion damage evolution model was applied to a synthesis tower (R-101). The tower's material is 2.25Cr-1Mo, its design pressure is 32 MPa, its design temperature is 500℃, and the medium is hydrogen-containing synthesis gas. Using historical inspection data from the last inspection on September 15, 2023 (with a damage value of 0.23), as the initial state, real-time operational monitoring data from September 15, 2023 to the present was input into the model segment by segment. The calculated quantitative index of the current damage state of the synthesis tower was 0.400, the predicted remaining life was 10.2 years, the failure mechanism matching accuracy was 94.0%, and the algorithm confidence level was 95.0%. Based on this, a routine inspection plan was generated, which includes performing fixed-point thickness measurement, visual inspection, and key checks on the heat-affected zone of the main weld in the next planned inspection cycle. The synthesis tower was subsequently tested during the periodic inspection in 2026. The measured damage was 0.39 to 0.40, and the relative error between the measured and model predictions was no more than 2.5%, which verified the effectiveness and engineering applicability of the model predictions.

[0029] Example 2: Figure 4 As shown, the present invention also discloses an intelligent decision-making system for pressure vessel inspection, comprising the following modules: Database construction module: used to obtain several failure mechanisms of pressure vessels, build a damage evolution model for each failure mechanism, perform quantization-based logical verification on the damage evolution model of each failure mechanism, and build a failure mechanism database containing several failure mechanisms and damage evolution models of each failure mechanism based on the verification results; Data acquisition module: used to acquire real-time operational monitoring data of the pressure vessel under inspection from the last inspection to the present, historical test data obtained from the last inspection, and basic equipment information; Data matching module: Connected to the database construction module and data acquisition module, it is used to match the failure mechanism that is suitable for the current pressure vessel under inspection under the current operating conditions from the failure mechanism database based on the real-time operation monitoring data and basic equipment information of the pressure vessel under inspection since the last inspection, and call the current damage evolution model according to the current failure mechanism. Data calculation module: connected to the data acquisition module and the data matching module, it is used to input the historical detection data obtained from the last inspection and the real-time operation monitoring data from the last inspection to the current damage evolution model, and calculate the remaining life prediction value and damage status quantitative index of the pressure vessel under inspection at the current moment. Solution generation module: Connected to the data calculation module, it is used to generate an inspection plan for the pressure vessel under inspection based on the predicted remaining life and quantitative indicators of damage status at the current moment.

[0030] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent decision-making method for pressure vessel inspection, characterized in that, Includes the following steps: S1. Obtain several failure mechanisms of the pressure vessel, and construct a damage evolution model for each failure mechanism. Perform quantization-based logical verification on the damage evolution model of each failure mechanism, and construct a failure mechanism database containing several failure mechanisms and damage evolution models of each failure mechanism based on the verification results. S2. Obtain real-time operational monitoring data of the pressure vessel to be inspected from the last inspection to the present, historical test data obtained from the last inspection, and basic equipment information; S3. Based on the real-time operation monitoring data and basic equipment information of the pressure vessel under inspection since the last inspection, the failure mechanism that is compatible with the pressure vessel under inspection under the current operating conditions is matched from the failure mechanism database, and the current damage evolution model is called according to the current failure mechanism. S4. Input the historical test data obtained from the last inspection and the real-time operation monitoring data from the last inspection to the current damage evolution model to calculate the remaining life prediction value and damage status quantitative index of the pressure vessel under inspection at the current moment. S5. Based on the predicted remaining life and quantitative indicators of damage status of the pressure vessel to be inspected at the current moment, generate an inspection plan for the pressure vessel to be inspected.

2. The intelligent decision-making method for pressure vessel inspection according to claim 1, characterized in that, The construction of the damage evolution model for each failure mechanism includes: S1a1. Determine a failure mechanism as the modeling object; S1a2. Based on the current failure mechanism, collect historical inspection data, real-time operation monitoring data, and basic equipment information generated during the actual use of several pressure vessels already in operation. The historical inspection data refers to the damage amount related to the current failure mechanism measured in each pressure vessel during each inspection. The real-time operation monitoring data refers to the operating parameters continuously recorded between two inspections for each pressure vessel. The basic equipment information includes material type, design temperature, design pressure, medium composition, and structural form. S1a3. Based on the physical or chemical action principle of the current failure mechanism, establish a mathematical relationship between the damage amount and time. The mathematical relationship takes time as the independent variable and the damage amount as the dependent variable, and takes the operating parameters in the real-time operation monitoring data as the input variables affecting the damage evolution rate. S1a4. Based on the material type, medium composition, design temperature, design pressure and structural form in the basic information of the equipment, obtain the initial material constants, initial environmental sensitivity coefficients and initial dynamic rate coefficients related to the current failure mechanism from the material performance database, and use these initial parameters as adjustable parameters in the mathematical relationship; S1a5. For each pressure vessel, select the damage amount of two historical detection data obtained from any two adjacent inspections, take the damage amount at the time of the previous inspection as the initial damage amount, and take the real-time operation monitoring data recorded between the two inspections as the operation parameter input. Substitute it into the mathematical relationship established in S1a3, and calculate the damage evolution model of the current failure mechanism by adjusting the adjustable parameters in the mathematical relationship.

3. The intelligent decision-making method for pressure vessel inspection according to claim 2, characterized in that, The damage evolution model for each failure mechanism undergoes quantization-based logical verification, including: S1b1. For any failure mechanism, for the damage evolution model, obtain historical detection data, real-time operation monitoring data and basic equipment information of several pressure vessels that have been put into operation to form a verification dataset. S1b2. Input the historical inspection data, real-time operation monitoring data and basic equipment information of each pressure vessel in the verification dataset into the damage evolution model of the current failure mechanism to generate a predicted damage sequence for each pressure vessel at one or more subsequent inspection times. S1b3. Compare the predicted damage sequence with the historical detection data of the same pressure vessel at the same inspection time in the validation dataset, and calculate the first quantitative index characterizing the prediction accuracy, the second quantitative index characterizing the consistency of the damage evolution trend, and the third quantitative index characterizing the stability of the model under different operating conditions. S1b4. Based on preset weights, the first quantitative indicator, the second quantitative indicator, and the third quantitative indicator are integrated to generate a comprehensive matching score. S1b5. Based on the first quantitative index, the second quantitative index, the third quantitative index and the comprehensive matching score, the degree of matching between the damage evolution model of the current failure mechanism and the current failure mechanism is judged, and the damage evolution model of the current failure mechanism is selected to be retained or reconstructed based on the judgment result.

4. The intelligent decision-making method for pressure vessel inspection according to claim 3, characterized in that, The process involves comparing the predicted damage sequence with historical detection data of the same pressure vessel at the same inspection time in the validation dataset, and calculating a first quantitative index characterizing the prediction accuracy, a second quantitative index characterizing the consistency of the damage evolution trend, and a third quantitative index characterizing the stability of the model under different operating conditions. These include: S1b31. For each pressure vessel in the verification dataset, at each same inspection time, extract the predicted damage amount from the predicted damage amount sequence, calculate the absolute error between the predicted damage amount and the damage amount in the historical inspection data, and perform statistical processing on the absolute error of all pressure vessels at all the same inspection times to generate the first quantitative index characterizing the prediction accuracy. S1b32. Based on the predicted damage sequence and the damage sequence in the historical detection data of each pressure vessel, fit their respective damage evolution trend curves, calculate the similarity between the two trend curves, and then aggregate the similarity of all pressure vessels to generate a second quantitative index characterizing the consistency of the damage evolution trend. S1b33. Based on the operating condition category of each pressure vessel, the validation dataset is divided into several operating condition subsets. S1b31 and S1b32 are repeated in each operating condition subset to obtain the first and second quantitative indices for each operating condition subset. Then, the first variance of the first quantitative indices and the second variance of the second quantitative indices for all operating condition subsets are calculated. After normalizing the first and second variances, the average value is taken to obtain the third quantitative index characterizing the stability of the model under different operating conditions.

5. The intelligent decision-making method for pressure vessel inspection according to claim 4, characterized in that, The process of fusing the first, second, and third quantitative indicators based on preset weights to generate a comprehensive matching score includes: S1b41. Based on the physical characteristics and engineering safety sensitivity of the current failure mechanism, the initial weights of the first quantitative index, the second quantitative index, and the third quantitative index are set respectively. S1b42. Obtain the historical verification case library, which contains several verification datasets of pressure vessels. Each verification dataset is explicitly assigned to a known failure mechanism during its construction. S1b43. Input each verification case in the historical verification case library into the damage evolution model of the current failure mechanism, and execute steps S1b2 to S1b3 to obtain the first quantitative index, the second quantitative index and the third quantitative index of the current verification case. S1b44. Based on the consistency between the known failure mechanism to which each verification case belongs and the current failure mechanism, an objective function is constructed. The objective function takes the comprehensive matching score as the output, the first quantitative index, the second quantitative index, and the third quantitative index as the input, and the initial weight as an adjustable parameter. When the known failure mechanism of the verification case is the same as the current failure mechanism, the objective function expects the comprehensive matching score to approach the preset high score threshold; when the two are different, the objective function expects the comprehensive matching score to approach the preset low score threshold. S1b45. The objective function is solved iteratively using an optimization algorithm, and the initial weights are adjusted to make the overall discrimination performance of the comprehensive matching score on the historical verification case library optimal, so as to obtain the optimized preset weights. S1b46. Using preset weights, the first, second, and third quantitative indicators of the damage evolution model of the current failure mechanism are weighted and fused to generate a comprehensive matching score for the damage evolution model of the current failure mechanism.

6. The intelligent decision-making method for pressure vessel inspection according to claim 5, characterized in that, The process of judging the degree of matching between the damage evolution model of the current failure mechanism and the current failure mechanism based on the first quantitative index, the second quantitative index, the third quantitative index, and the comprehensive matching score, and selecting whether to retain or reconstruct the damage evolution model of the current failure mechanism based on the judgment result, includes: S1b51. Set the primary allowable threshold and the strict allowable threshold for the first quantitative indicator, wherein the primary allowable threshold is greater than the strict allowable threshold; set the trend qualified threshold and the trend excellent threshold for the second quantitative indicator, wherein the trend excellent threshold is greater than the trend qualified threshold; set the stability lower limit threshold and the stability upper limit threshold for the third quantitative indicator, wherein the stability upper limit threshold is greater than the stability lower limit threshold; set the entry threshold value and the preferred certification value for the comprehensive matching score, wherein the preferred certification value is greater than the entry threshold value. S1b52. Determine whether the comprehensive matching score of the damage evolution model of the current failure mechanism is less than the entry threshold. If yes, determine that the damage evolution model of the current failure mechanism does not match the current failure mechanism and perform reconstruction. If no, proceed to S1b53. S1b53. Determine whether the first quantitative index of the damage evolution model of the current failure mechanism is less than or equal to the primary tolerance threshold, whether the second quantitative index is greater than or equal to the trend qualified threshold, and whether the third quantitative index is between the lower stability threshold and the upper stability threshold. If all three conditions are met, proceed to S1b54; if any one condition is not met, it is determined that the damage evolution model of the current failure mechanism does not match the current failure mechanism, and reconstruction is performed. S1b54. Determine whether the comprehensive matching score is greater than or equal to the preferred certification value. If yes, directly retain the damage evolution model of the current failure mechanism; otherwise, proceed to S1b55. S1b55. When the overall matching score is greater than or equal to the entry threshold but less than the preferred certification score, perform multi-dimensional indicator collaborative reinforcement verification: determine whether the following two conditions are met simultaneously: First, the first quantitative indicator is less than or equal to the strict allowable threshold; Second, the second quantitative indicator is greater than or equal to the excellent trend threshold; If both conditions are met, the damage evolution model of the current failure mechanism is retained; if either condition is not met, reconstruction is performed until the damage evolution model of the current failure mechanism is retained; and all successfully retained failure mechanisms and damage evolution models are stored in the database to obtain a failure mechanism database containing several failure mechanisms and the damage evolution model of each failure mechanism.

7. A pressure vessel inspection intelligent decision-making system, used to execute the pressure vessel inspection intelligent decision-making method according to any one of claims 1-6, characterized in that, Includes the following modules: Database construction module: used to obtain several failure mechanisms of pressure vessels, build a damage evolution model for each failure mechanism, perform quantization-based logical verification on the damage evolution model of each failure mechanism, and build a failure mechanism database containing several failure mechanisms and damage evolution models of each failure mechanism based on the verification results; Data acquisition module: used to acquire real-time operational monitoring data of the pressure vessel under inspection from the last inspection to the present, historical test data obtained from the last inspection, and basic equipment information; Data matching module: Connected to the database construction module and data acquisition module, it is used to match the failure mechanism that is suitable for the current pressure vessel under inspection under the current operating conditions from the failure mechanism database based on the real-time operation monitoring data and basic equipment information of the pressure vessel under inspection since the last inspection, and call the current damage evolution model according to the current failure mechanism. Data calculation module: connected to the data acquisition module and the data matching module, it is used to input the historical detection data obtained from the last inspection and the real-time operation monitoring data from the last inspection to the current damage evolution model, and calculate the remaining life prediction value and damage status quantitative index of the pressure vessel under inspection at the current moment. Solution generation module: Connected to the data calculation module, it is used to generate an inspection plan for the pressure vessel under inspection based on the predicted remaining life and quantitative indicators of damage status at the current moment.

Images