Reliability detection and model automation training method for sealing structure of wellhead

Abstract

The embodiment of the invention provides a reliability detection and model automatic training method for a sealing structure of a wellhead. The method comprises the steps of obtaining a size data set of the sealing structure; the size data set comprises a plurality of size data; the size data represents information of a physical structure of the sealing structure; performing matching processing on the size data set and a data set in a preset model library to obtain multiple groups of matched data sets; determining an initial model corresponding to the matched data set based on a preset model library; based on a Bayesian optimization algorithm, processing the size data set and each initial model to obtain a reliability detection model; the reliability detection model is used for processing the size data of the sealing structure to obtain a reliability detection result of the sealing structure. The method is used for achieving the effect of accurately and effectively detecting the reliability of the sealing structure.

Description

Technical Field

[0001] This application relates to the field of structural reliability, and in particular to a method for reliability testing and automated model training of a wellhead sealing structure. Background Technology

[0002] In the field of petroleum equipment, the sealing structure, as the core sealing element, is crucial for ensuring the safe and stable operation of equipment under harsh conditions such as high temperature, high pressure, and corrosive media. The reliability of the sealing structure directly determines the sealing performance of the entire equipment. Once it fails, it may not only lead to media leakage and safety accidents, but also cause environmental pollution and huge economic losses.

[0003] Therefore, how to accurately and effectively detect the reliability of the sealing structure is an urgent problem to be solved. Summary of the Invention

[0004] This application provides a method for reliability detection and automated model training of wellhead sealing structures, so as to achieve accurate and effective detection of the reliability of the sealing structure.

[0005] In a first aspect, embodiments of this application provide an automated training method for a reliability detection model of a wellhead sealing structure, including:

[0006] Obtain a dimensional dataset of the sealing structure; the dimensional dataset includes multiple dimensional data; the dimensional data characterizes information about the physical structure of the sealing structure.

[0007] The size dataset is matched with datasets in a preset model library to obtain multiple sets of matching datasets; and an initial model corresponding to the matching dataset is determined based on the preset model library.

[0008] Based on the Bayesian optimization algorithm, the size dataset and each initial model are processed to obtain a reliability detection model; the reliability detection model is used to process the size data of the sealing structure to obtain the reliability detection results of the sealing structure.

[0009] In one possible implementation, the dataset in the preset model library includes meta-features corresponding to the dataset; the size dataset is matched with the dataset in the preset model library to obtain multiple sets of matched datasets, including:

[0010] Based on the size dataset, size meta-features are determined; wherein, the size meta-features characterize the overall statistical properties of the size dataset;

[0011] Based on cosine similarity, the size meta-features of the size dataset are matched with the corresponding meta-features of the dataset in the preset model library to obtain multiple sets of matched datasets.

[0012] In one possible implementation, the initial model includes an initial hyperparameter range corresponding to the initial model; based on a Bayesian optimization algorithm, the size dataset and each initial model are processed to obtain a reliability detection model, including:

[0013] Based on the Bayesian optimization algorithm, the initial hyperparameter range corresponding to the initial model is iteratively searched to determine the hyperparameter information of the initial model; the initial hyperparameter range represents the hyperparameters within a reasonable range of the initial model.

[0014] Based on the initial models configured with the hyperparameter information, the size dataset is processed to obtain a reliability detection model.

[0015] In one possible implementation, based on each initial model configured with the hyperparameter information, the size dataset is processed to obtain a reliability detection model, including:

[0016] Based on the configured hyperparameter information of each initial model, the size dataset is processed to obtain the performance index of each initial model; wherein, the performance index is used to measure the processing ability of the initial model on the size dataset;

[0017] Based on the performance metrics of each initial model, a reliability testing model is determined among the initial models.

[0018] In one possible implementation, prior to obtaining the dimensional dataset of the sealing structure, the following steps are also included:

[0019] Obtain the initial dimensional data of the sealing structure; and obtain the preset reasonable error range corresponding to the initial dimensional data of the sealing structure;

[0020] Based on the initial size data and the preset reasonable error range, multiple sets of initial size ranges are determined; wherein, the multiple sets of initial size ranges represent the reasonable error range of the initial size data of the sealing structure;

[0021] The multiple initial size ranges are sampled to determine the size dataset of the sealing structure.

[0022] In one possible implementation, sampling the multiple sets of initial size ranges to determine the size dataset of the sealing structure includes:

[0023] The initial size range is divided into multiple sets of size intervals with equal probability.

[0024] Based on a random sampling algorithm, random sampling is performed on the multiple sets of size intervals to obtain multiple size parameters;

[0025] The dimensions of the sealing structure are determined by randomly combining multiple dimensional parameters.

[0026] In one possible implementation, the method further includes:

[0027] The size dataset of the sealing structure is preprocessed, wherein the preprocessing includes one or more of the following: outlier removal, missing value completion, and feature normalization.

[0028] Secondly, embodiments of this application provide a method for detecting the reliability of a wellhead sealing structure, including:

[0029] Obtain the dimensional data of the sealing structure;

[0030] The dimensional data of the sealing structure is input into the reliability detection model to obtain the reliability detection result of the sealing structure; the reliability detection model is trained through the first aspect and / or various possible implementations of the first aspect.

[0031] Thirdly, embodiments of this application provide an automated training device for a reliability detection model of a wellhead sealing structure, comprising:

[0032] An acquisition module is used to acquire a dimensional dataset of the sealing structure; the dimensional dataset includes multiple dimensional data; the dimensional data characterizes information about the physical structure of the sealing structure.

[0033] The matching module is used to match the size dataset with the dataset in the preset model library to obtain multiple sets of matching datasets; and to determine the initial model corresponding to the matching dataset based on the preset model library.

[0034] The processing module is used to process the size dataset and each initial model based on the Bayesian optimization algorithm to obtain a reliability detection model; the reliability detection model is used to process the size data of the sealing structure to obtain the reliability detection result of the sealing structure.

[0035] In one possible implementation, the dataset in the preset model library includes meta-features corresponding to the dataset; the matching module includes:

[0036] Based on the size dataset, size meta-features are determined; wherein, the size meta-features characterize the overall statistical properties of the size dataset;

[0037] Based on cosine similarity, the size meta-features of the size dataset are matched with the corresponding meta-features of the dataset in the preset model library to obtain multiple sets of matched datasets.

[0038] In one possible implementation, the initial model includes an initial hyperparameter range corresponding to the initial model; the processing module includes:

[0039] Based on the Bayesian optimization algorithm, the initial hyperparameter range corresponding to the initial model is iteratively searched to determine the hyperparameter information of the initial model; the initial hyperparameter range represents the hyperparameters within a reasonable range of the initial model.

[0040] Based on the initial models configured with the hyperparameter information, the size dataset is processed to obtain a reliability detection model.

[0041] In one possible implementation, based on each initial model configured with the hyperparameter information, the size dataset is processed to obtain a reliability detection model, including:

[0042] Based on the configured hyperparameter information of each initial model, the size dataset is processed to obtain the performance index of each initial model; wherein, the performance index is used to measure the processing ability of the initial model on the size dataset;

[0043] Based on the performance metrics of each initial model, a reliability testing model is determined among the initial models.

[0044] In one possible implementation, prior to obtaining the module, the following is also included:

[0045] Obtain the initial dimensional data of the sealing structure; and obtain the preset reasonable error range corresponding to the initial dimensional data of the sealing structure;

[0046] Based on the initial size data and the preset reasonable error range, multiple sets of initial size ranges are determined; wherein, the multiple sets of initial size ranges represent the reasonable error range of the initial size data of the sealing structure;

[0047] The multiple initial size ranges are sampled to determine the size dataset of the sealing structure.

[0048] In one possible implementation, sampling the multiple sets of initial size ranges to determine the size dataset of the sealing structure includes:

[0049] The initial size range is divided into multiple sets of size intervals with equal probability.

[0050] Based on a random sampling algorithm, random sampling is performed on the multiple sets of size intervals to obtain multiple size parameters;

[0051] The dimensions of the sealing structure are determined by randomly combining multiple dimensional parameters.

[0052] In one possible implementation, the device further includes:

[0053] The size dataset of the sealing structure is preprocessed, wherein the preprocessing includes one or more of the following: outlier removal, missing value completion, and feature normalization.

[0054] Fourthly, embodiments of this application provide a wellhead sealing structure reliability testing device, comprising:

[0055] The acquisition module is used to acquire the dimensional data of the sealing structure;

[0056] The detection module is used to input the dimensional data of the sealing structure into the reliability detection model to obtain the reliability detection result of the sealing structure; the reliability detection model is trained through the first aspect and / or various possible implementations of the first aspect.

[0057] Fifthly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0058] The memory stores computer-executed instructions;

[0059] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0060] In a sixth aspect, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0061] In a seventh aspect, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0062] This application provides a method for reliability testing and automated model training of wellhead sealing structures. First, it constructs a dataset by collecting multi-dimensional dimensional data to ensure that the model input covers key features of the physical structure, providing a foundation for subsequent matching. Second, it utilizes pre-trained models from a pre-defined model library for rapid matching, avoiding training the model from scratch and shortening the development cycle. Finally, it employs a Bayesian optimization algorithm to dynamically adjust the model's hyperparameters, efficiently locating the optimal solution in the search space, enabling the model to accurately capture the nonlinear relationship between dimensional data and reliability. This method solves the problems of traditional testing methods relying on empirical formulas and lacking accuracy, achieving a quantitative assessment of the reliability of the sealing structure throughout its entire lifecycle. The resulting reliability testing model can be directly applied to the production process, providing a scientific basis for the optimized design, fault warning, and maintenance decisions of the sealing structure. Attached Figure Description

[0063] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0064] Figure 1 A flowchart illustrating an automated training method for a reliability detection model of a wellhead sealing structure provided in this application embodiment. Figure 1 ;

[0065] Figure 2 A flowchart illustrating an automated training method for a reliability detection model of a wellhead sealing structure provided in this application embodiment. Figure 2 ;

[0066] Figure 3 This application provides a method for detecting the reliability of the sealing structure of a wellhead.

[0067] Figure 4 A schematic diagram of the dimensions of a BX type flange sealing gasket provided in this application embodiment;

[0068] Figure 5 Histogram of equivalent stress for a BX type flange sealing gasket dimensional dataset provided in this application embodiment;

[0069] Figure 6 A schematic diagram of the actual value and the predicted value provided in an embodiment of this application;

[0070] Figure 7 A schematic diagram illustrating the relationship between predicted reliability and actual reliability, provided for an embodiment of this application;

[0071] Figure 8 A schematic diagram of an automated training device for a reliability detection model of a wellhead sealing structure provided in this application embodiment;

[0072] Figure 9 A schematic diagram of a wellhead sealing structure reliability testing device provided in this application embodiment;

[0073] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0074] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0075] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0076] As the core sealing element of petroleum equipment, the reliability of the sealing structure directly determines the safe and stable operation of the equipment under harsh conditions such as high temperature, high pressure and corrosive media.

[0077] Traditional reliability analysis of sealing structures is based on fixed design parameters and safety factors, which cannot effectively quantify the uncertainties brought about by random factors such as material properties, load fluctuations, and manufacturing tolerances in actual working conditions. This leads to a significant discrepancy between reliability predictions and actual failure risks. To overcome this limitation, current mainstream reliability analysis methods have shifted to a probabilistic framework, the most representative of which are variance reduction techniques based on Monte Carlo simulations and approximate analytical methods based on Taylor series expansions (such as the first-order second-moment method). The Monte Carlo method uses a large number of random samples to statistically determine the failure probability, which has high accuracy, but suffers from huge computational costs and low efficiency in simulating rare events. Approximate analytical methods solve the problem through mathematical simplification, which has high computational efficiency, but often face challenges of convergence difficulties and insufficient accuracy when dealing with highly nonlinear, multi-peak sealing failure problems.

[0078] Machine learning-based reliability analysis of sealing structures often relies on experience to select a specific machine learning model (such as random forest or support vector machine) and pre-fix its hyperparameters. The model is then trained using historical operating data of the sealing structure, and finally, the trained model is used to predict and classify the reliability of new sealing structure dimensions or operating conditions. However, model selection and hyperparameter settings are often directly fixed based on expert experience, lacking a systematic optimization and verification process. This empirical configuration may result in the model not being optimal, introducing two risks: first, the model may underfit, failing to fully learn the complex nonlinear relationships of sealing failure, leading to insufficient prediction accuracy; second, the model may overfit, performing well on training data but showing a significant performance drop when generalizing to new operating conditions.

[0079] Therefore, this application provides an automated training method for a reliability detection model of a wellhead sealing structure. By acquiring the size dataset of the sealing structure, matching the size dataset with the dataset in a preset model library, multiple matching datasets are obtained and corresponding initial models are determined. Based on the Bayesian optimization algorithm, hyperparameters of the size dataset and each initial model are tuned, and finally a reliability detection model is generated, which improves the efficiency and accuracy of sealing structure reliability detection and reduces model training costs.

[0080] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0081] Figure 1 A flowchart illustrating an automated training method for a reliability detection model of a wellhead sealing structure provided in this application embodiment. Figure 1 ,like Figure 1 As shown, the method includes:

[0082] S101. Obtain the dimensional dataset of the sealing structure; wherein, the dimensional dataset includes multiple dimensional data; the dimensional data characterizes the physical structure information of the sealing structure.

[0083] For example, a sealing structure refers to a component used to prevent fluid or gas leakage. In oil equipment, such as valves, pipe connections, or sealing rings in drilling equipment, its function is to ensure the safe operation of the equipment under high pressure and high temperature environments. It is understood that the sealing structure of a wellhead can be a sealing structure for oil or gas wellheads, or for wellheads of other substances; this application does not impose specific limitations here. It is also understood that this application can be applied to the inspection of sealing structures in other equipment; this application does not impose specific limitations here.

[0084] A dimensional dataset is a collection of multiple dimensional data points. Dimensional data are key parameters used to characterize the physical structure of a sealing structure. These parameters cover various dimensions of the sealing structure, such as geometric dimensions like length, width, height, diameter, and thickness, as well as specific dimensions related to assembly and function, such as the inclination angle of the sealing surface, and the depth and width of the groove. This dimensional data is crucial for accurately describing the shape and size of the sealing structure and directly affects sealing performance. For example, if the inner diameter of a sealing ring is too large, it may result in excessive clearance with mating components, thus failing to effectively prevent fluid leakage; while an unsuitable outer diameter may affect its fixation and sealing effect at the installation location.

[0085] For example, obtain a dataset of the dimensions of the sealing structure.

[0086] S102. Match the size dataset with the dataset in the preset model library to obtain multiple matching datasets; and determine the initial model corresponding to the matching dataset based on the preset model library.

[0087] For example, the acquired raw size dataset is cleaned and standardized to eliminate differences in units and formats, facilitating comparison with a model library. For instance, all size units are standardized to millimeters, and the data is normalized while extracting key feature vectors, such as the mean, variance, or principal components of the main dimensions.

[0088] The system iterates through each standard dataset in the pre-defined model library, using a specific matching algorithm to compare the preprocessed measured features with the model features in the library. Commonly used algorithms include Euclidean distance calculation, cosine similarity analysis, or more advanced machine learning-based classifiers (such as the K-nearest neighbors algorithm). The algorithm calculates a similarity score for each pair (measured data and model in the library).

[0089] The system will filter out all matching pairs based on a preset similarity threshold (e.g., a score higher than 0.85), forming multiple matching datasets. Each matching dataset contains the actual test data, the corresponding standard model data, and their similarity index.

[0090] The system will select the standard model corresponding to the set with the highest similarity from multiple matching datasets and determine it as the initial model.

[0091] S103. Based on the Bayesian optimization algorithm, the size dataset and each initial model are processed to obtain the reliability detection model; the reliability detection model is used to process the size data of the sealing structure to obtain the reliability detection results of the sealing structure.

[0092] For example, Bayesian optimization is a sequential design strategy for globally optimizing a black-box function. It approximates the objective function by constructing a probabilistic surrogate model (typically a Gaussian process) and uses an acquisition function (such as the desired improvement in EI) to determine the next most promising evaluation point, thereby finding the global optimum with as few evaluations as possible. Its core idea is to comprehensively utilize existing prior knowledge and newly acquired data to update the model, belonging to a type of sequential optimization method.

[0093] For example, the system defines a loss function to be minimized or a utility function to be maximized. The input to this function is a set of hyperparameters (denoted as x) of the initial model to be optimized, and the output is a metric measuring the model's predictive performance under that set of hyperparameters. For instance, the objective function could be: using a model with hyperparameters x, predict the reliability of all samples in a training-sized dataset, compare the predictions with the known true reliability of these samples, and calculate the negative of the error rate.

[0094] The system specifies the range of values ​​for each hyperparameter that needs to be optimized. For example, if the initial model is a logistic regression classifier, then the hyperparameter that needs to be optimized might be the regularization strength C, whose space can be set to [0.01, 100] (log scale).

[0095] Within the hyperparameter space, a small number (e.g., 5-10) of initial hyperparameter combinations are selected through random sampling or Latin hypercube sampling, and their corresponding objective function values ​​are calculated to form the initial dataset for Bayesian optimization.

[0096] Using all the currently evaluated (hyperparameter, target value) data points, construct a Gaussian process model. This model not only predicts the target value at any hyperparameter point x, but also provides the uncertainty of that prediction.

[0097] Based on the current Gaussian process model, a data acquisition function (such as the desired improvement in EI or upper confidence limit UCB) is calculated across the entire hyperparameter space. The data acquisition function automatically balances utilization (searching in regions where the model predicts the target value well) and exploration (searching in regions where the model has high uncertainty).

[0098] Select the hyperparameter point x_new that maximizes the acquisition function value as the next evaluation point. Run the computationally intensive objective function to obtain the true objective value f(x_new) at that point.

[0099] Add the new evaluation result (x_new, f(x_new)) to the historical dataset.

[0100] This cyclical process will continue until a preset iteration limit (e.g., 50 times) is reached or the improvement of the objective function falls below a certain threshold multiple times in a row.

[0101] After the loop ends, select the set of hyperparameters that optimizes the objective function from all evaluated combinations. Configure the initial model using this optimal set of hyperparameters; this calibrated new model is the final reliability testing model.

[0102] This application provides an automated training method for a reliability detection model of a wellhead sealing structure. First, it constructs a dataset by collecting multi-dimensional dimensional data to ensure the model input covers key features of the physical structure, providing a foundation for subsequent matching. Second, it utilizes pre-trained models from a pre-defined model library for rapid matching, avoiding training the model from scratch and shortening the development cycle. Finally, it employs a Bayesian optimization algorithm to dynamically adjust the model's hyperparameters, efficiently locating the optimal solution in the search space, enabling the model to accurately capture the nonlinear relationship between dimensional data and reliability. This method solves the problems of traditional detection methods relying on empirical formulas and lacking accuracy, achieving a quantitative assessment of the reliability of the sealing structure throughout its entire lifecycle. The resulting reliability detection model can be directly applied to the production process, providing a scientific basis for the optimized design, fault warning, and maintenance decisions of the sealing structure.

[0103] Figure 2 A flowchart illustrating an automated training method for a reliability detection model of a wellhead sealing structure provided in this application embodiment. Figure 2 ,like Figure 2 As shown, in this embodiment... Figure 1 Based on the examples, an automated training method for a reliability detection model of a wellhead sealing structure is described in detail. This method includes:

[0104] S201. Obtain the initial dimension data of the sealing structure; and obtain the preset reasonable error range corresponding to the initial dimension data of the sealing structure; based on the initial dimension data and the preset reasonable error range, determine multiple initial dimension ranges; wherein, the multiple initial dimension ranges represent the reasonable error range of the initial dimension data of the sealing structure; perform sampling processing on the multiple initial dimension ranges to determine the dimension dataset of the sealing structure.

[0105] For example, a preset reasonable error range is a range of legal deviations allowed from the initial size (nominal value) of a product, explicitly specified in the product design or manufacturing standards. It is typically expressed as tolerances, such as "50 ± 0.2 mm" or "+0.1 / - 0.05 mm". This range defines the acceptable fluctuation range for a single dimensional parameter during the manufacturing process.

[0106] For example, obtain the initial values ​​of all critical dimensions of the sealing structure and their corresponding preset tolerance ranges from design documents or standard specifications. For instance, for a simple gasket, critical dimensions might include the outer diameter, inner diameter, and thickness; for a complex oil seal, they might also include the lip angle, spring groove depth, etc. Each dimensional parameter to be considered and its tolerances must be explicitly listed.

[0107] Each dimensional parameter and its tolerance range are defined as an independent dimension. If a sealing structure has n critical dimensions, then an n-dimensional hypercube space is constructed. Each point in this space corresponds to a specific combination of dimensions, representing a theoretically possible seal.

[0108] To avoid the uneven distribution problems that may result from simple random sampling, Latin hypercube sampling (LHS) can be used. LHS ensures that each dimension is uniformly stratified, and each stratum in each dimension is sampled only once, thus achieving uniform and unbiased coverage of the multidimensional parameter space with a relatively small number of samples. In implementation, the total number of samples to be generated is determined first.

[0109] The values ​​for each dimension (size) obtained from the sampling are combined to form a complete virtual sample. For example, the first sampling yields an inner diameter of 50.15 mm and a cross-sectional diameter of 4.92 mm, which constitutes the first sample in the dataset; the second sampling yields another combination, and so on. Finally, all these samples are organized into a row list table, and metadata such as sample IDs may be added, and then exported as a standard format file for further analysis.

[0110] In one possible implementation, a rectangular rubber sealing ring for an oil valve is taken as an example. Its initial dimensional data are: groove width (nominal value 10mm) and sealing strip height (nominal value 8mm). A preset reasonable error range is derived from the machining drawings: groove width is 10 + 0.2 / - 0.1 mm, and sealing strip height is 8 ± 0.15 mm. This determines two initial dimensional ranges: groove width range [9.9mm, 10.2mm], and sealing strip height range [7.85mm, 8.15mm].

[0111] Based on LHS technology, the groove width range [9.9mm, 10.2mm] and the sealing strip height range [7.85mm, 8.15mm] of rectangular rubber sealing rings were sampled to obtain a size dataset.

[0112] In one example, the initial size range is divided into multiple size interval sets with equal probability; based on a random sampling algorithm, random sampling is performed on the multiple size interval sets to obtain multiple size parameters; and the multiple size parameters are randomly combined to determine the size dataset of the sealing structure.

[0113] For example, this step includes interval partitioning, stratified sampling, and full factorial combination.

[0114] During the interval partitioning phase, the initial range of each key dimension needs to be independently partitioned with equal probability. The number of intervals to be partitioned (denoted as N) is determined; this number determines the size of the final dataset and the granularity of the coverage. For example, if you want to generate 10 sample points for each dimension, then you divide its initial range evenly into 10 intervals. This process creates a size interval set containing N intervals for each size parameter.

[0115] In the stratified sampling phase, a random sampling algorithm is used to operate within the interval set of each dimensional parameter. Specifically, it iterates through each sub-interval under that parameter and independently and randomly selects a value within each sub-interval. For example, 49.87 mm is randomly selected from the first interval [49.8, 49.9) of the inner diameter, 49.92 mm is randomly selected from the second interval [49.9, 50.0), and so on, until a value is selected for all N intervals. This operation generates a sample set containing N specific values ​​for each dimensional parameter.

[0116] Finally, in the random combination stage, all sample sets of dimensional parameters are factored together. That is, each sampled value of the first parameter (e.g., inner diameter) is paired with each sampled value of the second parameter (e.g., thickness), and then with each sampled value of all subsequent parameters. If each parameter has N sampled values, and there are M parameters in total, then N×M complete dimensional combinations will be generated. Each combination is a row of data in the dimensional dataset, representing a virtual sealing structure with a specific combination of dimensional parameters within a reasonable range.

[0117] In one possible implementation, consider an O-ring with two key dimensions: inner diameter (range [49.8, 50.2] mm) and cross-sectional diameter (range [4.9, 5.1] mm). The goal is to generate a size dataset containing nine complete samples.

[0118] Divide the range of each size evenly into 3 intervals.

[0119] The size interval sets are {[49.8,50.0),[50.0,50.2),[50.2]} and {[4.9,5.0),[5.0,5.1),[5.1]}. Then, a value is randomly selected from each size interval, resulting in {49.87,50.13,50.19} and {4.93,5.04,5.09}.

[0120] The two interval sets were randomly combined to obtain 9 samples.

[0121] Sample 1: (49.87, 4.93); Sample 2: (49.87, 5.04); Sample 3: (49.87, 5.09); Sample 4: (50.13, 4.93)... up to Sample 9: (50.19, 5.09).

[0122] These nine sets of data constitute the dimensional dataset for this study, ensuring that each sub-region of the inner diameter and cross-sectional diameter is uniformly represented.

[0123] In one example, the size dataset of the sealed structure is preprocessed, where the preprocessing includes one or more of the following: outlier removal, missing value completion, and feature normalization.

[0124] For example, the size dataset of the sealed structure is preprocessed, wherein the preprocessing includes one or more of the following: outlier removal, missing value completion, and feature normalization.

[0125] Outlier removal involves detecting outliers for each size feature that has already had missing values ​​filled. Outlier removal methods include box plot analysis or the Z-score method.

[0126] Box plot rule: Calculate the quartiles of the data, and consider values ​​less than Q1-1.5IQR or greater than Q3+1.5IQR as outliers.

[0127] Z-score method: Calculates the standard deviation of each data point from the mean. Data points with |Z-score| > 3 are usually considered outliers.

[0128] Remove the entire row of samples corresponding to the identified outliers from the dataset.

[0129] Missing value completion is the process of identifying feature columns that contain missing values.

[0130] If the data distribution is approximately normal and there are no serious outliers, impute with the mean; if there is a skewed distribution, impute with the median is more robust. Optionally, a model can be used to impute missing values, such as using the K-nearest neighbor algorithm to predict missing values ​​based on other complete size features.

[0131] Fill the determined fill value into the corresponding missing position.

[0132] Feature normalization is a process that scales all numerical feature columns in a cleaned dataset. Specifically, it scales the data to the [0,1] range. The scaled values ​​are then updated in the dataset.

[0133] S202. The datasets in the preset model library include the meta-features corresponding to the datasets; based on the size dataset, determine the size meta-features; wherein, the size meta-features represent the overall statistical characteristics of the size dataset; based on cosine similarity, match the size meta-features of the size dataset with the meta-features corresponding to the datasets in the preset model library to obtain multiple sets of matching datasets.

[0134] And based on the preset model library, determine the initial model corresponding to the matching dataset.

[0135] For example, statistical analysis is performed on the current size dataset to be analyzed to calculate a set of predefined meta-features, thereby constructing a size meta-feature vector. These features typically include:

[0136] Central tendency: mean and median of each size dimension.

[0137] Dispersion: Standard deviation and variance of each dimension.

[0138] Distribution shape: skewness (measures asymmetry) and kurtosis (measures sharpness) in each dimension.

[0139] Relationship metric: Pearson correlation coefficient between different dimensions.

[0140] For each dataset in the library, its pre-computed and stored meta-feature vector is read. Then, the cosine similarity between the meta-feature vector of the current-sized dataset and the meta-feature vector of each dataset in the library is calculated one by one.

[0141] The system receives all calculated cosine similarity scores and sorts them from highest to lowest. Based on a preset strategy (e.g., setting a similarity threshold of 0.85, or directly selecting the top K most similar datasets, such as Top 3), the system filters out pre-defined model datasets that meet the criteria. Then, the current-size dataset is paired with each of the filtered pre-defined model datasets, forming multiple sets of matched datasets. Each set of matches records the corresponding cosine similarity score, providing a basis for decision-making.

[0142] From the generated multiple matching datasets, the one with the highest similarity score is selected. The standard model from the pre-defined model library corresponding to this matching dataset is determined as the initial model required for this task. This model will serve as the basis for subsequent in-depth analysis and optimization.

[0143] In one possible implementation, assume the current dataset has 200 samples, containing two dimensions: sealing surface width and rubber hardness. Calculate its dimensional feature vector A, which may include: mean width (25.1 mm), standard deviation of width (0.12 mm), width skewness (0.15), mean hardness (75 Shore A), standard deviation of hardness (2.5), hardness skewness (-0.3), and correlation coefficient between width and hardness (0.05).

[0144] The preset model library contains three model datasets: M1 (standard API model), M2 (high pressure enhancement model), and M3 (low temperature special model). Their corresponding meta-feature vectors are B, C, and D, respectively.

[0145] Calculate cosine similarity and matching:

[0146] cosine_similarity(A, B) = 0.96

[0147] cosine_similarity(A, C) = 0.78

[0148] cosine_similarity(A, D) = 0.65

[0149] Setting the threshold to 0.8 yields a matching dataset: (current dataset, M1 dataset), with a similarity of 0.96.

[0150] Determine the initial model: The model corresponding to the matching dataset (current dataset, M1 dataset) is M1 (standard API model). Therefore, the initial model is determined to be M1.

[0151] S203. The initial model includes the initial hyperparameter range corresponding to the initial model; based on the Bayesian optimization algorithm, the initial hyperparameter range corresponding to the initial model is iteratively searched to determine the hyperparameter information of the initial model; the initial hyperparameter range represents the hyperparameters within a reasonable range of the initial model; based on each initial model with configured hyperparameter information, the size dataset is processed to obtain the reliability detection model.

[0152] For example, first, a function needs to be defined to evaluate the quality of a set of hyperparameters. The input to this function is a specific combination of hyperparameters, and the output is a performance metric (the loss to be minimized or the utility to be maximized). For instance, the function might internally train an initial model using this set of hyperparameters and then evaluate its accuracy or F1-score in predicting seal failures on a reserved validation set.

[0153] Determine each hyperparameter that needs to be optimized and its initial range; for example, search the learning rate on a logarithmic scale between [0.0001, 0.1] and the depth of the tree model on an integer scale between [3, 15].

[0154] A small number (e.g., 5-10) of initial hyperparameter combinations are selected through random sampling or Latin hypercube sampling within the hyperparameters and their initial range, and their corresponding objective function values ​​are calculated as the initial training data for the Bayesian optimization surrogate model.

[0155] Using all the currently evaluated (hyperparameters, performance scores) data points, a Gaussian process model is constructed. This model provides an estimate of the probability distribution of the objective function over the entire hyperparameter space.

[0156] Based on the current Gaussian process model, the value of a sampling function (such as the desired improvement in EI) is calculated over the entire hyperparameter space. The EI function quantifies the expected performance improvement relative to the current best value when evaluated at a certain point.

[0157] Select the hyperparameter combination that maximizes the acquisition function value as the next evaluation point. Run the computationally intensive objective function (i.e., train and evaluate the model using this set of hyperparameters) and obtain its true performance score.

[0158] Add the new evaluation results to the historical dataset.

[0159] This cycle will continue until a preset iteration limit (e.g., 50 times) is reached or the performance improvement falls below a certain threshold multiple times in a row.

[0160] After the loop ends, select the set of hyperparameters that optimizes the objective function value from all evaluated combinations.

[0161] The initial model is retrained on the entire training set (optionally, the initial size dataset is divided into training and validation sets) using this set of optimal hyperparameters. This calibrated new model is the final reliability detection model.

[0162] In one example, based on the configuration hyperparameter information of each initial model, the size dataset is processed to obtain the performance index of each initial model; the performance index is used to measure the processing ability of the initial model on the size dataset; based on the performance index of each initial model, the reliability detection model is determined among the initial models.

[0163] For example, based on the Bayesian optimization algorithm, the range of initial hyperparameters corresponding to the initial model is iteratively searched to determine the hyperparameter information of the initial model.

[0164] The hyperparameter information is used to configure the initial model, resulting in the configured initial model.

[0165] The configured initial model performs reliability predictions on all sealed structure size samples in the test set.

[0166] The model's predictions are compared one by one with the true label of each sample in the test set (i.e., the true reliability status of the sealing structure corresponding to the sample, which is kept secret from the model during the training phase).

[0167] Based on the comparison results, a set of predefined performance metrics is calculated for each model. The technological advantage lies in the ability to evaluate the model from different perspectives by calculating multiple metrics such as precision, recall, and F1-Score. For example, in a high-value equipment sealing inspection scenario, recall might be more important because the risk of a false negative (an unreliable seal being mistakenly identified as reliable) is far greater than a false positive (a reliable seal being mistakenly identified as unreliable). This multi-metric evaluation system ensures that model selection closely aligns with actual business needs.

[0168] The system collects performance metric reports from all candidate models and performs horizontal comparisons. The comparison strategy could be:

[0169] Single metric dominance: If a certain metric has veto power (such as recall rate must be higher than 99.9%), then first screen out the qualified models, and then select the best among them based on other metrics.

[0170] Overall score: Assign weights to different indicators, calculate the weighted total score for each model, and select the model with the highest total score.

[0171] Business Priority: Sort and select based on the most critical business metrics (such as F1-Score).

[0172] Ultimately, the selected model was officially adopted as the reliability testing model, completing the entire process.

[0173] This application provides an automated training method for a reliability detection model of a wellhead sealing structure. The method acquires initial dimensional data of the sealing structure and its corresponding preset reasonable error range. Based on this, multiple initial dimensional ranges are determined to characterize the reasonable error of the initial dimensional data. These multiple initial dimensional ranges are then sampled to generate a dimensional dataset of the sealing structure, ensuring that the training data covers dimensional variations in actual manufacturing and enhancing the representativeness and comprehensiveness of the data. Subsequently, dimensional meta-features are extracted from the dimensional dataset. These features describe the overall statistical characteristics of the dataset. Based on cosine similarity, the dimensional meta-features are matched with dataset meta-features in a preset model library to obtain multiple matching datasets. This intelligently selects model data that best matches the characteristics of the current sealing structure, improving the accuracy and efficiency of model selection. Simultaneously, the initial model includes a corresponding initial hyperparameter range. A Bayesian optimization algorithm iteratively searches the initial hyperparameter range to determine the optimal hyperparameter information, thereby efficiently optimizing the model's hyperparameters and avoiding the subjectivity and time-consuming problems of manual parameter tuning. Finally, based on the configured hyperparameter information of each initial model, the dimensional dataset is processed to train a reliability detection model. This process, through the synergistic effect of data generation, model matching, and hyperparameter optimization, enables the model to more accurately capture performance variations of the sealing structure within the error range, improving generalization ability and robustness. Thus, it achieves the beneficial effect of improving the accuracy and efficiency of sealing structure reliability testing, effectively solving the problems of low detection reliability and high false positive rate caused by insufficient data, unsuitable models, or improper hyperparameter settings in traditional methods.

[0174] Figure 3 This application provides a method for testing the reliability of a sealing structure, such as... Figure 3 The method includes:

[0175] S301. Obtain the dimensional data of the sealing structure.

[0176] For example, the dimensional data of the sealing structure is obtained.

[0177] S302. Input the dimensional data of the sealing structure into the reliability testing model to obtain the reliability testing results of the sealing structure.

[0178] The reliability detection model is trained using the automated training method described in the above steps.

[0179] For example, the dimensional data of the sealing structure is input into the reliability testing model to obtain the reliability testing results of the sealing structure.

[0180] This application provides a method for reliability testing of a sealing structure. By acquiring the dimensional data of the sealing structure and inputting this data into a reliability testing model, the reliability testing result of the sealing structure is obtained. This method utilizes the reliability testing model trained in this embodiment of the invention, enabling accurate and rapid detection of the reliability of the sealing structure.

[0181] This application also provides a reliability testing method for BX type flange sealing gaskets.

[0182] Step 1: Obtain the initial dimension data of the BX type flange sealing gasket.

[0183] Figure 4 This application provides a dimensional schematic diagram of a BX type flange sealing gasket as an embodiment; for example... Figure 4 First, establish a fixed XY coordinate system centered at the origin, and set an adjustable vertical axis Y1, whose position is controlled by parameter c1, as an auxiliary reference.

[0184] Next, based on the basic structural features of the BX type washer, draw horizontal lines, vertical lines, and bevel contour lines at specific angles that are parallel to each other on the coordinate axes.

[0185] Then, strict constraints are imposed on the geometric elements: on the one hand, the included angles between each line segment are limited to meet the design requirements, and on the other hand, the key feature points are ensured to be symmetrically distributed about the X-axis and Y1-axis.

[0186] Finally, dimensional control is achieved through parametric driving. The three parameters c2, c3, and c4 are used to precisely constrain the key dimensions of the washer, such as the inner diameter, outer diameter, and chamfer, respectively, so as to realize the parametric linkage adjustment of the entire model.

[0187] Step 2: Establish an automated finite element analysis model with extractable parametric scripts based on the cross-sectional profile of the BX type flange sealing gasket.

[0188] Input the basic parameters required for simulation, including c1, c2, c3, and c4, to provide data support for subsequent modeling.

[0189] A geometric model of the BX type flange sealing gasket is established based on the input parameters, and material properties are assigned. The material is 316L, and a mesh is generated using CAX4R as the mesh element type. An automated script is also invoked.

[0190] Apply boundary constraints to the established model, ensuring the right bevel of the flange gasket is flush with the flange groove, and apply a fully fixed constraint to the reference point of the lower flange groove. Set appropriate loads according to the actual working conditions, applying the load conditions in two steps: in analysis step 1, apply a preload displacement load to the upper flange groove; in analysis step 2, apply pressure loads to the upper, left, and lower flange groove areas. Submit the calculation job to start the simulation analysis process.

[0191] Extract key data such as maximum equivalent stress and maximum contact stress from the calculation results, and output a complete simulation analysis report.

[0192] Step 3: Calculate the true reliability by statistically analyzing the equivalent stress of the sample through Latin hypercube sampling.

[0193] The optimal Latin hypercube was used to sample the four design parameters to form a design form, and the performance response of the corresponding size BX type flange gasket was extracted according to the design form.

[0194] Referring to national tolerance standards, a tolerance of ±0.15 mm is adopted for parameters with dimensions greater than 100 mm, and a tolerance of ±0.1 mm is adopted for parameters less than 100 mm. Under the condition of tolerance, the structural design space of the flange sealing gasket is an initial size range composed of four dimensional parameters within their respective constraints. For example, the value range of c1 is [100 mm, 110 mm], the value range of c2 is [10 mm, 20 mm], the value range of c3 is [7 mm, 12 mm], and the value range of c4 is [5 mm, 10 mm]; a total of 10,000 sets of design parameter combinations are generated.

[0195] Figure 5 A histogram of equivalent stress for a BX type flange sealing gasket dimensional dataset provided in this application embodiment; as shown Figure 4 The 10,000 sets of dimensional data for BX type flange sealing gaskets correspond to 10,000 sets of maximum equivalent stresses. The proportion of samples with maximum equivalent stresses lower than the actual tensile strength represents the true reliability.

[0196] Step 4: Determine the reliability testing model.

[0197] Set the distribution of the dataset to The distribution of a single dataset can be obtained by sampling from it. :

[0198] in, As input to the dataset, For the output of the dataset, This represents the distribution of the sampled dataset.

[0199] Generate a set of hyperparameters Parameterization can automatically generate from Predictive training set model .

[0200] Because the dataset requires a set of independent observations Let's observe and then approximate the generalization error using sample data:

[0201]

[0202] During model training, the obtained dataset Divide the data into two disjoint parts and represent them as the training set. and test set Searching for the optimal set model hyperparameters At that time, only the training set Access was made, ultimately in the test set. Evaluate performance and search for the optimal ensemble model. The process is represented as:

[0203]

[0204] The generalization error GE is calculated using cross-validation.

[0205]

[0206] Considering search time and computational resource limitations, it can be represented as:

[0207]

[0208] The final machine learning system It should be in the entire distribution Good performance is indicated as:

[0209]

[0210] With a limited training dataset It can be approximated as:

[0211]

[0212] Optionally, 500, 1000, 2000 and 10000 sub-sample datasets are extracted from the 10000 sample datasets obtained by Latin hypercube sampling, respectively, and divided into training and test sets in a 7:3 ratio for training and optimization of the reliability detection model.

[0213] Figure 6 This is a schematic diagram illustrating the relationship between the actual value and the predicted value, provided as an embodiment of this application; Figure 6As shown in the figure, the blue dots represent the predicted equivalent stress, and the red line segments represent the actual equivalent stress.

[0214] Optionally, the final reliability testing model can be determined by comprehensively comparing factors such as the number of hyperparameters, the hyperparameter adjustment method, the computation time, and the model quality, as shown in Table 1.

[0215] Table 1 Parameters of the Reliability Testing Model

[0216]

[0217] Optionally, the input parameter for material tensile strength can be set from 350MPa to 600MPa, with an interval of 1MPa. The proportion of samples with the maximum equivalent stress less than the input tensile strength is the percentage of the subsample data, which is the prediction reliability.

[0218] Figure 7 A schematic diagram illustrating the relationship between predicted reliability and actual reliability is provided for embodiments of this application; as shown. Figure 7 The mean square error of the predicted reliability is 0.00338. The predicted minimum tensile strength at a reliability of 99% is 536 MPa, which is 18 MPa less than the actual value, with a relative percentage error of 3.25%.

[0219] Figure 8 A schematic diagram of an automated training device for a reliability detection model of a wellhead sealing structure provided in this application embodiment is shown below. Figure 8 As shown, the automated training device 80 for a reliability testing model of a wellhead sealing structure provided in this embodiment includes:

[0220] The acquisition module 801 is used to acquire the dimensional dataset of the sealing structure; the dimensional dataset includes multiple dimensional data; the dimensional data characterizes information about the physical structure of the sealing structure.

[0221] The matching module 802 is used to match the size dataset with the dataset in the preset model library to obtain multiple sets of matching datasets; and to determine the initial model corresponding to the matching dataset based on the preset model library.

[0222] The processing module 803 is used to process the size dataset and each initial model based on the Bayesian optimization algorithm to obtain the reliability detection model; the reliability detection model is used to process the size data of the sealing structure to obtain the reliability detection result of the sealing structure.

[0223] In one possible implementation, the dataset in the preset model library includes meta-features corresponding to the dataset; the matching module 802 includes:

[0224] Based on the size dataset, determine the size meta-features; where the size meta-features characterize the overall statistical properties of the size dataset.

[0225] Based on cosine similarity, the size meta-features of the size dataset are matched with the corresponding meta-features of the dataset in the preset model library to obtain multiple sets of matching datasets.

[0226] In one possible implementation, the initial model includes an initial hyperparameter range corresponding to the initial model; the processing module 803 includes:

[0227] Based on the Bayesian optimization algorithm, the initial hyperparameter range corresponding to the initial model is iteratively searched to determine the hyperparameter information of the initial model; the initial hyperparameter range represents the hyperparameters within a reasonable range of the initial model.

[0228] Based on the configuration hyperparameter information of each initial model, the size dataset is processed to obtain a reliability detection model.

[0229] In one possible implementation, based on each initial model with configured hyperparameter information, the size dataset is processed to obtain a reliability detection model, including:

[0230] Based on the configuration hyperparameter information of each initial model, the size dataset is processed to obtain the performance index of each initial model; the performance index is used to measure the processing ability of the initial model on the size dataset.

[0231] Based on the performance metrics of each initial model, a reliability testing model is determined among the initial models.

[0232] In one possible implementation, prior to obtaining module 801, the following is also included:

[0233] The acquisition submodule 804 is used to acquire the initial dimension data of the sealing structure and to acquire the preset reasonable error range corresponding to the initial dimension data of the sealing structure.

[0234] The determination module 805 is used to determine multiple sets of initial size ranges based on the initial size data and the preset reasonable error range; wherein, the multiple sets of initial size ranges represent the reasonable error range of the initial size data of the sealing structure.

[0235] Multiple initial size ranges are sampled to determine the size dataset of the sealing structure.

[0236] In one possible implementation, sampling is performed on multiple sets of initial size ranges to determine the size dataset of the sealing structure, including:

[0237] The initial size range is divided into multiple sets of size intervals with equal probability.

[0238] Based on a random sampling algorithm, random sampling is performed on multiple sets of size intervals to obtain multiple size parameters.

[0239] The dimensions of the sealing structure are determined by randomly combining multiple dimensional parameters.

[0240] In one possible implementation, the device 80 further includes:

[0241] The size dataset of the sealed structure is preprocessed, including one or more of the following: outlier removal, missing value completion, and feature normalization.

[0242] This embodiment provides an automated training device for a reliability detection model of a wellhead sealing structure. It can execute the method provided in the above-described method embodiment, and its implementation principle and technical effect are similar. This embodiment will not elaborate further here.

[0243] Figure 9 This is a schematic diagram of a wellhead sealing structure reliability testing device provided in an embodiment of this application, as shown below. Figure 9 As shown, the wellhead sealing structure reliability testing device 90 provided in this embodiment includes:

[0244] Module 901 is used to acquire the dimensional data of the sealing structure;

[0245] The detection module 902 is used to input the dimensional data of the sealing structure into the reliability detection model to obtain the reliability detection result of the sealing structure; the reliability detection model is trained through the above method embodiment.

[0246] This embodiment provides a wellhead sealing structure reliability testing device, which can perform the method provided in the above-described method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0247] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 10 As shown, the electronic device 100 provided in this embodiment includes at least one processor 1001 and a memory 1002. Optionally, the device 100 further includes a communication component 1003. The processor 1001, memory 1002, and communication component 1003 are connected via a bus 1004.

[0248] In a specific implementation, at least one processor 1001 executes computer execution instructions stored in memory 1002, causing at least one processor 1001 to perform the above-described method.

[0249] The specific implementation process of processor 1001 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0250] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0251] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0252] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0253] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0254] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0255] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0256] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0257] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0258] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0259] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0260] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0261] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0262] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. An automated training method for a reliability detection model of a wellhead sealing structure, characterized in that, include: Obtain a dimensional dataset of the sealing structure; wherein the dimensional dataset includes multiple dimensional data; the dimensional data characterizes information about the physical structure of the sealing structure; The size dataset is matched with datasets in a preset model library to obtain multiple sets of matching datasets; and an initial model corresponding to the matching dataset is determined based on the preset model library. Based on the Bayesian optimization algorithm, the size dataset and each initial model are processed to obtain a reliability detection model; the reliability detection model is used to process the size data of the sealing structure to obtain the reliability detection results of the sealing structure.

2. The method according to claim 1, characterized in that, The datasets in the preset model library include the meta-features corresponding to the datasets; The size dataset is matched with datasets in a preset model library to obtain multiple sets of matched datasets, including: Based on the size dataset, size meta-features are determined; wherein, the size meta-features characterize the overall statistical properties of the size dataset; Based on cosine similarity, the size meta-features of the size dataset are matched with the corresponding meta-features of the dataset in the preset model library to obtain multiple sets of matched datasets.

3. The method according to claim 1, characterized in that, The initial model includes the range of initial hyperparameters corresponding to the initial model; Based on the Bayesian optimization algorithm, the size dataset and each initial model are processed to obtain a reliability detection model, including: Based on the Bayesian optimization algorithm, the initial hyperparameter range corresponding to the initial model is iteratively searched to determine the hyperparameter information of the initial model; the initial hyperparameter range represents the hyperparameters within a reasonable range of the initial model. Based on the initial models configured with the hyperparameter information, the size dataset is processed to obtain a reliability detection model.

4. The method according to claim 3, characterized in that, Based on the initial models configured with the hyperparameter information, the size dataset is processed to obtain a reliability detection model, including: Based on the configured hyperparameter information of each initial model, the size dataset is processed to obtain the performance index of each initial model; wherein, the performance index is used to measure the processing ability of the initial model on the size dataset; Based on the performance metrics of each initial model, a reliability testing model is determined among the initial models.

5. The method according to claim 1, characterized in that, Before obtaining the dimensional dataset of the sealing structure, the following steps are also included: Obtain the initial dimensional data of the sealing structure; and obtain the preset reasonable error range corresponding to the initial dimensional data of the sealing structure; Based on the initial size data and the preset reasonable error range, multiple sets of initial size ranges are determined; wherein, the multiple sets of initial size ranges represent the reasonable error range of the initial size data of the sealing structure; The multiple initial size ranges are sampled to determine the size dataset of the sealing structure.

6. The method according to claim 5, characterized in that, Sampling is performed on the multiple initial size ranges to determine the size dataset of the sealing structure, including: The initial size range is divided into multiple sets of size intervals with equal probability. Based on a random sampling algorithm, random sampling is performed on the multiple sets of size intervals to obtain multiple size parameters; The dimensions of the sealing structure are determined by randomly combining multiple dimensional parameters.

7. The method according to any one of claims 1-6, characterized in that, The method further includes: The size dataset of the sealing structure is preprocessed, wherein the preprocessing includes one or more of the following: outlier removal, missing value completion, and feature normalization.

8. A method for testing the reliability of a wellhead sealing structure, characterized in that, include: Obtain the dimensional data of the sealing structure; The dimensional data of the sealing structure are input into the reliability testing model to obtain the reliability testing results of the sealing structure; The reliability detection model is trained using the training method described in any one of claims 1-7.

9. An automated training device for a reliability testing model of a wellhead sealing structure, characterized in that, include: The acquisition module is used to obtain the dimensional dataset of the sealing structure; The size dataset includes multiple size data; The dimensional data characterizes information about the physical structure of the sealing structure; The matching module is used to match the size dataset with the dataset in the preset model library to obtain multiple sets of matching datasets; And based on the preset model library, determine the initial model corresponding to the matching dataset; The processing module is used to process the size dataset and each initial model based on the Bayesian optimization algorithm to obtain a reliability detection model; the reliability detection model is used to process the size data of the sealing structure to obtain the reliability detection result of the sealing structure.

10. A device for testing the reliability of a wellhead sealing structure, characterized in that, include: The acquisition module is used to acquire the dimensional data of the sealing structure; The detection module is used to input the dimensional data of the sealing structure into the reliability detection model to obtain the reliability detection result of the sealing structure; the reliability detection model is trained by the training method of any one of claims 1-7.

Images