A method for predicting deformation of a pressure vessel structure based on wall thickness time-varying analysis

Abstract

The application provides a pressure vessel structure deformation prediction method based on wall thickness time-varying analysis, and belongs to the technical field of prediction analysis. The method comprises the following steps: searching a global wall thickness monitoring sample set; performing wall thickness partitioning according to pressure vessel design parameters; searching a local inner wall pressure monitoring sample set, and counting the inner wall pressure in the partitioning; combining the partitioning container structure, size distribution and material constraints, searching a local wall thickness monitoring sample set, and counting wall thickness change time sequence information; performing confidence analysis on the wall thickness change time sequence information based on the global wall thickness monitoring sample set; when the confidence is greater than or equal to a threshold value, the wall thickness change time sequence information is added to the prediction result. The application solves the technical problem that the prediction accuracy is insufficient due to the scarcity of training samples in the prior art for pressure vessel structure deformation prediction, and achieves the technical effect of improving the prediction accuracy of pressure vessel structure deformation by wall thickness partitioning, sample set expansion and confidence verification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of predictive analysis, and particularly relates to a pressure vessel structure deformation prediction method based on wall thickness time-varying analysis. BACKGROUND

[0002] As key equipment in the petroleum chemical industry, energy industry, pharmaceutical industry and other industries, the structural safety of pressure vessels is directly related to production safety. In the long-term service process, the pressure vessel is affected by internal medium pressure, temperature change and corrosion and other factors, and the wall thickness of the container will degenerate in time-varying, which will cause structural deformation, and in severe cases, may lead to safety accidents.

[0003] At present, the pressure vessel structure deformation prediction mainly adopts finite element analysis method and machine learning method. Although the finite element analysis method has a solid theoretical basis, it has high computational complexity and is difficult to accurately simulate the time-varying characteristics of material performance. The machine learning method has strong nonlinear fitting ability, but in practical application, it faces significant difficulties: on the one hand, the general trained machine learning model cannot effectively predict the specific structure of the container, because the structural parameters and working conditions of different containers are quite different; on the other hand, when a special prediction model is established for a specific structure of the container, due to the limited historical monitoring data of the structure of the container, the training sample is scarce, which leads to poor generalization ability of the model and insufficient prediction accuracy. SUMMARY

[0004] The present application provides a pressure vessel structure deformation prediction method based on wall thickness time-varying analysis to solve the technical problem of insufficient prediction accuracy caused by the scarcity of training samples in the prior art.

[0005] The technical solution of the present application to solve the above technical problems is as follows:

[0006] The present application provides a pressure vessel structure deformation prediction method based on wall thickness time-varying analysis, comprising: retrieving a global wall thickness monitoring sample set with container model, service time and storage medium parameters as constraints; according to the design parameters of the pressure vessel, performing wall thickness partitioning to obtain a plurality of wall thickness partitions; traversing the plurality of wall thickness partitions to retrieve a plurality of local inner wall pressure monitoring sample sets with storage medium parameters, partition container structure, and partition container size distribution as constraints, and counting a plurality of partition inner wall pressures; traversing the plurality of partition inner wall pressures, combining the partition container structure, the partition container size distribution and the partition container material as constraints to retrieve a plurality of local wall thickness monitoring sample sets, and counting a plurality of wall thickness change time series information; based on the global wall thickness monitoring sample set, traversing the plurality of wall thickness change time series information to perform confidence analysis to obtain a plurality of confidence degrees; when the plurality of confidence degrees are all greater than or equal to a confidence threshold, adding the plurality of wall thickness change time series information into the pressure vessel structure deformation prediction result.

[0007] Optionally, according to the pressure vessel design parameters, the wall thickness partitioning is performed to obtain a plurality of wall thickness partitions, and the method further comprises: when the global wall thickness monitoring sample set is greater than or equal to the fitting data amount threshold, global wall thickness change time sequence information is counted according to the global wall thickness monitoring sample set, and the pressure vessel structure deformation prediction result is added; when the global wall thickness monitoring sample set is less than the fitting data amount threshold, the wall thickness partitioning is performed according to the pressure vessel design parameters to obtain a plurality of wall thickness partitions.

[0008] Optionally, according to the pressure vessel design parameters, the wall thickness partitioning is performed to obtain a plurality of wall thickness partitions, and the method further comprises: step one, extracting wall thickness design distribution information from the pressure vessel design parameters; step two, randomly extracting a first position design wall thickness from the wall thickness design distribution information; step three, extracting a first adjacent position design wall thickness of the first position design wall thickness, and comparing the first position design wall thickness and the first adjacent position design wall thickness to obtain a first wall thickness deviation; step four, when the first wall thickness deviation is less than or equal to a wall thickness deviation threshold, adding the first adjacent position and the first position into a first position homologous point; step five, taking the first position homologous point as a reference, returning to step three to perform a loop; step six, when the first wall thickness deviation is greater than the wall thickness deviation threshold, setting the first adjacent position as a second position, setting the first adjacent position design wall thickness as a second position design wall thickness, and returning to step two to perform a loop; and when all distribution positions are traversed, constructing the plurality of wall thickness partitions based on a plurality of groups of position homologous points.

[0009] Optionally, the plurality of wall thickness partitions are traversed to retrieve a plurality of local inner wall pressure monitoring sample sets with the storage medium parameters, the partition container structure, and the partition container size distribution as constraints, and a plurality of partition inner wall pressures are counted, which comprises: extracting a first wall thickness partition from the plurality of wall thickness partitions; extracting a first partition container structure and a first partition size distribution from the pressure vessel design parameters; retrieving a first local inner wall pressure monitoring sample set with the storage medium parameters, the first partition container structure, and the first partition size distribution as constraints; removing outlier inner wall pressure monitoring samples in the first local inner wall pressure monitoring sample set, performing mean value calculation to obtain a first partition inner wall pressure, and adding the first partition inner wall pressure into the plurality of partition inner wall pressures.

[0010] Optionally, retrieving the first local inner wall pressure monitoring sample set with the storage medium parameter, the first partition container structure, and the first partition size distribution as constraints comprises: obtaining a to-be-analyzed inner wall pressure monitoring sample, wherein the to-be-analyzed inner wall pressure monitoring sample has a storage medium record parameter, a container structure record parameter, a size distribution record parameter, and a to-be-analyzed inner wall pressure monitoring value; when the storage medium record parameter is consistent with the storage medium parameter, adding the container structure record parameter, the size distribution record parameter, and the to-be-analyzed inner wall pressure monitoring value into a first-level local inner wall pressure monitoring sample set; traversing the first-level local inner wall pressure monitoring sample set, and when the container structure record parameter is consistent with the first partition container structure, adding the size distribution record parameter and the to-be-analyzed inner wall pressure monitoring value into a second-level local inner wall pressure monitoring sample set; and traversing the second-level local inner wall pressure monitoring sample set, and when the size distribution record parameter is consistent with the first partition size distribution, adding the to-be-analyzed inner wall pressure monitoring value into the first local inner wall pressure monitoring sample set.

[0011] Optionally, traversing the plurality of partition inner wall pressures, retrieving a plurality of local wall thickness monitoring sample sets with the partition container structure, the partition container size distribution, and the partition container material as constraints, and counting a plurality of wall thickness change time sequence information comprises: extracting a first partition inner wall pressure from the plurality of partition inner wall pressures; extracting a first partition container structure, a first partition size distribution, and a first partition container material from the pressure container design parameters; multiplying the service length by a preset multiple to obtain an index service length, wherein 2≥ the preset multiple ≥1.2; retrieving a first local wall thickness monitoring sample set with the first partition inner wall pressure, the first partition container structure, the first partition size distribution, the first partition container material, and the index service length as constraints; performing pairwise sequence similarity calculation on the first local wall thickness monitoring sample set to obtain a sequence similarity set; and based on the sequence similarity set, traversing the first local wall thickness monitoring sample set to obtain a first partition wall thickness change time sequence information, and adding the first partition wall thickness change time sequence information into the plurality of wall thickness change time sequence information.

[0012] Optionally, based on the global wall thickness monitoring sample set, confidence analysis is performed on the plurality of wall thickness change time sequence information to obtain a plurality of confidence degrees, including: extracting first partition wall thickness change time sequence information from the plurality of wall thickness change time sequence information; extracting a first global wall thickness monitoring sample from the global wall thickness monitoring sample set, wherein the first global wall thickness monitoring sample has first global wall thickness change time sequence information; after aligning the first global wall thickness change time sequence information and the first partition wall thickness change time sequence information in head time, performing sequence similarity calculation on the aligned part to obtain a first sequence similarity; until a Qth global wall thickness monitoring sample is extracted from the global wall thickness monitoring sample set, wherein the Qth global wall thickness monitoring sample has Qth global wall thickness change time sequence information, and Q represents the number of global wall thickness monitoring sample sets; after aligning the Qth global wall thickness change time sequence information and the first partition wall thickness change time sequence information in head time, performing sequence similarity calculation on the aligned part to obtain a Qth sequence similarity; and the proportion of sequence similarities greater than or equal to a sequence similarity threshold in the first sequence similarity to the Qth sequence similarity is taken as a first partition confidence degree, which is added to the plurality of confidence degrees.

[0013] Optionally, the sequence similarity represents the proportion of time points at which the wall thickness deviation at the same time is less than or equal to a wall thickness deviation threshold.

[0014] The beneficial effects of the present application are:

[0015] The global wall thickness monitoring sample set is retrieved with the container model, service length and storage medium parameters as constraints, thereby providing data support for subsequent analysis; the wall thickness partition is performed according to the pressure container design parameters to obtain a plurality of wall thickness partitions, and the overall structure is refined into local areas through partition processing, thereby creating conditions for expanding the sample quantity; the plurality of local inner wall pressure monitoring sample sets are retrieved with the storage medium parameters, partition container structure, partition container size distribution as constraints, the plurality of partition inner wall pressures are counted, and thus the pressure load information of each partition is obtained, thereby providing the load boundary condition for the wall thickness change analysis; the plurality of local wall thickness monitoring sample sets are retrieved with the partition container structure, partition container size distribution and partition container material as constraints, the plurality of wall thickness change time sequence information are counted, and thus the change rule of the wall thickness of each partition with time is obtained, thereby realizing the expansion of the local sample set; the confidence analysis is performed on the plurality of wall thickness change time sequence information based on the global wall thickness monitoring sample set to obtain a plurality of confidence degrees, thereby verifying the reliability of the local prediction result by using the global sample, and ensuring the accuracy of the prediction result; when the plurality of confidence degrees are all greater than or equal to a confidence degree threshold, the plurality of wall thickness change time sequence information are added to the pressure container structure deformation prediction result, the reliable prediction result is selected by setting the confidence degree threshold, and the quality of the output result is ensured.

[0016] By the technical scheme, the global search is combined with the local partition, on the one hand, the whole container is divided into multiple local areas by the wall thickness partition, so that more sample data similar in structure can be searched in each local area, the training sample set is effectively expanded, and the key problem of sample scarcity is solved; on the other hand, the local prediction result is verified by the global wall thickness monitoring sample set, a reliable prediction result screening mechanism is established, and only the prediction result with high confidence is adopted. Both the technical defects that the general model is not applicable and the special model is insufficient in sample in the traditional machine learning method are overcome, and the problem of high calculation complexity in the finite element analysis method is avoided, accurate prediction of the structural deformation of the pressure container of different specifications and different working conditions is realized, and the accuracy of the pressure container structural deformation prediction and the practicality of the engineering application are improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A flowchart of a pressure container structure deformation prediction method based on wall thickness time-varying analysis is provided.

[0018] Figure 2 A flowchart of obtaining a plurality of wall thickness partitions is provided. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0020] In the description of the present application, the terms "first", "second" are only used for description purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0021] In the description of the present application, the term "for example" is used to mean "serving as an example, instance, or illustration." Any embodiment described as "for example" in the present application is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to make and use the application. In the following description, for the purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present application. It will be apparent, however, to one skilled in the art that the present application can be practiced without using these specific details. In other instances, well-known structures and processes are not elaborated upon in order to avoid unnecessary detail, which can obscure the description of the present application. Thus, the present application is not intended to be limited by the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0022] As Figure 1 shown, the embodiment of the present application provides a pressure vessel structure deformation prediction method based on wall thickness time-varying analysis, comprising:

[0023] S1, with the container model, service life and storage medium parameters as constraints, retrieving the global wall thickness monitoring sample set.

[0024] Specifically, first, the basic parameter information of the pressure vessel to be analyzed is obtained, including the container model, service life and storage medium parameters. Among them, the container model is used to identify the structure type and specification characteristics of the pressure vessel, such as the geometric shape (spherical, cylindrical, elliptical, etc.) of the pressure vessel to be analyzed, the nominal diameter, the nominal volume, the design pressure grade, etc. Standardized parameters are recorded in the design drawings or product nameplate of the pressure vessel to be analyzed; the service life represents the actual service life of the pressure vessel, that is, the actual running time of the pressure vessel to be analyzed from the time of putting into use to the current analysis point, recorded in units of years; the storage medium parameters include the medium type (such as the specific name of chemical medium such as liquefied petroleum gas, liquid ammonia, liquid chlorine and its corrosion grade), medium pressure (numerical range of working pressure, unit: MPa), medium temperature (numerical range of working temperature, unit: ℃).

[0025] A wall thickness monitoring database is established in advance, which adopts a structured storage method. Each record in the wall thickness monitoring database contains fields such as container model, service life, storage medium parameters, monitoring time sequence, corresponding wall thickness measurement value sequence, and monitoring position coordinates. A large amount of historical wall thickness monitoring data of pressure vessels under different conditions is stored in the wall thickness monitoring database.

[0026] Next, the container model, service time and storage medium parameters of the pressure container to be analyzed are taken as constraint conditions, and a retrieval operation is performed from the pre-established wall thickness monitoring database. Specifically, historical monitoring data with the same or similar container model as the container model of the pressure container to be analyzed is selected, while the service time of these data is required to be within a reasonable range, and the storage medium parameters meet the consistency requirements. For example, for the container model constraint, when the container model of the pressure container to be analyzed is completely consistent with the container model of a record in the database; or by selecting the records with similarity greater than a set threshold (such as 0.85) through container structure similarity calculation (Euclidean distance based on geometric parameters). For the service time constraint, records with service time within ±20% of the service time of the container to be analyzed are selected. For the storage medium parameter constraint, the medium type requires complete matching, and the medium pressure and temperature allow a deviation range of ±10%.

[0027] After the above constraint screening, the monitoring data records that meet all conditions are collected to form a global wall thickness monitoring sample set. Each sample in the global wall thickness monitoring sample set contains complete wall thickness change time series data, which provides a reliable data basis for subsequent wall thickness change analysis and structure deformation prediction. Such retrieval method ensures the relevance and applicability of sample data, avoids prediction deviation caused by unmatched data, and improves the accuracy and reliability of prediction results.

[0028] S2, according to the pressure container design parameters, performing wall thickness partitioning, obtaining a plurality of wall thickness partitions.

[0029] Specifically, first, the wall thickness design distribution information is extracted from the pressure container design parameters of the pressure container to be analyzed. The wall thickness design distribution information includes the design wall thickness values of each position on the surface of the pressure container to be analyzed, stored in the form of position coordinates in a three-dimensional coordinate system and corresponding design wall thickness values. The wall thickness design distribution information can be obtained from the design drawings of the pressure container to be analyzed.

[0030] Next, a region growing algorithm based on wall thickness similarity is used to perform the partitioning operation. Specifically, a random unassigned location is selected from the wall thickness design distribution information as a seed point, and the design wall thickness of this location is extracted as a reference wall thickness. Then, the neighboring locations of the seed point are searched (usually using a six-neighborhood or twenty-six-neighborhood search strategy), and the wall thickness deviation between the design wall thickness of the neighboring location and the reference wall thickness is calculated. When the wall thickness deviation is less than or equal to a preset wall thickness deviation threshold (such as 5%-15% of the reference wall thickness), the neighboring location is assigned to the current partition, and the neighboring location is used as a new expansion point to continue searching outward. When the wall thickness deviation of all neighboring locations is greater than the wall thickness deviation threshold, the growth process of the current partition ends. Then, the above partitioning process is repeated until all locations in the wall thickness design distribution information are assigned to the corresponding partition. During the partitioning process, the boundary information, average wall thickness value, location range, and other characteristic parameters of each partition are recorded.

[0031] After the above partitioning process, a plurality of wall thickness partitions are obtained, and the locations within each wall thickness partition have similar wall thickness design values, and there are obvious wall thickness differences between the partitions. Each wall thickness partition includes location coordinates, average design wall thickness, wall thickness coefficient of variation, and other description information of the partition. This partitioning strategy can effectively identify the wall thickness distribution characteristics of the pressure vessel, provide a spatial division basis for subsequent local analysis based on partitioning, and increase the number of similar samples in the local region, solving the problem of insufficient overall samples of fixed structure containers.

[0032] S3, traverse the plurality of wall thickness partitions, and retrieve a plurality of local inner wall pressure monitoring sample sets with the storage medium parameters, partition container structures, and partition container size distribution as constraints, and count the number of inner wall pressures in the partitions.

[0033] Specifically, first, the plurality of wall thickness partitions obtained are traversed, and independent inner wall pressure analysis is performed on each wall thickness partition. For the wall thickness partition being processed, the partition container structure and the partition container size distribution corresponding to the wall thickness partition are extracted. The partition container structure describes the geometric shape characteristics of the wall thickness partition, such as planar, cylindrical, spherical, ellipsoidal, and other curved surface types, as well as the position attribute of the wall thickness partition in the overall container (such as the cylinder part, the head part, the nozzle part, etc.); the partition container size distribution includes quantitative description information of the geometric size parameters of the wall thickness partition, such as the local curvature radius, the partition area, and the partition boundary size.

[0034] Next, multiple constraints are established for sample retrieval. Using the storage medium parameters of the pressure vessel to be analyzed, the structure of the vessel within the current wall thickness zone, and the size distribution of the vessel within the current wall thickness zone as constraints, matching historical data are retrieved from a pre-established internal wall pressure monitoring database. This database stores a large amount of internal wall pressure monitoring data for different pressure vessels under various operating conditions. Each record in the database includes fields such as storage medium parameters, local structural features of the vessel, size distribution information, monitoring time series, and the corresponding internal wall pressure value series.

[0035] Then, a hierarchical screening and matching process is performed. First, a primary screening is conducted based on storage medium parameters, requiring a perfect match in medium type and that medium pressure and temperature are within allowable deviation ranges (e.g., ±10%). Next, a secondary screening is performed based on the partition container structure, calculating structural similarity using the cross-union ratio (CUP) of structural feature descriptors, and selecting records with a CUP greater than or equal to a threshold (e.g., 0.7). Finally, a tertiary screening is performed based on the partition container size distribution, calculating the CUP of the size distribution, and selecting records with a CUP greater than or equal to a threshold, ultimately obtaining the local internal wall pressure monitoring sample set for the current wall thickness partition. Then, the obtained local internal wall pressure monitoring sample set is processed. First, outlier internal wall pressure monitoring samples are identified and removed, typically using an outlier detection method based on quartile intervals, considering data deviating from the normal range by more than three standard deviations as outliers. Then, the mean is calculated for the remaining valid samples to obtain the representative internal wall pressure value for that partition, i.e., the partition internal wall pressure.

[0036] Repeat the above process until the internal wall pressure statistics for all wall thickness zones are completed. Then, summarize the internal wall pressure of each zone to form several zone internal wall pressures. This internal wall pressure analysis based on wall thickness zones can accurately reflect the stress conditions in different areas of the pressure vessel, providing support for subsequent wall thickness variation analysis. At the same time, the data relevance and representativeness are ensured through sample retrieval with multiple constraints.

[0037] S4. Traverse the pressure of the inner wall of the several partitions, and combine the partition container structure, the partition container size distribution and the partition container material as constraints to retrieve several local wall thickness monitoring sample sets and statistically analyze several wall thickness change time series information.

[0038] Specifically, firstly, the obtained wall pressures of several zones are sequentially traversed, and an independent wall thickness variation analysis is performed on each zone. For the wall pressure of the currently processed zone, the zone's container structure, container size distribution, and container material are extracted. The container material includes material grade (e.g., 16MnR, Q345R), material mechanical property parameters (e.g., yield strength, tensile strength, elastic modulus), and material corrosion resistance level, among other material characteristic parameters. These parameters directly affect the wall thickness variation behavior of the zone under specific conditions.

[0039] Next, comprehensive constraints are established for sample retrieval. Using the intra-zone wall pressure, container structure, container size distribution, and container material of the current zone as constraints, matching historical data are retrieved from a pre-established wall thickness monitoring database. This database stores a large amount of time-series wall thickness monitoring data for local areas of pressure vessels. Each record in the database includes fields such as intra-zone pressure range, local structural features, size distribution, material parameters, and the corresponding wall thickness change time series.

[0040] Then, a hierarchical screening and matching process is executed. First, screening is performed based on the internal wall pressure of the partition, selecting records whose internal wall pressure falls within the allowable deviation range for the current partition's internal wall pressure. Next, screening is performed based on the partition's container structure, using the intersection-union ratio (CUI) of structural feature descriptors to select records with structural similarity greater than a set threshold. Then, screening is performed based on the partition's container size distribution, calculating the similarity of size distribution parameters. Finally, screening is performed based on the partition's container material, requiring matching material types and key performance parameters within the allowable deviation range, thus obtaining the local wall thickness monitoring sample set for the current partition. Afterwards, statistical analysis is performed on the local wall thickness monitoring sample set to extract typical wall thickness variation time-series patterns, obtaining the wall thickness variation time-series information for that partition. This process is repeated until the internal wall pressure of all partitions has been processed, and the wall thickness variation time-series information for each partition is summarized to form several wall thickness variation time-series information sets.

[0041] By retrieving and statistically analyzing wall thickness monitoring samples based on multiple constraints, we can accurately capture the wall thickness evolution patterns of different zones under specific working conditions and material conditions, laying the foundation for subsequent confidence analysis and prediction result verification.

[0042] S5. Based on the global wall thickness monitoring sample set, perform confidence analysis by traversing the several wall thickness change time series information to obtain several confidence levels.

[0043] Specifically, firstly, the obtained time-series information on wall thickness changes is sequentially traversed, and the confidence level of the time-series information on wall thickness changes in each partition is independently evaluated. For the current partition's time-series information on wall thickness changes, it is used as the prediction result to be verified, and is prepared to be compared and verified with the actual monitoring data in the global wall thickness monitoring sample set.

[0044] Next, each global wall thickness monitoring sample in the obtained global wall thickness monitoring sample set is extracted one by one. Each global wall thickness monitoring sample contains complete global wall thickness change time sequence information, which records the actual wall thickness change history of the pressure vessel under similar working conditions as verification reference data.

[0045] Then, time sequence data alignment and similarity calculation are performed. The current partition wall thickness change time sequence information is aligned with each global wall thickness change time sequence information at the head time, that is, the starting time of the two time sequences is taken as the reference for time axis calibration. On the basis of alignment, the sequence similarity of the aligned part is calculated for the overlapping time period. The calculation of sequence similarity is based on the statistical analysis of wall thickness deviation at the same time. Specifically, the number of time points at which the wall thickness deviation is less than or equal to the wall thickness deviation threshold is counted, and the proportion of the number of time points at which the wall thickness deviation is less than or equal to the wall thickness deviation threshold in the total number of time points is calculated. The proportion is the sequence similarity.

[0046] Further, the above comparison process is repeated until the similarity calculation of the current partition wall thickness change time sequence information and all samples in the global wall thickness monitoring sample set is completed, a series of sequence similarity values are obtained, and a sequence similarity set is formed.

[0047] After that, confidence statistics are performed based on the sequence similarity set. The number of sequence similarities in the sequence similarity set that are greater than or equal to the sequence similarity threshold (such as 0.75) is counted, and the proportion of the number of sequence similarities in the total number of global wall thickness monitoring samples is calculated. The proportion is the confidence of the current partition. The above process is repeated until the confidence of all wall thickness change time sequence information is evaluated, and the confidences of the partitions are summarized to form a number of confidences.

[0048] Through the confidence analysis based on the global sample verification, the reliability of the prediction results of the wall thickness change of each partition can be objectively evaluated. Through comparison with a large amount of actual monitoring data, the accuracy and practicality of the prediction results are ensured, and a quantitative confidence index is provided for the final prediction result screening.

[0049] S6, when the number of confidences are greater than or equal to the confidence threshold, the number of wall thickness change time sequence information is added to the pressure vessel structure deformation prediction result.

[0050] Specifically, first, the obtained number of confidences is subjected to threshold judgment. The confidence threshold is set as the minimum standard of the confidence of the prediction result, and the confidence threshold is determined according to engineering practice experience and safety requirements, and is generally set between 0.7 and 0.9. The confidence corresponding to each partition is checked one by one to determine whether it meets the confidence threshold requirement.

[0051] When the confidence of all partitions is greater than or equal to the confidence threshold, it indicates that the wall thickness change prediction results of each partition are reliable enough. At this time, it is considered that the wall thickness change time series information obtained based on the partition analysis is reliable and can be used as an effective result of the pressure vessel structure deformation prediction. Subsequently, the wall thickness change time series information verified by the confidence is integrated and processed as the pressure vessel structure deformation prediction result, including the position information of each partition, the wall thickness change time series data, and the corresponding confidence.

[0052] On the contrary, if the confidence of any partition is less than the confidence threshold, it indicates that the wall thickness change prediction result of the partition is not reliable enough, and the relevant wall thickness change time series information will not be added to the final prediction result to ensure the overall reliability and engineering practicability of the prediction result.

[0053] Through the result screening mechanism based on the confidence threshold, the quality and reliability of the pressure vessel structure deformation prediction result are effectively guaranteed, and the problem of affecting the overall prediction effect due to inaccurate local prediction is avoided, which provides reliable technical support for the safe operation of the pressure vessel.

[0054] Further, according to the pressure vessel design parameters, wall thickness partitioning is performed to obtain a plurality of wall thickness partitions, which further includes:

[0055] S7, when the global wall thickness monitoring sample set is greater than or equal to the fitting data amount threshold, global wall thickness change time series information is counted according to the global wall thickness monitoring sample set, and is added to the pressure vessel structure deformation prediction result;

[0056] S8, when the global wall thickness monitoring sample set is less than the fitting data amount threshold, wall thickness partitioning is performed according to the pressure vessel design parameters to obtain a plurality of wall thickness partitions.

[0057] In a feasible implementation, before performing the wall thickness partitioning operation in step S2, the size of the global wall thickness monitoring sample set is first evaluated, and a corresponding processing strategy is selected according to the sufficiency of the sample quantity.

[0058] When the sample quantity of the obtained global wall thickness monitoring sample set is greater than or equal to the fitting data amount threshold, it indicates that the existing global sample data is sufficient to support direct global analysis. Specifically, first, the fitting data amount threshold is set as a standard for judging the sufficiency of the sample, which is determined according to statistical principles and engineering experience, and is generally set between 50-100 samples to ensure the effectiveness of the statistical analysis and the reliability of the result.

[0059] Then, when the number of samples in the global wall thickness monitoring sample set meets the threshold requirement, global wall thickness change analysis is directly performed based on the global wall thickness monitoring sample set. By statistically processing the wall thickness change time series data of all samples in the global wall thickness monitoring sample set, a typical wall thickness change mode is extracted by using a weighted average, median statistics or trend fitting method, and global wall thickness change time series information is obtained. The global wall thickness change time series information reflects the wall thickness evolution law of the pressure vessel as a whole under similar working conditions. Then, the obtained global wall thickness change time series information is directly added to the pressure vessel structure deformation prediction result as the final output of the overall prediction. Since the number of samples is sufficient, complex partition analysis is not required at this time, and reliable prediction results can be quickly obtained, thereby improving the analysis efficiency.

[0060] When the number of samples in the global wall thickness monitoring sample set obtained in step S1 is less than the fitting data amount threshold, it indicates that the existing global sample data is insufficient to support reliable global analysis, and global statistics based on limited samples may lead to insufficient representativeness and low reliability of the prediction results. At this time, the partition analysis strategy is started, wall thickness partitioning is performed according to the pressure vessel design parameters, and a plurality of wall thickness partitions are obtained. By converting the overall analysis problem into a plurality of local analysis problems, more samples of the same type can be retrieved in each partition, effectively solving the problem of insufficient global samples. This partition strategy can fully utilize the local structural similarity and expand the range of available samples, laying a foundation for subsequent intra-partition wall pressure analysis and wall thickness change prediction.

[0061] Through the adaptive analysis strategy selection mechanism, the analysis method is automatically adjusted according to the data availability, which ensures the analysis efficiency when the number of samples is sufficient and ensures the prediction reliability when the number of samples is insufficient.

[0062] Further, as shown in Figure 2 According to the pressure vessel design parameters, wall thickness partitioning is performed to obtain a plurality of wall thickness partitions, including:

[0063] Step one: extracting wall thickness design distribution information from the pressure vessel design parameters;

[0064] Step two: randomly extracting a first position design wall thickness from the wall thickness design distribution information;

[0065] Step three: extracting a first adjacent position design wall thickness of the first position design wall thickness, and comparing the first position design wall thickness and the first adjacent position design wall thickness to obtain a first wall thickness deviation;

[0066] Step four: when the first wall thickness deviation is less than or equal to a wall thickness deviation threshold, adding the first adjacent position and the first position into the first position same type point;

[0067] Step five: take the first position homologous point as the reference, return to step three to execute the loop;

[0068] Step six: when the first wall thickness deviation is greater than the wall thickness deviation threshold, set the first adjacent position as the second position, set the first adjacent position design wall thickness as the second position design wall thickness, and return to step two to execute the loop;

[0069] When all the distribution positions are traversed, the wall thickness partitions are constructed based on a plurality of position homologous points.

[0070] In a preferred embodiment, the wall thickness partition operation is performed based on wall thickness similarity, and the pressure vessel to be analyzed is divided into a plurality of partitions with similar wall thickness characteristics through step-by-step expansion and clustering, thereby obtaining a plurality of wall thickness partitions.

[0071] First, step one is executed to extract the wall thickness design distribution information from the pressure vessel design parameters. The wall thickness design distribution information includes the design wall thickness values of each discrete position on the surface of the pressure vessel, which is stored in a grid form, with each grid point corresponding to a spatial position coordinate and a corresponding design wall thickness value. These data can be obtained from the design drawings or CAD model of the pressure vessel, forming complete wall thickness design distribution information. Then, step two is executed to randomly select a position that has not been assigned to any partition from the wall thickness design distribution information as the first position, and extract the design wall thickness of the first position as the first position design wall thickness. The first position serves as the seed point for the growth of the current partition, and is used to start the new partition construction process. The random selection strategy ensures the unbiasedness and stability of the partition.

[0072] Subsequently, step three is executed to search for the adjacent positions of the first position, extract one of the adjacent positions as the first adjacent position, and take the design wall thickness corresponding to the first adjacent position as the first adjacent position design wall thickness. The determination of the adjacent position can use spatial adjacency relationship, such as four-neighborhood, eight-neighborhood or higher-order neighborhood search strategy. The absolute difference between the first position design wall thickness and the first adjacent position design wall thickness is calculated to obtain the first wall thickness deviation, which is used to judge whether the two positions have sufficient wall thickness similarity.

[0073] Then, the first wall thickness deviation is compared with a preset wall thickness deviation threshold value. When the first wall thickness deviation is less than or equal to the preset wall thickness deviation threshold value, it is considered that the first adjacent position and the first position have similar wall thickness characteristics and should be attributed to the same partition. At this time, step four is executed to add the first adjacent position and the first position into the first position homologous point at the same time. The first position homologous point records all positions with similar wall thickness characteristics in the current partition. Among them, the wall thickness deviation threshold value is usually set to 5%-15% of the reference wall thickness to balance the fineness and rationality of the partition. Then, step five is executed to continue expanding the current partition based on the updated first position homologous point. For each position in the first position homologous point, the adjacent position search and wall thickness deviation calculation process of step three are repeated to form an iterative loop. This set-based expansion strategy can make the partition boundary grow outward gradually until a region with significant wall thickness difference is encountered.

[0074] After that, when the first wall thickness deviation is greater than the wall thickness deviation threshold value, it indicates that the first adjacent position has significant difference in wall thickness characteristics from the current partition and should not be attributed to the current partition. At this time, step six is executed to re-label the first adjacent position as a second position and re-label the corresponding first adjacent position design wall thickness as a second position design wall thickness, and return to step two to start a new partition construction process. This processing method ensures that a region with significant wall thickness difference can form an independent partition.

[0075] The above steps are repeated until all distribution positions in the wall thickness design distribution information are traversed and assigned to the corresponding partitions. At this time, based on the obtained several groups of position homologous points, the corresponding several wall thickness partitions are constructed. Each partition contains a set of point sets with continuous spatial positions and similar wall thickness characteristics, and there is obvious wall thickness difference between the partitions. The finally obtained wall thickness partitions not only maintain the wall thickness consistency within each partition, but also ensure the integrity and rationality of the partition division, providing an analysis space unit for subsequent local analysis.

[0076] Further, the several wall thickness partitions are traversed to retrieve several local inner wall pressure monitoring sample sets with the storage medium parameters, partition container structure, and partition container size distribution as constraints, and the inner wall pressures of the several partitions are counted, including:

[0077] S31, extracting a first wall thickness partition from the several wall thickness partitions;

[0078] S32, extracting a first partition container structure and a first partition size distribution from the pressure container design parameters;

[0079] S33, retrieving a first local inner wall pressure monitoring sample set with the storage medium parameters, the first partition container structure, and the first partition size distribution as constraints;

[0080] S34, rejecting outlier inner wall pressure monitoring samples of the first local inner wall pressure monitoring sample set, performing mean value calculation to obtain a first partition inner wall pressure, and adding the first partition inner wall pressure into the plurality of partition inner wall pressures.

[0081] In a preferred embodiment, first, each partition is extracted in a predetermined order (such as spatial position order or partition number order) from the obtained plurality of wall thickness partitions for processing, one wall thickness partition is extracted each time, denoted as a first wall thickness partition, as the current processing object, the first wall thickness partition contains a set of position points with continuous spatial positions and similar wall thickness characteristics, as well as corresponding partition boundary information, average wall thickness value and other characteristic parameters.

[0082] Subsequently, the structure and size information corresponding to the first wall thickness partition is extracted from the pressure vessel design parameters as the first partition container structure and the first partition size distribution. The first partition container structure describes the geometric characteristics of the first wall thickness partition, including the local surface type (such as plane, cylindrical surface, spherical surface, ellipsoidal surface, etc.), curvature information, and the functional attribute of the partition in the overall container (such as cylinder part, head part, nozzle part, support part, etc.); the first partition size distribution includes the geometric size parameters of the first wall thickness partition, such as the length, width, area, volume, local curvature radius, wall thickness variation gradient, etc. quantitative description information, which can accurately represent the spatial geometric characteristics of the partition.

[0083] Then, multiple constraint conditions are established for sample retrieval. The storage medium parameters, the first partition container structure, and the first partition size distribution of the pressure vessel to be analyzed are used as constraint conditions to retrieve matching historical data from the pre-established inner wall pressure monitoring database. The inner wall pressure monitoring database stores a large amount of inner wall pressure monitoring data of local regions of different pressure vessels under various working conditions, each record contains complete information such as storage medium information, local container structure characteristics, size distribution parameters, monitoring time sequence and corresponding inner wall pressure value sequence. Through layer-by-layer screening and matching, a first local inner wall pressure monitoring sample set highly matched with the working condition and structure characteristics of the first wall thickness partition is obtained.

[0084] Subsequently, data quality control and statistical analysis are performed on the first local inner wall pressure monitoring sample set. First, statistical methods are used to identify and remove outlier inner wall pressure monitoring samples in the first local inner wall pressure monitoring sample set. For example, the mean and standard deviation of all inner wall pressure values in the sample set are calculated, and the 3σ criterion is used to detect outlier inner wall pressure monitoring samples, i.e. when the absolute deviation of a certain inner wall pressure monitoring value from the sample mean is greater than 3σ, the sample is identified as an outlier inner wall pressure monitoring sample and is removed. For another example, an outlier detection method based on interquartile range is used, and the specific process is as follows: first, the first quartile and the third quartile of the first local inner wall pressure monitoring sample set are calculated, and then the interquartile range between the first quartile and the third quartile is calculated; then the upper and lower boundaries of the outlier judgment are determined, the lower boundary is the first quartile minus 1.5 times the interquartile range, and the upper boundary is the third quartile plus 1.5 times the interquartile range; samples beyond the upper and lower boundaries are identified as outliers and are removed to ensure the reliability of the data. Then, the mean value of the first local inner wall pressure monitoring sample set after removing the outlier inner wall pressure monitoring samples is calculated to obtain the representative inner wall pressure value of the partition, i.e. the first partition inner wall pressure. The first partition inner wall pressure reflects the typical inner wall pressure level of the first partition under specific operating conditions and structural conditions. Then, the first partition inner wall pressure is added to the several partition inner wall pressures to provide support for subsequent wall thickness change analysis.

[0085] The above steps S31 to S34 are repeated to perform inner wall pressure statistical analysis on all wall thickness partitions in turn until the partition inner wall pressure corresponding to each partition is obtained. Finally, a complete set of several partition inner wall pressures is formed, which accurately reflects the stress condition of each local area of the pressure vessel under the current operating condition, providing accurate and reliable basic data for subsequent partition-based wall thickness change prediction.

[0086] Further, the storage medium parameters, the first partition container structure, and the first partition size distribution are used as constraints to retrieve the first local inner wall pressure monitoring sample set, including:

[0087] S331, obtaining a to-be-analyzed inner wall pressure monitoring sample, wherein the to-be-analyzed inner wall pressure monitoring sample has storage medium recording parameters, container structure recording parameters, size distribution recording parameters, and a to-be-analyzed inner wall pressure monitoring value;

[0088] S332, when the storage medium recording parameters and the storage medium parameters are consistent, adding the container structure recording parameters, the size distribution recording parameters, and the to-be-analyzed inner wall pressure monitoring value into the first local inner wall pressure monitoring sample set;

[0089] S333, traversing the primary local inner wall pressure monitoring sample set, when the container structure record parameter is consistent with the first partition container structure, adding the size distribution record parameter and the to-be-analyzed inner wall pressure monitoring value into the secondary local inner wall pressure monitoring sample set;

[0090] S334, traversing the secondary local inner wall pressure monitoring sample set, when the size distribution record parameter is consistent with the first partition size distribution, adding the to-be-analyzed inner wall pressure monitoring value into the first local inner wall pressure monitoring sample set.

[0091] In a preferred embodiment, first, to-be-analyzed inner wall pressure monitoring samples are obtained one by one from a pre-established inner wall pressure monitoring database. Each to-be-analyzed inner wall pressure monitoring sample contains a storage medium record parameter, a container structure record parameter, a size distribution record parameter, and a to-be-analyzed inner wall pressure monitoring value. The storage medium record parameter records the medium type, medium pressure, medium temperature, and other working condition information corresponding to the sample; the container structure record parameter describes the local container geometry, surface characteristics, functional attributes, and other structure information corresponding to the sample; the size distribution record parameter contains the geometric size, spatial distribution, boundary characteristics, and other quantitative parameters corresponding to the sample; and the to-be-analyzed inner wall pressure monitoring value is the actual measured inner wall pressure value of the sample.

[0092] Then, a first-level screening is performed based on the storage medium parameter for accurate matching. The storage medium record parameter of each to-be-analyzed inner wall pressure monitoring sample is directly compared with the storage medium parameter of the current analysis target. Specifically, when the medium type is completely the same, and the medium pressure and medium temperature deviations are both less than the corresponding deviation threshold, it is considered that the storage medium record parameter and the storage medium parameter are consistent. At this time, the container structure record parameter, the size distribution record parameter, and the to-be-analyzed inner wall pressure monitoring value of the sample are extracted and added to the primary local inner wall pressure monitoring sample set, thereby effectively eliminating samples that do not match the working condition, reducing the data amount for subsequent processing.

[0093] Subsequently, a second level screening is performed for similarity matching based on the container structure features. Each sample in the first level local inner wall pressure monitoring sample set is traversed, and its container structure record parameters are compared with the first partition container structure. Specifically, the structural similarity is evaluated by the intersection over union calculation of the structure feature descriptors, and when the intersection over union is greater than or equal to a preset intersection over union threshold, it is considered that the container structure record parameters are consistent with the first partition container structure. The structure feature descriptors include key structural elements such as geometric shape type, curvature feature, boundary shape, etc., and these feature elements constitute a feature set. The calculation of the structural similarity uses the intersection over union method, that is, the number of intersection elements of two structure feature sets divided by the number of union elements, specifically: structural similarity = number of intersection elements of feature set / number of union elements of feature set. The intersection of the feature set represents the number of feature elements common to both structures, and the union of the feature set represents the total number of non-repeating feature elements contained in both structures. The intersection over union value ranges from 0 to 1, and the larger the value, the higher the structural similarity. When the calculated intersection over union is greater than or equal to the preset intersection over union threshold (usually set to 0.7), it is considered that the container structure record parameters are consistent with the first partition container structure. The sample that meets the condition is added to the second level local inner wall pressure monitoring sample set with its size distribution record parameters and the to-be-analyzed inner wall pressure monitoring value.

[0094] Subsequently, a third level screening is performed for fine matching based on the size distribution features. Each sample in the second level local inner wall pressure monitoring sample set is traversed, and its size distribution record parameters are compared with the first partition size distribution. Specifically, the similarity of the two size distributions is calculated using the intersection over union method. The size distribution is composed of multiple independent geometric size parameters, including key size features such as feature length, feature width, local curvature radius, wall thickness gradient, etc. For each geometric size parameter, the overlap degree is calculated respectively: single geometric size parameter overlap degree = overlap interval length / total coverage interval length. The overlap interval length is the intersection interval length of the sample size parameter value range and the corresponding size parameter value range of the first partition, and the total coverage interval length is the union interval length of the two value ranges. Then, the overlap degrees of all geometric size parameters are integrated to calculate the overall size distribution intersection over union: size distribution intersection over union = weighted average of geometric size parameter overlap degrees. The weight is determined according to the importance of each geometric size parameter in describing the partition features. When the calculated size distribution intersection over union is greater than or equal to the preset intersection over union threshold (usually set to 0.7), it is considered that the size distribution record parameters are consistent with the first partition size distribution, and the to-be-analyzed inner wall pressure monitoring value of the sample is added to the first local inner wall pressure monitoring sample set.

[0095] Through the hierarchical screening strategy of layer-by-layer comparison, the matching range is gradually narrowed down, greatly reducing the data volume of each level of comparison, and improving the data processing efficiency. Secondly, different levels use different matching strategies to ensure the accuracy and reliability of the screening results, effectively optimize the consumption of computing resources, and are suitable for large-scale local inner wall pressure monitoring sample efficient retrieval.

[0096] Further, traversing the several partition inner wall pressures, combining the partition container structure, the partition container size distribution and the partition container material as constraints, retrieving several local wall thickness monitoring sample sets, and counting several wall thickness change time series information, including:

[0097] S41, extracting a first partition inner wall pressure from the several partition inner wall pressures;

[0098] S42, extracting a first partition container structure, a first partition size distribution and a first partition container material from the pressure vessel design parameters;

[0099] S43, multiplying the service length by a preset multiple to obtain an index service length, wherein 2≥preset multiple≥1.2;

[0100] S44, retrieving a first local wall thickness monitoring sample set with the first partition inner wall pressure, the first partition container structure, the first partition size distribution, the first partition container material and the index service length as constraints;

[0101] S45, calculating the sequence similarity of the first local wall thickness monitoring sample set, obtaining a sequence similarity set;

[0102] S46, based on the sequence similarity set, traversing the first local wall thickness monitoring sample set to obtain the first partition wall thickness change time series information, and adding it to the several wall thickness change time series information.

[0103] In a preferred embodiment, first, the inner wall pressure of each partition is sequentially extracted from the obtained inner wall pressures of several partitions in a predetermined order for processing. The first partition inner wall pressure is extracted first as the current processing object, which reflects the typical stress level of the first partition under specific working conditions. Then, detailed feature information corresponding to the first partition is extracted from the pressure vessel design parameters. Specifically, it includes: first partition container structure, describing the geometric shape characteristics, surface type, functional attributes and other structural information of the partition; first partition size distribution, containing the characteristic length, characteristic width, local curvature radius, partition area and other geometric size parameters of the partition; first partition container material, including material brand, material mechanical property parameters (such as yield strength, tensile strength, elastic modulus, etc.), material corrosion resistance performance level and other key material characteristic parameters. These parameters jointly determine the wall thickness variation behavior of the partition under specific conditions.

[0104] Subsequently, the retrieval time range of the wall thickness monitoring sample is determined. The service life of the pressure vessel to be analyzed is multiplied by a preset multiple to obtain an index service life, wherein the preset multiple is valued between 1.2 and 2.0. By expanding the retrieval time range, longer time span wall thickness change history data can be obtained, providing sufficient data support for wall thickness change trend prediction, while avoiding the introduction of irrelevant data due to too large time range.

[0105] Then, the first partition inner wall pressure, the first partition container structure, the first partition size distribution, the first partition container material and the index service life are used as constraint conditions to retrieve matching historical data from the pre-established wall thickness monitoring database. The retrieval process adopts the same layer-by-layer comparison strategy as step S33, gradually reducing the comparison data amount and transmission data amount, and improving the data processing efficiency. Through multi-constraint hierarchical screening, a first local wall thickness monitoring sample set highly matched with the working condition, structure characteristics, size distribution and material properties of the first partition is finally obtained.

[0106] Next, similarity analysis is performed on the first local wall thickness monitoring sample set. For the wall thickness change time series data in the first local wall thickness monitoring sample set, pairwise sequence similarity calculation is performed to obtain a sequence similarity set. Sequence similarity calculation is based on comparison analysis of equal-length sequences, which quantifies the similarity of wall thickness change patterns between different samples through dynamic time warping, Pearson correlation coefficient or cosine similarity, etc., providing a basis for subsequent abnormal sample identification and representative sample screening.

[0107] After that, sample quality assessment and screening are performed based on the sequence similarity set. Each sample in the first local wall thickness monitoring sample set is traversed, the similarity distribution with other samples is calculated, and abnormal samples are identified through outlier factor analysis. The outlier factor minimum sorting strategy is adopted to preferentially select typical samples with high similarity to most samples and to remove abnormal samples deviating from the mainstream mode. Based on the high-quality samples after screening, the first partition wall thickness change time series information is obtained through weighted average, median statistics or trend fitting, and is added to the wall thickness change time series information.

[0108] The processing procedures of steps S41 to S46 are repeated to sequentially perform wall thickness change time series analysis on the wall pressure in all partitions, until the wall thickness change time series information corresponding to each partition is obtained, and finally a complete wall thickness change time series information is formed, reflecting the wall thickness evolution law of each local area of the pressure vessel under specific working conditions and material conditions, and providing a reliable data basis for subsequent confidence analysis and prediction result verification.

[0109] Further, based on the global wall thickness monitoring sample set, confidence analysis is performed on the wall thickness change time series information to obtain a plurality of confidence degrees, including:

[0110] S51, extracting first partition wall thickness change time series information from the wall thickness change time series information;

[0111] S52, extracting a first global wall thickness monitoring sample from the global wall thickness monitoring sample set, wherein the first global wall thickness monitoring sample has first global wall thickness change time series information;

[0112] S53, after aligning the head time of the first global wall thickness change time series information and the first partition wall thickness change time series information, performing sequence similarity calculation on the aligned part to obtain a first sequence similarity;

[0113] S54, until the Qth global wall thickness monitoring sample is extracted from the global wall thickness monitoring sample set, wherein the Qth global wall thickness monitoring sample has Qth global wall thickness change time series information, and Q represents the number of global wall thickness monitoring sample sets;

[0114] S55, after aligning the head time of the Qth global wall thickness change time series information and the first partition wall thickness change time series information, performing sequence similarity calculation on the aligned part to obtain a Qth sequence similarity;

[0115] S56, the proportion of sequence similarities greater than or equal to the sequence similarity threshold in the first sequence similarity to the Qth sequence similarity is calculated and set as the first partition confidence degree, which is added to the plurality of confidence degrees.

[0116] In a preferred embodiment, the reliability of the prediction result of each partition is evaluated by comparing the wall thickness change time sequence information of each partition with the global wall thickness monitoring sample set, and the corresponding confidence is obtained.

[0117] First, from the obtained several wall thickness change time sequence information, the wall thickness change time sequence information of each partition is extracted in a predetermined order for confidence evaluation. Each time a wall thickness change time sequence information is extracted, it is recorded as the first partition wall thickness change time sequence information as the current processing object, which records the wall thickness evolution law of the first partition under specific working conditions.

[0118] Then, from the obtained global wall thickness monitoring sample set, the global wall thickness monitoring samples are extracted in sequence for comparison and verification. The first global wall thickness monitoring sample is extracted first, which contains complete first global wall thickness change time sequence information, recording the actual wall thickness change history of the pressure vessel under similar working conditions, as the verification reference data. Subsequently, the first global wall thickness change time sequence information is aligned with the first partition wall thickness change time sequence information at the head time, i.e. the starting time of the two time sequences is used as the reference for time axis calibration, to ensure the consistency of the comparison analysis. Based on the alignment, the sequence similarity calculation of the aligned part is performed for the time overlap part, and the first sequence similarity is obtained. The sequence similarity reflects the degree of agreement between the partition prediction result and the actual monitoring data, and the higher the value, the better the prediction reliability.

[0119] The above comparison process is repeated to process all samples in the global wall thickness monitoring sample set in sequence. The second, third, and Qth global wall thickness monitoring samples are extracted from the global wall thickness monitoring sample set one by one, where Q represents the total number of the global wall thickness monitoring sample set. For each global wall thickness monitoring sample, its corresponding global wall thickness change time sequence information is extracted and aligned with the first partition wall thickness change time sequence information at the head time, and then the sequence similarity calculation of the aligned part is performed to obtain the second sequence similarity, the third sequence similarity, and the Qth sequence similarity.

[0120] Then, the number of sequence similarities greater than or equal to the preset sequence similarity threshold value is counted from the first sequence similarity to the Qth sequence similarity, and the proportion of the total number Q of the global wall thickness monitoring sample set is calculated. This proportion reflects the degree of agreement between the first partition wall thickness change time sequence information and most of the actual monitoring data, which is set as the first partition confidence, and is added to the several confidences. The higher the confidence value, the more reliable the wall thickness change prediction result of the partition.

[0121] The processes of steps S51 to S56 are repeated to sequentially evaluate the confidence of all wall thickness change time sequence information until the confidence corresponding to each partition is obtained, and finally a complete number of confidences is formed, reflecting the confidence degree of the prediction result of the wall thickness change of each partition, thereby providing a quantitative evaluation basis for the screening and quality control of the final prediction result.

[0122] Further, the sequence similarity represents the proportion of the number of time points at which the wall thickness deviation at the same time point is less than or equal to the wall thickness deviation threshold value.

[0123] Specifically, the sequence similarity represents the proportion of the number of time points at which the wall thickness deviation at the same time point is less than or equal to the wall thickness deviation threshold value. When performing the sequence similarity calculation, first, the two aligned wall thickness change time sequences are compared time point by time point, and the wall thickness deviation at each corresponding time point is calculated, i.e., the absolute difference between the wall thickness values of the two time sequences at the same time point. Then, the wall thickness deviation at each time point is compared with the preset wall thickness deviation threshold value, and the number of time points at which the wall thickness deviation is less than or equal to the wall thickness deviation threshold value is counted. Finally, the number of time points satisfying the condition is divided by the total number of time points in the aligned part to obtain the sequence similarity, i.e., the calculation method of the sequence similarity is: sequence similarity = number of time points satisfying the condition / total number of time points in the aligned part. Wherein, the time point satisfying the condition refers to the time point at which the wall thickness deviation at the same time point is less than or equal to the wall thickness deviation threshold value, and the total number of time points in the aligned part is the total number of time points in the overlapping time period of the two time sequences.

[0124] The numerical range of the sequence similarity is between 0 and 1, and the closer the value is to 1, the higher the similarity of the two wall thickness change time sequences, and the better the consistency of the prediction result with the actual monitoring data. The wall thickness deviation threshold value is determined according to the engineering precision requirement and the measurement error range to ensure the rationality and practicability of the similarity evaluation.

[0125] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0126] Those skilled in the art will understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0127] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or blocks of the flowcharts. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or blocks of the flowcharts.

[0128] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or blocks of the flowcharts. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or blocks of the flowcharts.

[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or blocks of the flowcharts. Figure 1 one or more flowcharts and / or blocks in the flowcharts and / or blocks of the flowcharts.

[0130] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments described and shown, without departing from the spirit and scope of the application.

[0131] It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, it is intended that the present application cover the modifications and variations of this application and its equivalent.

Claims

1. A method for predicting deformation of a pressure vessel structure based on wall thickness time-varying analysis, characterized by, The method comprises the following steps: retrieving a global wall thickness monitoring sample set with container model, service time and storage medium parameters as constraints; performing wall thickness partitioning according to pressure vessel design parameters to obtain a plurality of wall thickness partitions; traversing the plurality of wall thickness partitions to retrieve a plurality of local inner wall pressure monitoring sample sets with storage medium parameters, partition container structure, and partition container size distribution as constraints, and to count a plurality of partition inner wall pressures; traversing the plurality of partition inner wall pressures to retrieve a plurality of local wall thickness monitoring sample sets with the partition container structure, the partition container size distribution and the partition container material as constraints, and to count a plurality of wall thickness change time series information; based on the global wall thickness monitoring sample set, traversing the plurality of wall thickness change time series information to perform confidence analysis to obtain a plurality of confidence degrees; when the plurality of confidence degrees are all greater than or equal to a confidence threshold, adding the plurality of wall thickness change time series information into the pressure vessel structure deformation prediction result.

2. The method of claim 1, wherein, According to the pressure vessel design parameters, the wall thickness partitioning is performed to obtain a plurality of wall thickness partitions, which further comprises the following steps: when the global wall thickness monitoring sample set is greater than or equal to a fitting data amount threshold, counting global wall thickness change time series information from the global wall thickness monitoring sample set and adding it into the pressure vessel structure deformation prediction result; when the global wall thickness monitoring sample set is less than the fitting data amount threshold, performing wall thickness partitioning according to the pressure vessel design parameters to obtain a plurality of wall thickness partitions.

3. The method of claim 1, wherein, According to the pressure vessel design parameters, the wall thickness partitioning is performed to obtain a plurality of wall thickness partitions, which comprises the following steps: Step 1: extracting wall thickness design distribution information from the pressure vessel design parameters; Step 2: randomly extracting a first position design wall thickness from the wall thickness design distribution information; Step 3: extracting a first adjacent position design wall thickness of the first position design wall thickness, and comparing the first position design wall thickness with the first adjacent position design wall thickness to obtain a first wall thickness deviation; Step 4: when the first wall thickness deviation is less than or equal to a wall thickness deviation threshold, adding the first adjacent position and the first position into a first position homologous point; Step 5: taking the first position homologous point as a reference to return to Step 3 for loop execution; Step 6: when the first wall thickness deviation is greater than the wall thickness deviation threshold, setting the first adjacent position as a second position, setting the first adjacent position design wall thickness as a second position design wall thickness, and returning to Step 2 for loop execution; when all distribution positions are traversed, constructing the plurality of wall thickness partitions based on a plurality of groups of position homologous points.

4. The method of claim 1, wherein, Traversing the plurality of wall thickness partitions to retrieve a plurality of local inner wall pressure monitoring sample sets with storage medium parameters, partition container structure, and partition container size distribution as constraints, and to count a plurality of partition inner wall pressures, which comprises the following steps: extracting a first wall thickness partition from the plurality of wall thickness partitions; extracting a first partition container structure and a first partition size distribution from the pressure vessel design parameters; retrieving a first local inner wall pressure monitoring sample set with the storage medium parameters, the first partition container structure, and the first partition size distribution as constraints; The outlier inner wall pressure monitoring sample of the first local inner wall pressure monitoring sample set is removed, a mean value calculation is performed, a first partition inner wall pressure is obtained, and the first partition inner wall pressure is added to the plurality of partition inner wall pressures.

5. The method of claim 4, wherein, The first local inner wall pressure monitoring sample set is retrieved with the storage medium parameter, the first partition container structure, and the first partition size distribution as constraints, and includes the following steps: Obtain the to-be-analyzed inner wall pressure monitoring sample, wherein the to-be-analyzed inner wall pressure monitoring sample has a storage medium record parameter, a container structure record parameter, a size distribution record parameter, and a to-be-analyzed inner wall pressure monitoring value; When the storage medium record parameter is consistent with the storage medium parameter, the container structure record parameter, the size distribution record parameter, and the to-be-analyzed inner wall pressure monitoring value are added to the first-level local inner wall pressure monitoring sample set; Traverse the first-level local inner wall pressure monitoring sample set, and when the container structure record parameter is consistent with the first partition container structure, the size distribution record parameter and the to-be-analyzed inner wall pressure monitoring value are added to the second-level local inner wall pressure monitoring sample set; Traverse the second-level local inner wall pressure monitoring sample set, and when the size distribution record parameter is consistent with the first partition size distribution, the to-be-analyzed inner wall pressure monitoring value is added to the first local inner wall pressure monitoring sample set.

6. The method of claim 1, wherein, Traverse the plurality of partition inner wall pressures, retrieve a plurality of local wall thickness monitoring sample sets in combination with the partition container structure, the partition container size distribution, and the partition container material as constraints, and count a plurality of wall thickness change time sequence information, including the following steps: Extract the first partition inner wall pressure from the plurality of partition inner wall pressures; Extract the first partition container structure, the first partition size distribution, and the first partition container material from the pressure container design parameters; Multiply the service time length by a preset multiple to obtain an index service time length, wherein 2≥preset multiple≥1.2; Retrieve the first local wall thickness monitoring sample set with the first partition inner wall pressure, the first partition container structure, the first partition size distribution, the first partition container material, and the index service time length as constraints; Perform pairwise sequence similarity calculation on the first local wall thickness monitoring sample set to obtain a sequence similarity set; Based on the sequence similarity set, traverse the first local wall thickness monitoring sample set to obtain the first partition wall thickness change time sequence information, and add the first partition wall thickness change time sequence information to the plurality of wall thickness change time sequence information.

7. The method of claim 1, wherein, Based on the global wall thickness monitoring sample set, traverse the plurality of wall thickness change time sequence information to perform confidence analysis and obtain a plurality of confidence degrees, including the following steps: Extract the first partition wall thickness change time sequence information from the plurality of wall thickness change time sequence information; Extract the first global wall thickness monitoring sample from the global wall thickness monitoring sample set, wherein the first global wall thickness monitoring sample has a first global wall thickness change time sequence information; After aligning the first global wall thickness change time sequence information and the first partition wall thickness change time sequence information at the head time, perform sequence similarity calculation on the aligned part to obtain a first sequence similarity; until from the global wall thickness monitoring sample set, extract the Qth global wall thickness monitoring sample, wherein the Qth global wall thickness monitoring sample has Qth global wall thickness change time sequence information, and Q represents the number of global wall thickness monitoring sample sets; After aligning the Qth global wall thickness change time sequence information and the first partition wall thickness change time sequence information in terms of head time, perform sequence similarity calculation on the aligned part to obtain the Qth sequence similarity. Count the proportion of sequence similarities greater than or equal to the sequence similarity threshold value in the first sequence similarity to the Qth sequence similarity, and set the proportion as a first partition confidence, and add the first partition confidence to the plurality of confidences.

8. The method of any one of claims 6 or 7, wherein, The sequence similarity represents the proportion of the number of time points at which the wall thickness deviation at the same time is less than or equal to the wall thickness deviation threshold value.

Images