A method and system for evaluating the data quality of pressure vessels based on big data

Abstract

The present application relates to the technical field of data processing, and discloses a method and system for evaluating the data quality of pressure vessels based on big data. The method includes: generating a mapped data set through physical constraint preprocessing and working condition synchronization processing according to the internal pressure, wall temperature and stress-strain data of the pressure vessel. Calculating physical constraint, working condition stability and data coordination indexes to obtain a feature set. Analyzing the coupling relationship by using a deep neural network driven by physical knowledge to obtain an abnormal type discrimination. Constructing a scoring matrix and dynamically calibrating to obtain an evaluation factor, and finally obtaining the evaluation result of the equipment data quality through multi-scale fuzzy evaluation. The present application realizes the double consistency of data statistical characteristics and physical constraints to improve the reliability and interpretability of the evaluation result. By introducing physical knowledge into the whole process of data quality evaluation, a "data-physics" double-layer evaluation system is established to ensure that the evaluation result not only conforms to the data distribution law but also meets the physical constraint requirements.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method and system for evaluating the data quality of pressure vessels based on big data. Background Art

[0002] During the operation of pressure vessels, data quality assessment is an important link to ensure the safe operation of equipment. Existing technologies mainly adopt data quality assessment methods based on statistical features, and through the analysis of the integrity, consistency, and accuracy of the collected data, the quantitative assessment of data quality is realized. Such methods usually rely on mathematical statistics models, combined with a set evaluation index system, to analyze the statistical features of data, including the distribution characteristics, fluctuation rules, and trend changes of data. At the same time, some evaluation methods also introduce machine learning algorithms, and through establishing the mapping relationship between data features and quality levels, the automatic evaluation of data quality is realized. These methods have been widely applied in industrial practice and provide important support for equipment operation monitoring and fault diagnosis.

[0003] However, the existing data quality assessment methods have obvious deficiencies. These methods overly rely on the statistical characteristics of data and ignore the physical constraint relationships during the operation of pressure vessels, resulting in the lack of physical meaning in the evaluation results. Secondly, the coupling relationship between multi-source sensing data is not fully considered during the evaluation process, making the evaluation results unable to reflect the true operating state of the equipment. Thirdly, the existing methods mainly identify abnormal data based on statistical thresholds, lacking the verification of physical laws and prone to misjudgment. Finally, the interpretability of the evaluation results is insufficient and it is difficult to provide effective guidance for operation and maintenance decisions. Summary of the Invention

[0004] This application provides a method and system for evaluating the data quality of pressure vessels based on big data, which is used to achieve the double consistency of data statistical characteristics and physical constraints, so as to improve the reliability and interpretability of the evaluation results. By introducing physical knowledge into the whole process of data quality assessment and establishing a "data-physics" double-layer evaluation system, it is ensured that the evaluation results not only conform to the data distribution law but also meet the physical constraint requirements.

[0005] In a first aspect, the present application provides a method for evaluating the data quality of pressure vessels based on big data. The method for evaluating the data quality of pressure vessels based on big data includes: obtaining a physical condition mapping data set through physical constraint preprocessing and condition synchronization processing according to the operating condition data collected from the pressure vessel, including the internal pressure data, wall temperature data, and stress-strain data of the vessel; obtaining a physical condition feature set by calculating physical constraint indicators, condition stability indicators, and data coordination indicators according to the physical condition mapping data set; analyzing the physical coupling relationship and change trend between the physical condition mapping data sets through a physical knowledge-driven deep neural network according to the physical condition feature set to obtain a discrimination result of the abnormal condition type; constructing a scoring matrix through physical constraint scoring according to the discrimination result of the abnormal condition type, and dynamically calibrating the scoring matrix to obtain a physical condition evaluation factor; calculating the operating quality grade of the pressure vessel through multi-scale fuzzy evaluation according to the physical condition evaluation factor and the discrimination result of the abnormal condition type to obtain an evaluation result of the equipment data quality.

[0006] In a second aspect, the present application provides a system for evaluating the data quality of pressure vessels based on big data. The system for evaluating the data quality of pressure vessels based on big data includes:

[0007] A synchronization module, configured to obtain a physical condition mapping data set through physical constraint preprocessing and condition synchronization processing according to the operating condition data collected from the pressure vessel, including the internal pressure data, wall temperature data, and stress-strain data of the vessel;

[0008] A calculation module, configured to obtain a physical condition feature set by calculating physical constraint indicators, condition stability indicators, and data coordination indicators according to the physical condition mapping data set;

[0009] An analysis module, configured to analyze the physical coupling relationship and change trend between the physical condition mapping data sets through a physical knowledge-driven deep neural network according to the physical condition feature set to obtain a discrimination result of the abnormal condition type;

[0010] A construction module, configured to construct a scoring matrix through physical constraint scoring according to the discrimination result of the abnormal condition type, and dynamically calibrate the scoring matrix to obtain a physical condition evaluation factor;

[0011] An evaluation module, configured to calculate the operating quality grade of the pressure vessel through multi-scale fuzzy evaluation according to the physical condition evaluation factor and the discrimination result of the abnormal condition type to obtain an evaluation result of the equipment data quality.

[0012] In a third aspect of the present invention, a computer device is provided, comprising: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor invokes the instructions in the memory to cause the computer device to execute the above-mentioned method for evaluating the data quality of pressure vessels based on big data.

[0013] In a fourth aspect of the present invention, a computer-readable storage medium is provided, in which instructions are stored. When it runs on a computer, it causes the computer to execute the above-mentioned method for evaluating the data quality of pressure vessels based on big data.

[0014] In the technical solution provided by the present application, through the physical constraint preprocessing and working condition synchronization processing of the working condition data collected by the pressure vessel, the construction of the physical mapping relationship of multi-source data is realized, effectively avoiding the deviation caused by simply relying on the statistical characteristics of the data; by calculating the physical constraint index, working condition stability index and data coordination index, a complete set of physical working condition characteristics is established, enabling the evaluation process to fully consider the physical law constraints; a deep neural network driven by physical knowledge is used to analyze the physical coupling relationship and change trend between the physical working condition mapping data sets. This network innovatively adopts a dual-stream structure, and respectively processes the time series features and physical constraint features through the time series flow branch structure and the physical flow branch structure, and realizes the dynamic fusion of features through the attention mechanism, significantly improving the accuracy and interpretability of anomaly recognition, and fully reflecting the deep integration of artificial intelligence algorithms and physical knowledge; through the construction of a scoring matrix by physical constraint scoring and dynamic calibration, the real-time update and adaptive adjustment of the evaluation results are realized, improving the timeliness and adaptability of the evaluation; the operation quality grade of the pressure vessel is calculated by multi-scale fuzzy evaluation, and the operation state of the equipment is analyzed on different time scales, enhancing the comprehensiveness and reliability of the evaluation results; at the algorithm level, the dual-stream structure design of the deep neural network makes full use of the advantages of the long short-term memory network in processing time series data and the characteristics of the graph convolutional network in extracting physical constraint relationships, and the introduction of the attention mechanism realizes the intelligent fusion of time series features and physical constraint features. This algorithm architecture innovation effectively solves the problems of insufficient feature extraction and unclear physical constraint expression in traditional methods; in terms of the evaluation index system, through the organic combination of the physical constraint index, working condition stability index and data coordination index, an evaluation system that conforms to physical laws and meets the actual engineering requirements is constructed, making the evaluation results have stronger physical significance and practical guiding value, and being able to accurately identify various abnormal states in the operation process of the pressure vessel, significantly improving the operation safety management level of the pressure vessel. Description of the Drawings

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0016] Figure 1 It is a schematic diagram of an embodiment of the method for evaluating the data quality of pressure vessels based on big data in the embodiments of the present application;

[0017] Figure 2 It is a timing diagram of preprocessing by physical constraints and synchronous processing of working conditions in the embodiments of the present application;

[0018] Figure 3 It is a data processing flowchart of a deep neural network in the embodiments of the present application;

[0019] Figure 4 It is a schematic diagram of an embodiment of the system for evaluating the data quality of pressure vessels based on big data in the embodiments of the present application;

[0020] Figure 5 It is a schematic block diagram of the structure of a computer device in the embodiments of the present invention. Specific Embodiments

[0021] The embodiments of the present application provide a method and system for evaluating the data quality of pressure vessels based on big data. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned accompanying drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than those illustrated or described here. In addition, the term "comprising" or "having" and any deformation thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0022] For ease of understanding, the following describes the specific process of the embodiments of the present application. Please refer to Figure 1 , an embodiment of the method for evaluating the data quality of pressure vessels based on big data in the embodiments of the present application includes:

[0023] Step S101, according to the working condition data collected from the pressure vessel, including the internal pressure data, wall temperature data and stress-strain data of the vessel, through physical constraint preprocessing and working condition synchronization processing, obtain a physical working condition mapping data set;

[0024] Step S102: According to the physical condition mapping data set, calculate the physical constraint index, condition stability index, and data coordination index to obtain the physical condition feature set;

[0025] Step S103: According to the physical condition feature set, analyze the physical coupling relationship and change trend between the physical condition mapping data sets through a physics knowledge-driven deep neural network to obtain the discriminant result of the abnormal condition type;

[0026] Step S104: According to the discriminant result of the abnormal condition type, construct a scoring matrix through physical constraint scoring and dynamically calibrate the scoring matrix to obtain the physical condition evaluation factor;

[0027] Step S105: According to the physical condition evaluation factor and the discriminant result of the abnormal condition type, calculate the operation quality grade of the pressure vessel through multi-scale fuzzy evaluation to obtain the evaluation result of the equipment data quality.

[0028] It can be understood that the execution subject of this application can be a pressure vessel data quality evaluation system based on big data, or a terminal or a server. Specifically, it is not limited here. In this embodiment of the application, the server is taken as the execution subject for illustration.

[0029] Specifically, a sensor network is arranged at the key positions of the pressure vessel to collect the internal pressure data, wall temperature data, and stress and strain data of the container. For the internal pressure data of the container, a pressure sensor is used to collect data at the top, middle, and bottom of the container, with a sampling frequency of 10 Hz, and the pressure change is recorded in real time. For the wall temperature data, it is collected through a temperature sensor array at different height positions on the container wall, with a sampling frequency of 1 Hz, to monitor the temperature distribution. The stress and strain data are collected through strain gauges in the stress concentration area of the container, with a sampling frequency of 100 Hz, to reflect the local deformation degree. The collected original data undergoes physical constraint preprocessing and condition synchronization processing. The physical constraint preprocessing includes two links: data cleaning and physical rule verification. During the data cleaning process, median filtering is performed on the internal pressure data of the container to remove mutation points, smoothing processing is performed on the wall temperature data to eliminate noise, and normalization processing is performed on the stress and strain data to unify the dimensions. During the physical rule verification process, based on the pressure vessel design specifications, it is verified whether the data meets the physical constraint requirements. The condition synchronization processing is to align the data collected by different sensors according to a unified time reference to construct a physical condition mapping data set.

[0030] Based on the physical condition mapping dataset, by calculating physical constraint indicators, condition stability indicators, and data coordination indicators, a physical condition feature set is formed. Physical constraint indicators reflect the degree of compliance of the data with physical laws, including the correlation characteristics between pressure and temperature, and the coupling characteristics between temperature and stress. Condition stability indicators describe the fluctuation degree of the operating state, including pressure volatility, temperature change rate, and strain growth rate. Data coordination indicators reflect the consistency between different types of data, including temperature-pressure coordination degree and thermal stress coordination degree. Based on the physical condition feature set, a deep neural network driven by physical knowledge is used to analyze the physical coupling relationship and change trend between physical condition mapping datasets. The network adopts a two-stream structure. The time series stream branch uses a long short-term memory network to process time series features, and the physical stream branch uses a graph convolutional network to process physical constraint features, and feature fusion is achieved through an attention mechanism. The network inputs include time series data of pressure, temperature, stress and strain, and physical constraint relationships, and outputs the discriminant results of condition anomaly types.

[0031] A scoring matrix is constructed and dynamically calibrated according to the discriminant results of condition anomaly types. The discriminant results are divided into three categories: pressure anomaly, temperature anomaly, and strain anomaly. Pressure anomalies include overpressure anomalies and underpressure anomalies, temperature anomalies include over-temperature anomalies and temperature fluctuation anomalies, and strain anomalies include strain overlimit anomalies and strain mutation anomalies. Different weights are assigned to different types of anomalies. The weight range for pressure anomalies is 0.4 - 0.5, the weight range for temperature anomalies is 0.3 - 0.4, and the weight range for strain anomalies is 0.2 - 0.3. The degree of anomaly is quantified into a risk level through score mapping to form an initial scoring matrix. Considering the duration and degree of anomaly, the scoring matrix is dynamically calibrated to obtain a physical condition evaluation factor.

[0032] Based on the physical condition evaluation factor and the discriminant results of condition anomaly types, a multi-scale fuzzy evaluation method is used to calculate the operating quality level of the pressure vessel. The evaluation process is divided into three time scales: second level, minute level, and hour level. At the second level scale, the frequency of anomaly occurrence is counted. Less than 5 times / second is divided into the low-frequency value range, 5 - 10 times / second is divided into the medium-frequency value range, and more than 10 times / second is divided into the high-frequency value range. At the minute level scale, the persistence of the anomaly is analyzed. A duration less than 1 minute is divided into the instantaneous anomaly domain, 1 - 5 minutes is divided into the short-term anomaly domain, and more than 5 minutes is divided into the continuous anomaly domain. At the hour level scale, the periodicity of the anomaly is studied. A period less than 2 hours is divided into the high-frequency anomaly domain, 2 - 6 hours is divided into the medium-frequency anomaly domain, and more than 6 hours is divided into the low-frequency anomaly domain.

[0033] The data quality assessment of the ethylene storage tank of a certain chemical enterprise is used as an illustration. The volume of this storage tank is 1000 cubic meters, the design pressure is 0.6 MPa, and the operating temperature range is from -30°C to 40°C. One-week operation data is collected through a sensor network, including pressure data, temperature data, and stress-strain data. During the data preprocessing stage, it is found that there are 3 mutations in the pressure data, the temperature data contains random noise, and the stress-strain data is missing in some time periods. Physical constraint verification shows that some data does not meet the requirements of the temperature-pressure relationship. The effective data after cleaning and verification constitutes the physical condition mapping data set. The calculated condition characteristics show that the pressure stability index is good, but the temperature fluctuates greatly, and there is a coordination deviation with the stress change. Deep neural network analysis discriminates multiple abnormal temperature fluctuations and abnormal strain mutations. The physical condition assessment factor after scoring and calibration is 0.82. The multi-scale fuzzy evaluation results show that the operation quality level of this storage tank is in the normal operation range, but there are potential hazards in terms of temperature control and structural monitoring.

[0034] In the embodiments of the present application, through physical constraint preprocessing and working condition synchronization processing of the working condition data collected from the pressure vessel, the construction of the physical mapping relationship of multi-source data is realized, effectively avoiding the deviation caused by simply relying on the statistical characteristics of the data; by calculating physical constraint indicators, working condition stability indicators and data coordination indicators, a complete set of physical working condition characteristics is established, enabling the evaluation process to fully consider the constraints of physical laws; a deep neural network driven by physical knowledge is used to analyze the physical coupling relationship and change trend between physical working condition mapping data sets. This network innovatively adopts a two-stream structure, processes temporal features and physical constraint features through a temporal stream branch structure and a physical stream branch structure respectively, and realizes the dynamic fusion of features through an attention mechanism, significantly improving the accuracy and interpretability of anomaly recognition, and fully reflecting the deep integration of artificial intelligence algorithms and physical knowledge; through the construction of a scoring matrix by physical constraint scoring and dynamic calibration, the real-time update and adaptive adjustment of the evaluation results are realized, improving the timeliness and adaptability of the evaluation; a multi-scale fuzzy evaluation is used to calculate the operation quality grade of the pressure vessel, and the operation state of the equipment is analyzed on different time scales, enhancing the comprehensiveness and reliability of the evaluation results; at the algorithm level, the two-stream structure design of the deep neural network makes full use of the advantages of the long short-term memory network in processing temporal data and the characteristics of the graph convolutional network in extracting physical constraint relationships. The introduction of the attention mechanism realizes the intelligent fusion of temporal features and physical constraint features. This algorithm architecture innovation effectively solves the problems of insufficient feature extraction and unclear physical constraint expression in traditional methods; in the evaluation index system, through the organic combination of physical constraint indicators, working condition stability indicators and data coordination indicators, an evaluation system that conforms to physical laws and meets the actual engineering needs is constructed, making the evaluation results have stronger physical significance and practical guiding value, and being able to accurately identify various abnormal states in the operation process of the pressure vessel, significantly improving the operation safety management level of the pressure vessel.

[0035] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0036] (1) Convert the fluctuation points in the internal pressure data of the container into a pressure trend sequence, the temperature points in the wall temperature data into a thermal sequence, and the strain points in the stress and strain data into a deformation sequence;

[0037] (2) Perform segmented sampling on the pressure trend sequence, thermal sequence and deformation sequence through a time window to obtain working condition data segments;

[0038] (3) Correlate and pair the pressure trend sequence and the thermal sequence in the working condition data segment to generate a temperature-pressure mapping group;

[0039] (4) Correlate and pair the thermal sequence and the deformation sequence in the working condition data segment to generate a thermal stress mapping group;

[0040] (5) Perform working condition coupling processing on the temperature-pressure mapping group and the thermal stress mapping group to obtain working condition correlation characteristics;

[0041] (6) Perform physical constraint compensation and working condition synchronization calibration according to the working condition correlation characteristics to obtain a physical working condition mapping data set.

[0042] Specifically, as Figure 2 shown, it is a timing schematic diagram of physical constraint preprocessing and working condition synchronization processing in an embodiment of the present application. Among them, fluctuation point conversion processing is performed on the original collected data. For the internal pressure data of the container, the pressure change amplitude between adjacent sampling points is analyzed to identify the fluctuation points, and when the pressure change amplitude exceeds the preset threshold, it is marked as a fluctuation point. These fluctuation points are connected in chronological order to form a pressure trend sequence, which reflects the dynamic change trend of the pressure. Similarly, the temperature measurement points in the wall temperature data are screened and connected to construct a thermal sequence reflecting the temperature change law. After the stress-strain data is extracted for strain points, a deformation sequence describing the deformation state of the container is generated. In order to achieve unified processing of different types of data, time windows are used for segmented sampling. The size of the time window is determined according to the physical characteristics of the pressure vessel and the data sampling frequency, and a typical value is 10 seconds. Within each time window, the pressure trend sequence, the thermal sequence, and the deformation sequence are sampled respectively to obtain working condition data segments with the same time span. This segmented sampling method not only retains the time series characteristics of the data but also facilitates subsequent correlation analysis.

[0043] After obtaining the working condition data segments, establish the correlation relationship between different physical quantities. Correlate and pair the pressure trend sequence and the thermal sequence in the working condition data segments, and based on the principle of the gas state equation, analyze the mutual influence between pressure and temperature to generate a temperature-pressure mapping group. This mapping group reflects the corresponding relationship between pressure changes and temperature changes. Secondly, correlate and pair the thermal sequence with the deformation sequence, and according to the thermal stress theory, study the stress-strain response caused by temperature changes to generate a thermal stress mapping group.

[0044] Performing working condition coupling processing on the temperature-pressure mapping group and the thermal stress mapping group is a key link in data fusion. By analyzing the interaction between the temperature-pressure relationship and the thermal stress relationship, a coupling model among pressure, temperature, and stress-strain is established to obtain working condition correlation characteristics. These characteristics contain the physical state information of the pressure vessel under different working conditions. Perform physical constraint compensation and working condition synchronization calibration based on the working condition correlation characteristics. Physical constraint compensation means correcting the data according to the design parameters and material properties of the pressure vessel to make it meet the requirements of physical laws. Working condition synchronization calibration is to ensure the temporal consistency of different types of data and eliminate the influence of factors such as sampling delay. After these processes, a complete physical working condition mapping data set is formed.

[0045] Taking a liquefied natural gas storage tank as an example. The storage tank is equipped with multiple pressure sensors, temperature sensors and strain sensors, and the sampling frequencies are 10Hz, 1Hz and 100Hz respectively. During the data fluctuation point conversion stage, the pressure data is analyzed, and multiple pressure fluctuation points are found. These fluctuation points are related to the liquid level change and gasification fluctuation of the storage tank. By connecting these fluctuation points, a trend sequence reflecting the pressure change law is formed. The temperature data shows obvious stratification phenomenon, and the data of temperature sensors at different heights form a unique thermal sequence. The stress and strain data reflect the local deformation of the storage tank wall, especially the strain value is larger in the area where the temperature changes violently.

[0046] Using a 10-second time window to segment and sample the three types of sequences, a large number of working condition data segments are obtained. During the correlation pairing process, it is found that the pressure increases with the increase of temperature, but there is a time lag between them. The thermal stress mapping group shows that there is a certain deviation between the thermal strain caused by temperature change and the measured strain, and this deviation is related to the structural characteristics and constraint conditions of the storage tank. Through the working condition coupling process, the quantitative relationship among the pressure increase, temperature decrease and strain increase under the rapid gasification working condition is clarified. Finally, after physical constraint compensation and working condition synchronization calibration, a physical working condition mapping data set reflecting the true operating state of the storage tank is established.

[0047] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0048] (1) Divide the physical working condition mapping data set into working condition time period groups according to the time sequence, calculate the fluctuation values of the data in the working condition time period groups, and obtain physical constraint indexes;

[0049] (2) Calculate the change rate of the internal pressure data of the container in the working condition time period group, and determine the working condition stable interval according to the pressure change trend to obtain the working condition stability index;

[0050] (3) Conduct a correlation analysis on the wall temperature data and stress and strain data in the working condition time period group to generate a thermal stress correlation degree and obtain a data coordination index;

[0051] (4) Combine the physical constraint index, working condition stability index and data coordination index to construct a feature vector;

[0052] (5) Conduct a normalization mapping on the feature vector to generate a standardized feature set of the physical constraint index, working condition stability index and data coordination index;

[0053] (6) Combine and correlate the indexes in the standardized feature set to construct a working condition feature mapping relationship and obtain a physical working condition feature set.

[0054] Specifically, the time series division is performed on the physical condition mapping data set. According to the operating characteristics of the equipment, the data set is divided into multiple condition period groups, and the length of each period is 1 hour. Within each condition period group, the fluctuation value is calculated based on the pressure, temperature, and strain data. The coefficient of variation method is used to calculate the fluctuation value:

[0055]

[0056] Among them, is the fluctuation value, is the data standard deviation, is the data mean. The fluctuation value of each physical quantity is calculated and compared with the preset threshold to generate a physical constraint index.

[0057] For the internal pressure data of the container, the pressure change rate is calculated by the pressure difference between consecutive time points. The calculation formula of the pressure change rate is:

[0058]

[0059] Among them, is the pressure change rate (MPa / s), is the pressure change amount (MPa), is the time interval (s). According to the magnitude of the pressure change rate, the stable condition interval is determined. When the pressure change rate is lower than the set threshold, it is determined as the stable interval, and the duration of the stable interval is statistically counted to obtain the condition stability index. For the correlation analysis of the wall temperature data and the stress-strain data, the thermal stress correlation calculation method is used:

[0060]

[0061] Among them, is the thermal stress correlation, is the temperature value (°C), is the strain value (με), is the average temperature, is the average strain, and n is the number of data points. The correlation reflects the consistency between the temperature change and the strain response, and constitutes the data coordination index.

[0062] The physical constraint index, the condition stability index, and the data coordination index are combined to construct a feature vector:

[0063]

[0064] Among them, is the feature vector, is the physical constraint index, is the condition stability index, is the data coordination index. The feature vector is normalized:

[0065]

[0066] Among them, is the normalized eigenvector, is the minimum value of each index, is the maximum value of each index. The normalization process makes the indexes with different dimensions comparable and forms a standardized feature set.

[0067] Finally, the indexes in the standardized feature set are combined and correlated to construct a mapping relationship of operating condition features. Through the weighted combination method:

[0068]

[0069] Among them, is the mapping value, , , are the weight coefficients of the physical constraint index, the operating condition stability index and the data coordination index respectively, , , are the corresponding normalized index values. According to the distribution characteristics of the mapping value, a physical operating condition feature set is obtained.

[0070] Taking a certain liquefied natural gas storage tank as an example, within a group of operating condition time periods, the standard deviation of the pressure data is 0.02 MPa, the mean value is 0.5 MPa, and the calculated fluctuation value is 0.04. The maximum value of the pressure change rate appears during the unloading process and reaches 0.005 MPa / s. The correlation coefficient between the temperature and strain data is 0.85, indicating good thermal stress response. After normalization and weighted combination of these indexes, a feature set reflecting the operating state of the equipment during this period is formed.

[0071] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0072] (1) The deep neural network has a two-stream structure, including a time-series stream branch structure and a physical stream branch structure. The time-series stream branch structure uses a long short-term memory network to process time-series features, and the physical stream branch structure uses a graph convolutional network to process physical constraint features. The outputs of the two branches are fused through an attention mechanism to obtain a deep neural network driven by physical knowledge;

[0073] (2) Input the physical operating condition feature set into the long short-term memory network to extract time-series change features and obtain a time-series feature vector;

[0074] (3) Input the physical operating condition mapping data set into the graph convolutional network to extract physical constraint relationships and obtain a physical constraint vector;

[0075] (4) Perform weighted fusion on the temporal feature vector and the physical constraint vector through the attention mechanism to obtain a fused feature matrix;

[0076] (5) Calculate the abnormal probability distribution based on the fused feature matrix to generate an abnormal mapping vector;

[0077] (6) Conduct type attribution analysis on the abnormal mapping vector to obtain the discrimination result of the working condition abnormal type.

[0078] Specifically, as Figure 3 shown, it is the data processing flow chart of the deep neural network in the embodiment of this application. The deep neural network is constructed with a two-stream structure, divided into a temporal stream branch and a physical stream branch. The temporal stream branch uses a long short-term memory network (LSTM) to process temporal features. This network has three control units: an input gate, a forget gate, and an output gate, and can capture the long-term dependence relationship of data. The physical stream branch uses a graph convolutional network (GCN) to process physical constraint features, representing the physical parameter relationship of the pressure vessel as a graph structure, where nodes represent different physical quantities and edges represent the constraint relationships between physical quantities. After the physical working condition feature set is input into the LSTM network, the memory unit extracts features from the temporal data. The input gate determines the new information to be retained at the current moment, the forget gate controls the degree of forgetting of historical information, and the output gate determines the output content of the current state. After being processed by multiple layers of LSTM, a temporal feature vector reflecting the time evolution characteristics of parameters such as pressure, temperature, and strain is obtained. At the same time, the physical working condition mapping data set is input into the GCN network, and the graph convolutional layer extracts features from the physical constraint relationships. Each graph convolutional layer aggregates the information of adjacent nodes to realize the feature learning of the constraint relationships between physical quantities, and finally obtains a physical constraint vector.

[0079] The attention mechanism is used to fuse the temporal feature vector and the physical constraint vector. This mechanism calculates the correlation weights between the two types of feature vectors, assigns larger weights to important features, and smaller weights to secondary features. Through weighted summation, a fused feature matrix is obtained, which contains both the temporal change information of the data and reflects the physical constraint relationship. Based on the fused feature matrix, a softmax classifier is used to calculate the probability distribution of different anomaly types and generate an anomaly mapping vector. Finally, threshold judgment and type attribution analysis are performed on the anomaly mapping vector to determine the specific operating condition anomaly type. Taking the propylene storage tank of a petrochemical enterprise as an example to illustrate the application of this method. This storage tank is equipped with sensors such as pressure, temperature, and strain, and the operation data for a continuous week is collected. The temporal flow branch LSTM network extracts temporal features such as pressure fluctuations, temperature gradients, and strain mutations, and the physical flow branch GCN network identifies physical constraint features such as abnormal temperature-pressure relationships and thermal stress response deviations. The attention mechanism focuses on the physical constraint violations at the moment of pressure mutation and weakens the general fluctuations under normal operating conditions. Through anomaly probability calculation and type discrimination, multiple pressure anomaly events caused by improper temperature control are successfully identified. These anomalies show obvious characteristics in both temporal features and physical constraint features, verifying the effectiveness of the dual-stream network structure. In practical applications, each branch of the dual-stream neural network specifically processes different types of features. The temporal flow branch focuses on analyzing the temporal correlation of the data and capturing the temporal patterns of anomaly occurrence. The physical flow branch focuses on the compliance with physical laws and discovers anomalies that violate physical constraints. The two branches are organically combined through the attention mechanism, which not only ensures temporal continuity but also meets physical rationality, thus accurately identifying various abnormal operating conditions during the operation of pressure vessels.

[0080] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0081] (1) Classify the discriminant results of the operating condition anomaly types into three types: pressure anomaly, temperature anomaly, and strain anomaly. The pressure anomaly is divided into overpressure anomaly and underpressure anomaly, the temperature anomaly is divided into overtemperature anomaly and temperature fluctuation anomaly, and the strain anomaly is divided into strain overlimit anomaly and strain mutation anomaly, generating an anomaly feature group;

[0082] (2) Divide the weight intervals of the anomaly feature group according to the safety constraint criteria of pressure vessels. Among them, the weight interval of the pressure anomaly is 0.4 to 0.5, the weight interval of the temperature anomaly is 0.3 to 0.4, and the weight interval of the strain anomaly is 0.2 to 0.3, obtaining a feature weight matrix;

[0083] (3) Perform scoring mapping on the feature weight matrix, map overpressure anomalies and strain overlimit anomalies to a high-risk scoring range, map temperature fluctuation anomalies and strain mutation anomalies to a medium-risk scoring range, map underpressure anomalies to a low-risk scoring range, and construct an initial scoring matrix;

[0084] (4) Dynamically update the scoring values in the initial scoring matrix according to the duration and degree of anomalies in the discrimination result of the working condition anomaly type to obtain a calibrated scoring matrix;

[0085] (5) Weight and combine the scoring values in the calibrated scoring matrix according to the harm degree of the anomaly type to generate a working condition evaluation value;

[0086] (6) Perform normalization transformation on the working condition evaluation value to obtain a physical working condition evaluation factor.

[0087] Specifically, based on physical characteristics, working condition anomalies are divided into three categories: pressure anomalies, temperature anomalies, and strain anomalies. Among them, pressure anomalies are further divided into overpressure anomalies and underpressure anomalies, temperature anomalies are divided into overtemperature anomalies and temperature fluctuation anomalies, and strain anomalies are divided into strain overlimit anomalies and strain mutation anomalies. Through this multi-level anomaly classification system, a complete set of anomaly characteristics is constructed. When determining the danger degree of the anomaly type, a weight distribution method based on the safety constraint criteria of pressure vessels is adopted. As the most direct factor affecting the safety of the container, the weight range of pressure anomalies is set from 0.4 to 0.5; due to its influence on material properties and pressure changes, the weight range of temperature anomalies is 0.3 to 0.4; strain anomalies reflect structural responses, and the weight range is set from 0.2 to 0.3. This hierarchical weight is expressed in matrix form:

[0088]

[0089] Among them and represent the weights of overpressure and underpressure anomalies respectively, and represent the weights of overtemperature and temperature fluctuation anomalies respectively, and represent the weights of strain overlimit and strain mutation anomalies respectively. This matrix structure clearly shows the weight distribution relationship of different types of anomalies.

[0090] During the scoring mapping process, the anomaly types are divided into three intervals according to the risk level. Based on the harmfulness of the anomalies, a risk scoring vector is constructed:

[0091]

[0092] Among them is the comprehensive risk score, is the risk level coefficient (taking values of 0.9, 0.6, and 0.3 for high, medium, and low risks respectively), is the abnormal index value corresponding to the risk level. Taking the overpressure abnormality that occurs during the operation of a certain pressure vessel as an example, when the pressure value exceeds 1.1 times the design pressure, this abnormality is classified into the high-risk range, and its α value is taken as 0.9.

[0093] Considering the dynamic characteristics of the abnormality, the time factor is introduced to calibrate the score. The calibration coefficient is calculated using:

[0094]

[0095] where is the calibration coefficient, is the time impact factor, is the duration of the abnormality (in hours), is the amplitude impact factor, is the degree of abnormality exceeding the limit. For example, a certain storage tank has a temperature fluctuation abnormality for 2 hours, exceeding the normal range by 20%, taking = 0.1, = 0.2, then the corresponding calibration coefficient can be calculated.

[0096] The working condition evaluation value is obtained through the weighted combination of each abnormal type:

[0097]

[0098] where is the working condition evaluation value, is the weight of the abnormal type, is the abnormal score, is the corresponding calibration coefficient. Taking the data of a certain pressure vessel within an operating cycle as an example, when an overpressure abnormality (λ = 0.45) lasts for 1.5 hours and exceeds the limit by 15%, combined with the evaluation results of other abnormal types, the complete working condition evaluation value can be obtained.

[0099] Finally, the physical working condition evaluation factor is obtained through normalization:

[0100]

[0101] where is the normalized evaluation factor, , are the theoretical minimum and maximum values of the evaluation value respectively. Such an evaluation system can accurately reflect the operating status and potential risks of equipment in practical applications, such as the operation monitoring of a certain LNG storage tank.

[0102] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0103] (1)Divide the time scale of the physical condition evaluation factors, construct the second-level evaluation domain, minute-level evaluation domain, and hour-level evaluation domain, and obtain multi-scale evaluation indicators;

[0104] (2)Statistically analyze the abnormal frequency of the discriminant results of the working condition abnormal types in the second-level evaluation domain, analyze the abnormal persistence in the minute-level evaluation domain, and analyze the abnormal periodicity in the hour-level evaluation domain to obtain time-scale characteristic values;

[0105] (3)Establish a fuzzy membership function based on the time-scale characteristic values, divide the operating quality of the pressure vessel into excellent operating interval, normal operating interval, abnormal operating interval, and dangerous operating interval, and obtain a fuzzy evaluation space;

[0106] (4)Map the multi-scale evaluation indicators to the fuzzy evaluation space, calculate the membership degree values of each operating interval, and obtain a fuzzy evaluation vector;

[0107] (5)Perform multi-scale weighted fusion on the fuzzy evaluation vector to generate a quality grade score value;

[0108] (6)Classify and divide the quality grade score value according to the set threshold to obtain the evaluation result of the equipment data quality.

[0109] Specifically, in the final stage of the pressure vessel data quality evaluation, perform multi-scale time analysis on the physical condition evaluation factors. Establish three time-scale evaluation domains: the second-level evaluation domain (1 - 60 seconds), the minute-level evaluation domain (1 - 60 minutes), and the hour-level evaluation domain (1 - 24 hours). For the evaluation indicators of different time scales, use the scale decomposition function:

[0110]

[0111] where is the scale decomposition value, is the time-scale weight coefficient, is the evaluation indicator corresponding to the time scale. For example, in the operation monitoring of a certain LNG storage tank, the second-level data reflects the instantaneous change, the minute-level data reflects the process trend, and the hour-level data shows the long-term characteristics.

[0112] When analyzing the abnormal characteristics at each time scale, different statistical methods are used. The calculation of the second-level abnormal frequency uses:

[0113]

[0114] where is the abnormal frequency (times / second), is the number of abnormal occurrences, is the statistical time period (seconds). Taking the data of a pressure vessel during startup as an example, when 3 pressure fluctuations occur within 10 seconds, the abnormal frequency can be calculated as 0.3 times per second.

[0115] For the analysis of abnormal persistence at the minute level, the duration ratio is introduced:

[0116]

[0117] where is the duration ratio, is the abnormal weight factor, is the duration of a single abnormal event (minutes), is the total observation time (minutes), and m is the number of abnormal events. For example, if a storage tank has two temperature abnormalities within one hour, with durations of 3 minutes and 5 minutes respectively, its persistence index can be calculated.

[0118] For the analysis of abnormal periodicity at the hour level, the periodic characteristic function is adopted:

[0119]

[0120] where is the periodic characteristic value, is the weight of the periodic component, is the period length (hours), and n is the number of periodic components. This analysis method is suitable for discovering regular abnormalities in equipment operation, such as pressure fluctuations caused by the production cycle.

[0121] Based on the time-scale characteristics, a fuzzy membership function is established:

[0122]

[0123] where is the membership degree of quality level, is the evaluation index value, and are the parameters of the membership function. This fuzzy evaluation method divides the equipment operation state into different intervals and realizes the accurate evaluation of operation quality. The fusion of multi-scale evaluation indexes adopts the weighted summation method:

[0124]

[0125] where is the final quality score, is the interval weight coefficient, The evaluation value for each operating interval. For the actual operating data of a pressure vessel, when the proportion of high-quality operating time reaches 85% and the proportion of abnormal intervals is less than 5%, it can be determined that the equipment is in good operating condition. The scoring results are graded by setting different threshold intervals. The classification criteria are: above 90 points is high-quality operation, 80 - 90 points is normal operation, 60 - 80 points is abnormal operation, and below 60 points is dangerous operation. In practical applications, this multi-scale evaluation system can not only reflect the immediate state of the equipment but also predict potential operating risks.

[0126] In a specific embodiment, the process of performing the step of dividing the operating quality of the pressure vessel into high-quality operating intervals, normal operating intervals, abnormal operating intervals, and dangerous operating intervals may specifically include the following steps:

[0127] (1) Statistically analyze the second-level abnormal frequency in the time-scale characteristic values, divide those with an abnormal frequency less than 5 times / second into the low-frequency value range, those with an abnormal frequency between 5 - 10 times / second into the medium-frequency value range, and those with an abnormal frequency greater than 10 times / second into the high-frequency value range to obtain the frequency distribution characteristics;

[0128] (2) Statistically analyze the minute-level abnormal persistence in the time-scale characteristic values, divide those with an abnormal duration less than 1 minute into the instantaneous abnormal domain, those with an abnormal duration between 1 - 5 minutes into the short-term abnormal domain, and those with an abnormal duration greater than 5 minutes into the continuous abnormal domain to obtain the persistence characteristics;

[0129] (3) Statistically analyze the hour-level abnormal periodicity in the time-scale characteristic values, divide those with an abnormal period less than 2 hours into the high-frequency abnormal domain, those with an abnormal period between 2 - 6 hours into the medium-frequency abnormal domain, and those with an abnormal period greater than 6 hours into the low-frequency abnormal domain to obtain the period characteristics;

[0130] (4) Establish a membership function rule based on the frequency distribution characteristics, persistence characteristics, and period characteristics to obtain the interval membership relationship;

[0131] (5) Map the interval membership relationship to the operating quality interval, construct the membership boundaries of the high-quality operating interval, normal operating interval, abnormal operating interval, and dangerous operating interval to obtain the evaluation interval matrix;

[0132] (6) Perform a spatial mapping transformation on the evaluation interval matrix to obtain the fuzzy evaluation space.

[0133] Specifically, in the final link of the pressure vessel data quality assessment, the abnormal characteristics at different time scales are analyzed in detail and graded. For the second-level abnormal frequency, a quantitative analysis is carried out, and the frequency density function is introduced:

[0134]

[0135] wherein is the abnormal frequency density value (times / second), is the abnormal event weight coefficient, is the occurrence times of a single abnormal event, is the observation time window (seconds). Taking the pressure fluctuation of a certain LNG storage tank during the unloading process as an example, when the system detects 45 minor fluctuations within 10 seconds, the abnormal frequency density can be calculated as 4.5 times / second, belonging to the low-frequency range.

[0136] For the minute-level abnormal persistence analysis, a continuous feature evaluation function is constructed:

[0137]

[0138] wherein is the persistence feature value, is the duration weight factor, is the characteristic time constant (minutes), and m is the number of abnormal events. In practical applications, for example, when the temperature of a certain pressure vessel shows abnormal persistence for 3.5 minutes, it can be determined to belong to the short-term abnormal domain according to this function.

[0139] For the hour-level abnormal periodicity analysis, a period feature recognition function is introduced:

[0140]

[0141] wherein is the period feature index, is the amplitude of the periodic component, is the characteristic period (hours), is the phase factor, is the number of periodic components. For example, if a certain storage tank shows regular pressure fluctuations during a 24-hour operation cycle with a period of about 4 hours, it belongs to the medium-frequency abnormal domain.

[0142] Based on the feature analysis results of the above three time scales, a multi-dimensional membership function is constructed:

[0143]

[0144] wherein is the membership matrix, , , respectively represent the frequency, persistence, and periodicity feature values, is the corresponding membership coefficient. This matrix structure realizes the comprehensive evaluation of the features of the three time scales.

[0145] Spatial transformation is performed through the operation quality interval mapping function:

[0146]

[0147] Among them, is the mass interval mapping value, is the interval weight coefficient, is the interval characteristic function, is the interval reference value, and z is the comprehensive characteristic vector. In specific applications, such as when evaluating the operating status of a pressure vessel, when the frequency belongs to the low value range (<5 times / second), the duration belongs to the instantaneous abnormal range (<1 minute), and the period belongs to the low-frequency abnormal range (>6 hours), through the calculation of the mapping function, this working condition will be determined as the high-quality operation interval.

[0148] The above describes the method for evaluating the quality of pressure vessel data based on big data in the embodiments of the present application. Next, the system for evaluating the quality of pressure vessel data based on big data in the embodiments of the present application will be described. Please refer to Figure 4 , an embodiment of the system for evaluating the quality of pressure vessel data based on big data in the embodiments of the present application includes:

[0149] A synchronization module, configured to obtain a physical condition mapping data set through physical constraint preprocessing and condition synchronization processing according to the condition data collected from the pressure vessel, including the internal pressure data, wall temperature data, and stress and strain data of the container;

[0150] A calculation module, configured to obtain a physical condition feature set by calculating physical constraint indicators, condition stability indicators, and data coordination indicators according to the physical condition mapping data set;

[0151] An analysis module, configured to analyze the physical coupling relationship and change trend between the physical condition mapping data sets through a deep neural network driven by physical knowledge according to the physical condition feature set, and obtain a discriminant result of the condition abnormal type;

[0152] A construction module, configured to construct a scoring matrix through physical constraint scoring according to the discriminant result of the condition abnormal type, and perform dynamic calibration on the scoring matrix to obtain a physical condition evaluation factor;

[0153] An evaluation module, configured to calculate the operating quality grade of the pressure vessel through multi-scale fuzzy evaluation according to the physical condition evaluation factor and the discriminant result of the condition abnormal type, and obtain an evaluation result of the equipment data quality.

[0154] Through the collaborative cooperation of the above-mentioned various components, through the physical constraint preprocessing and working condition synchronization processing of the working condition data collected from the pressure vessel, the construction of the physical mapping relationship of multi-source data is realized, effectively avoiding the deviation caused by simply relying on the statistical characteristics of data; by calculating the physical constraint index, working condition stability index and data coordination index, a complete set of physical working condition characteristics is established, enabling the evaluation process to fully consider the constraints of physical laws; a deep neural network driven by physical knowledge is used to analyze the physical coupling relationship and change trend between the physical working condition mapping data sets. This network innovatively adopts a dual-stream structure, and processes the temporal features and physical constraint features through the temporal stream branch structure and the physical stream branch structure respectively, and realizes the dynamic fusion of features through the attention mechanism, significantly improving the accuracy and interpretability of anomaly recognition, and fully reflecting the deep integration of artificial intelligence algorithms and physical knowledge; through the construction of a scoring matrix by physical constraint scoring and dynamic calibration, the real-time update and adaptive adjustment of the evaluation results are realized, improving the timeliness and adaptability of the evaluation; the multi-scale fuzzy evaluation is used to calculate the operation quality grade of the pressure vessel, and the operation state of the equipment is analyzed on different time scales, enhancing the comprehensiveness and reliability of the evaluation results; at the algorithm level, the dual-stream structure design of the deep neural network makes full use of the advantages of the long short-term memory network in processing temporal data and the characteristics of the graph convolutional network in extracting physical constraint relationships. The introduction of the attention mechanism realizes the intelligent fusion of temporal features and physical constraint features. This algorithm architecture innovation effectively solves the problems of insufficient feature extraction and unclear physical constraint expression in traditional methods; in the evaluation index system, through the organic combination of physical constraint indexes, working condition stability indexes and data coordination indexes, an evaluation system that conforms to physical laws and meets the actual engineering needs is constructed, making the evaluation results have stronger physical significance and practical guiding value, and being able to accurately identify various abnormal states in the operation process of the pressure vessel, significantly improving the operation safety management level of the pressure vessel.

[0155] Referring to Figure 5 , in the embodiment of the present invention, a computer device is further provided. This computer device can be a server, and its internal structure can be as Figure 5 shown. This computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor of this computer design is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of this computer device is used to store the corresponding data in this embodiment. The network interface of this computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0156] Those skilled in the art can understand that Figure 5 the structure shown in Figure 5 is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0157] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0158] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0159] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0160] If the integrated unit is implemented in the form of 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 the present invention, in essence, or the part that contributes to the prior art, or all or 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 for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0161] As described above, the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A pressure vessel data quality assessment method based on big data, characterized in that: The pressure vessel data quality assessment method based on big data includes: According to the working condition data collected from the pressure vessel, including the internal pressure data of the vessel, the wall temperature data and the stress-strain data, a physical working condition mapping data set is obtained through physical constraint preprocessing and working condition synchronization processing; According to the physical working condition mapping data set, a physical working condition feature set is obtained by calculating a physical constraint index, a working condition stability index and a data coordination index; According to the physical working condition feature set, a physical coupling relationship and a change trend between the physical working condition mapping data sets are analyzed by a deep neural network driven by physical knowledge to obtain a result of distinguishing the type of working condition abnormality; According to the abnormal working condition type discrimination result, a scoring matrix is ​​constructed through physical constraint scoring, and the scoring matrix is ​​dynamically calibrated to obtain physical working condition evaluation factors, including: The results of the abnormal working condition type discrimination are divided into three types: pressure abnormality, temperature abnormality and strain abnormality, and the pressure abnormality is divided into overpressure abnormality and underpressure abnormality, the temperature abnormality is divided into overtemperature abnormality and temperature fluctuation abnormality, and the strain abnormality is divided into strain overlimit abnormality and strain mutation abnormality, to generate an abnormal feature group; the abnormal feature group is divided into weight intervals according to the pressure vessel safety constraint criteria, wherein the weight interval of the pressure abnormality is 0.4 to 0.5, the weight interval of the temperature abnormality is 0.3 to 0.4, and the weight interval of the strain abnormality is 0.2 to 0.3, to obtain a feature weight matrix; the feature weight matrix Perform scoring mapping, map the overpressure anomaly and the strain overlimit anomaly into a high-risk scoring interval, map the temperature fluctuation anomaly and the strain mutation anomaly into a medium-risk scoring interval, and map the underpressure anomaly into a low-risk scoring interval, and construct an initial scoring matrix; dynamically update the scoring values ​​in the initial scoring matrix according to the duration of the anomaly and the degree of the anomaly in the working condition anomaly type discrimination result to obtain a calibration scoring matrix; perform weighted combination of the scoring values ​​in the calibration scoring matrix according to the degree of hazard of the anomaly type to generate a working condition assessment value; perform normalization conversion on the working condition assessment value to obtain the physical working condition assessment factor; According to the physical operating condition evaluation factor and the abnormal operating condition type identification result, the operation quality grade of the pressure vessel is calculated through multi-scale fuzzy evaluation to obtain the equipment data quality evaluation result.

2. The pressure vessel data quality assessment method based on big data according to claim 1 is characterized in that: The physical condition mapping data set is obtained by performing physical constraint preprocessing and condition synchronization processing based on the condition data collected from the pressure vessel, including the vessel internal pressure data, wall temperature data and stress and strain data, including: Converting the fluctuation points in the container internal pressure data into a pressure trend sequence, converting the temperature points in the wall temperature data into a thermal sequence, and converting the strain points in the stress-strain data into a deformation sequence; Segmented sampling of the pressure trend sequence, the thermal sequence and the deformation sequence is performed through a time window to obtain working condition data segments; Associating and pairing the pressure trend sequence and the thermal sequence in the operating condition data segment to generate a temperature-pressure mapping group; Associating and pairing the thermal sequence and the deformation sequence in the operating condition data segment to generate a thermal stress mapping group; Performing working condition coupling processing on the temperature-pressure mapping group and the thermal stress mapping group to obtain working condition correlation characteristics; Physical constraint compensation and working condition synchronization calibration are performed according to the working condition association characteristics to obtain the physical working condition mapping data set.

3. The pressure vessel data quality assessment method based on big data according to claim 1 is characterized in that: The physical working condition feature set is obtained by calculating the physical constraint index, the working condition stability index and the data coordination index according to the physical working condition mapping data set, including: Dividing the physical working condition mapping data set into working condition time period groups according to a time series, calculating fluctuation values ​​for the data in the working condition time period groups, and obtaining physical constraint indicators; Calculating the change rate of the internal pressure data of the container in the working condition period group, determining the working condition stable interval according to the pressure change trend, and obtaining the working condition stability index; Performing correlation analysis on the wall temperature data and stress-strain data in the working condition period group, generating a thermal stress correlation, and obtaining a data coordination index; Combining the physical constraint index, the working condition stability index and the data coordination index to construct a feature vector; Performing normalized mapping on the feature vector to generate a standardized feature set of the physical constraint index, the working condition stability index, and the data coordination index; The indicators in the standardized feature set are combined and associated to construct a working condition feature mapping relationship to obtain the physical working condition feature set.

4. The pressure vessel data quality assessment method based on big data according to claim 1 is characterized in that: The physical coupling relationship and change trend between the physical working condition mapping data sets are analyzed by a deep neural network driven by physical knowledge according to the physical working condition feature set to obtain a working condition abnormality type discrimination result, including: The deep neural network is a dual-stream structure, including a temporal stream branch structure and a physical stream branch structure. The temporal stream branch structure uses a long short-term memory network to process temporal features, and the physical stream branch structure uses a graph convolutional network to process physical constraint features. The outputs of the two branches are fused through an attention mechanism to obtain a physical knowledge-driven deep neural network. Inputting the physical working condition feature set into the long short-term memory network, extracting the time series change features, and obtaining a time series feature vector; Inputting the physical working condition mapping data set into the graph convolutional network, extracting the physical constraint relationship, and obtaining a physical constraint vector; Performing weight fusion on the temporal feature vector and the physical constraint vector through the attention mechanism to obtain a fused feature matrix; Calculate the abnormal probability distribution according to the fusion feature matrix and generate an abnormal mapping vector; The abnormal mapping vector is subjected to type attribution analysis to obtain the abnormal type discrimination result of the operating condition.

5. The pressure vessel data quality assessment method based on big data according to claim 1 is characterized in that: The operation quality grade of the pressure vessel is calculated by multi-scale fuzzy evaluation according to the physical operating condition evaluation factor and the abnormal operating condition type identification result to obtain the equipment data quality evaluation result, including: Dividing the physical working condition evaluation factor into time scales, constructing a second-level evaluation domain, a minute-level evaluation domain, and an hour-level evaluation domain, and obtaining a multi-scale evaluation index; The abnormality type discrimination result of the working condition is used to perform abnormality frequency statistics in the second-level evaluation domain, to perform abnormality persistence analysis in the minute-level evaluation domain, and to perform abnormality periodicity analysis in the hour-level evaluation domain to obtain a time scale characteristic value; A fuzzy membership function is established according to the time scale characteristic value, and the operation quality of the pressure vessel is divided into a high-quality operation range, a normal operation range, an abnormal operation range and a dangerous operation range to obtain a fuzzy evaluation space; Mapping the multi-scale evaluation index to the fuzzy evaluation space, calculating the membership value of each operating interval, and obtaining a fuzzy evaluation vector; Performing multi-scale weighted fusion on the fuzzy evaluation vector to generate a quality grade score value; The quality grade score values ​​are graded and divided according to a set threshold to obtain the equipment data quality assessment result.

6. The pressure vessel data quality assessment method based on big data according to claim 5 is characterized in that: The fuzzy membership function is established according to the time scale characteristic value, and the operation quality of the pressure vessel is divided into a high-quality operation range, a normal operation range, an abnormal operation range and a dangerous operation range, and a fuzzy evaluation space is obtained, including: Performing statistical analysis on the second-level abnormal frequency in the time scale characteristic value, dividing the abnormal frequency less than 5 times / second into the low frequency value range, the abnormal frequency between 5 and 10 times / second into the medium frequency value range, and the abnormal frequency greater than 10 times / second into the high frequency value range, to obtain the frequency distribution characteristics; Statistically analyzing the minute-level abnormal persistence in the time scale characteristic value, dividing the abnormal duration of less than 1 minute into an instantaneous abnormal domain, the abnormal duration of 1-5 minutes into a short-term abnormal domain, and the abnormal duration of more than 5 minutes into a continuous abnormal domain, to obtain a persistence feature; Performing statistical analysis on the hourly abnormal periodicity in the time scale characteristic value, dividing the abnormal period less than 2 hours into a high-frequency abnormal domain, the abnormal period between 2 and 6 hours into a medium-frequency abnormal domain, and the abnormal period greater than 6 hours into a low-frequency abnormal domain, to obtain periodic characteristics; Establishing a membership function rule according to the frequency distribution feature, the persistence feature and the period feature to obtain an interval membership relationship; Mapping the interval membership relationship into an operation quality interval, constructing the membership boundaries of a high-quality operation interval, a normal operation interval, an abnormal operation interval, and a dangerous operation interval, and obtaining an evaluation interval matrix; The evaluation interval matrix is ​​subjected to space mapping transformation to obtain the fuzzy evaluation space.

7. A pressure vessel data quality assessment system based on big data, used to implement the pressure vessel data quality assessment method based on big data as described in any one of claims 1 to 6, characterized in that: The pressure vessel data quality assessment system based on big data includes: A synchronization module is used to obtain a physical working condition mapping data set through physical constraint preprocessing and working condition synchronization processing based on working condition data collected from the pressure vessel, including vessel internal pressure data, wall temperature data and stress-strain data; A calculation module, used for obtaining a physical working condition feature set by calculating a physical constraint index, a working condition stability index and a data coordination index according to the physical working condition mapping data set; An analysis module is used to analyze the physical coupling relationship and change trend between the physical working condition mapping data sets through a deep neural network driven by physical knowledge according to the physical working condition feature set, so as to obtain a result of distinguishing the type of working condition abnormality; A construction module is used to construct a scoring matrix through physical constraint scoring according to the abnormal working condition type discrimination result, and dynamically calibrate the scoring matrix to obtain a physical working condition evaluation factor, including: distinguishing the abnormal working condition type discrimination result into three types: pressure abnormality, temperature abnormality and strain abnormality, and dividing the pressure abnormality into overpressure abnormality and underpressure abnormality, dividing the temperature abnormality into overtemperature abnormality and temperature fluctuation abnormality, and dividing the strain abnormality into strain overlimit abnormality and strain mutation abnormality to generate an abnormal feature group; dividing the abnormal feature group into weight intervals according to the pressure vessel safety constraint criterion, wherein the weight interval of the pressure abnormality is 0.4 to 0.5, the weight interval of the temperature abnormality is 0.3 to 0.4, and the weight interval of the strain abnormality is 0.6 to 0.

8. The normal weight interval is 0.2 to 0.3, and a feature weight matrix is ​​obtained; the feature weight matrix is ​​scored and mapped, the overpressure anomaly and the strain overlimit anomaly are mapped as high-risk scoring intervals, the temperature fluctuation anomaly and the strain mutation anomaly are mapped as medium-risk scoring intervals, and the underpressure anomaly is mapped as a low-risk scoring interval, and an initial scoring matrix is ​​constructed; the score values ​​in the initial scoring matrix are dynamically updated according to the duration of the anomaly and the degree of anomaly in the result of the abnormal condition type discrimination, and a calibration scoring matrix is ​​obtained; the score values ​​in the calibration scoring matrix are weighted and combined according to the degree of hazard of the abnormal type to generate a working condition evaluation value; the working condition evaluation value is normalized and converted to obtain the physical working condition evaluation factor; The evaluation module is used to calculate the operation quality grade of the pressure vessel through multi-scale fuzzy evaluation according to the physical operating condition evaluation factor and the abnormal operating condition type discrimination result, so as to obtain the equipment data quality evaluation result.

8. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the pressure vessel data quality assessment method based on big data as described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the processor is enabled to execute the pressure vessel data quality assessment method based on big data as described in any one of claims 1 to 6.

Images