Petrochemical pipeline full life cycle detection and evaluation method and equipment

Abstract

The invention provides a petrochemical pipeline full-life-cycle detection and evaluation method and equipment. The method comprises the following steps: constructing a pipeline detection digital twin model comprising a geometric model and material attributes; collecting detection data of the pipeline through the deployed multi-source sensor network; inputting the detection data into the pipeline detection digital twin model to perform feature extraction and analysis of the pipeline state; constructing a heterogeneous tree network and executing hierarchical evaluation; optimizing the maintenance plan based on the evaluation result; and predicting the residual life of the pipeline. Through the digital twinning technology and multi-source detection data fusion, full-life-cycle accurate evaluation of the petrochemical pipeline is realized.

Description

Technical Field

[0001] The present application relates to the field of industrial intelligence and digital twin technology, and in particular to a method and equipment for detecting and evaluating the entire life cycle of a petrochemical pipeline. Background Art

[0002] As an important infrastructure of the petrochemical industry, the safe operation of petrochemical pipelines is directly related to production safety and environmental protection. With the development of industrial intelligence and digital transformation, pipeline detection and evaluation technology is also constantly innovating.

[0003] At present, the industrial field mainly adopts periodic inspection methods based on single detection technology, such as regular inspection by ultrasonic or X-ray means. Although this method can obtain pipeline status data at a certain moment, it is difficult to achieve continuous monitoring and timely warning, and the utilization efficiency of detection data is low.

[0004] The traditional pipeline condition assessment system adopts a multi-source detection data collection method, which can obtain a variety of detection data including ultrasonic, electromagnetic, and radiographic data. However, the system has deficiencies in data fusion and in-depth analysis, and it is difficult to give full play to the advantages of multi-source data, resulting in unsatisfactory accuracy and timeliness of the assessment results.

[0005] In particular, existing technologies are inefficient in collecting and fusing multi-source heterogeneous detection data, making it difficult to support comprehensive and accurate status assessment, and lack in-depth analysis of the dynamic characteristics of pipeline systems, which directly affects the accuracy of the assessment results and cannot meet the needs of full life cycle management of petrochemical pipelines. Summary of the invention

[0006] In view of this, the present application provides a petrochemical pipeline full life cycle detection and evaluation method and equipment, which solves the problem in the prior art that the collection and fusion of multi-source heterogeneous detection data is inefficient and difficult to support comprehensive and accurate status evaluation.

[0007] The present application embodiment provides a petrochemical pipeline full life cycle detection and evaluation method, including: A pipeline inspection digital twin model including a geometric model and material properties is constructed; pipeline inspection data is collected through a deployed multi-source sensor network; the inspection data is input into the pipeline inspection digital twin model, and pipeline status feature extraction and analysis are performed in the pipeline inspection digital twin model to generate a pipeline status feature evolution report; a heterogeneous tree network is constructed according to the pipeline status feature evolution report, and a hierarchical evaluation is performed on the pipeline system based on the hierarchical structure of the heterogeneous tree network to generate a hierarchical evaluation result; a maintenance plan is optimized based on the hierarchical evaluation result; and the remaining life of the pipeline is predicted according to the optimized maintenance plan.

[0008] The construction of the pipeline detection digital twin model includes: constructing a geometric model of the physical pipeline system based on the engineering drawings and field measurement data of the pipeline system; establishing a material property model including elastic modulus, Poisson's ratio and thermal expansion coefficient according to the physical characteristic parameters of the pipeline system; establishing a pipeline system operation parameter model including pressure, temperature and flow rate based on the geometric model and the material property model; constructing a coordinate mapping relationship between physical space and virtual space based on the geometric model, the material property model and the operation parameter model; establishing state parameter mapping of each component of the pipeline system based on the coordinate mapping relationship, and realizing a real-time data synchronization mechanism between the physical system and the virtual model.

[0009] The present application also provides a computer device, the computer device comprising: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned petrochemical pipeline full life cycle detection and evaluation method.

[0010] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned petrochemical pipeline full life cycle detection and evaluation method.

[0011] An embodiment of the present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the above-mentioned petrochemical pipeline full life cycle detection and evaluation method.

[0012] This application has the following technical effects: By constructing a heterogeneous tree network model with edge weight uncertainty, the efficient fusion of multi-source detection data is achieved, and the accuracy of pipeline status assessment is improved. The maintenance plan is optimized using the k-swap neighborhood smoothing analysis method, which significantly improves the scientificity and executability of maintenance decisions. Life prediction based on implicit high-order moment tensor estimation technology effectively improves the prediction accuracy and provides reliable data support for pipeline life cycle management. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0014] Figure 1 A schematic diagram of the process of the petrochemical pipeline full life cycle detection and evaluation method provided in the embodiment of the present application; Figure 2 This is a schematic diagram of the construction process of the pipeline detection digital twin model in the embodiment of the present application; Figure 3 A schematic diagram of the deployment of a multi-source sensor network in an embodiment of the present application; Figure 4 This is a schematic diagram of the structure of a heterogeneous tree network in an embodiment of the present application; Figure 5 Schematic diagram of the optimization process of the k-swap neighborhood smoothing analysis method in the embodiment of the present application; Figure 6 Schematic diagram of the life prediction process based on implicit high-order moment tensor estimation in an embodiment of the present application. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0016] See also Figure 1-Figure 6 The embodiment of the present application provides a petrochemical pipeline full life cycle detection and evaluation method and equipment. The technical solution of the embodiment of the present application is described in detail below in conjunction with the accompanying drawings.

[0017] like Figure 1 As shown, the technical solution of the petrochemical pipeline full life cycle detection and evaluation method provided by this application can be summarized into the following steps: S1: Build a pipeline inspection digital twin model including geometric model and material properties; like Figure 2As shown in FIG. 1 , the specific implementation process of constructing the pipeline detection digital twin model in step S1 is as follows: First, a geometric model of the physical pipeline system is constructed based on the engineering drawings and field measurement data of the pipeline system. The geometric model needs to accurately reflect the three-dimensional spatial structural characteristics of the pipeline system, including key parameters such as the direction, diameter, and wall thickness of the pipeline. Secondly, a material property model is established based on the physical property parameters of the pipeline system. The model contains material performance parameters such as elastic modulus, Poisson's ratio, and thermal expansion coefficient, which are of great significance for accurately evaluating the mechanical properties and deformation characteristics of the pipeline. Thirdly, based on the geometric model and the material property model, a pipeline system operation parameter model is established, which includes dynamic operation parameters such as pressure, temperature, and flow rate, which can reflect the real-time operation status of the pipeline system. Finally, by constructing the coordinate mapping relationship between the physical space and the virtual space, the state parameter mapping of each component of the pipeline system is realized, thereby establishing a real-time data synchronization mechanism between the physical system and the virtual model.

[0018] S1 specifically includes: S1.1: Construct a geometric model of the physical piping system based on the engineering drawings and field measurement data of the piping system; The construction of the geometric model of the physical pipeline system is the basis of the digital twin model. First, it is necessary to collect and organize the engineering drawings of the pipeline system, including the plan layout, longitudinal section and cross-section of the pipeline, etc. These drawings contain the basic geometric information of the pipeline system. Secondly, the actual dimensional data of the pipeline system is obtained through on-site field measurements, including key parameters such as the spatial position coordinates of the pipeline, pipe section length, pipeline diameter, wall thickness distribution, etc. Thirdly, based on the collected drawing information and measurement data, a precise geometric model of the pipeline system is constructed using a 3D modeling tool. In addition, during the modeling process, special attention should be paid to the geometric features of special parts such as pipeline joints, valves, elbows, etc. to ensure that the model can accurately reflect the structural details of these key parts.

[0019] S1.2: Based on the physical characteristic parameters of the pipeline system, establish a material property model including elastic modulus, Poisson's ratio and thermal expansion coefficient; The establishment of the material property model is directly related to the accuracy of the assessment of the performance of the pipeline system. First, according to the design specifications of the pipeline system and the actual materials used, determine the key physical property parameters that need to be modeled. For example, these parameters include the elastic modulus that characterizes the elastic deformation capacity of the material, the Poisson's ratio that reflects the ratio of lateral deformation to axial deformation, and the thermal expansion coefficient that describes the thermal expansion performance of the material. Secondly, for different materials that may be used in different pipe sections, establish a corresponding material property database. It should be noted that material property parameters may change with environmental factors such as temperature and pressure, so the dynamic characteristics of these parameters need to be considered when modeling.

[0020] S1.3: Based on the geometric model and the material property model, establish a pipeline system operation parameter model including pressure, temperature and flow rate; The operating parameter model is an important component that reflects the dynamic operating status of the pipeline system. First, based on the established geometric model and material property model, an operating parameter model including pressure distribution, temperature field and flow field is constructed. Specifically, the pressure model needs to consider the pressure distribution law and pressure fluctuation characteristics in the pipeline system, the temperature model needs to reflect the temperature distribution and heat conduction characteristics of the pipeline system, and the flow velocity model needs to describe the flow characteristics and velocity distribution of the medium in the pipe. Secondly, a coupling relationship model between these operating parameters is established to accurately reflect the mutual influence between the parameters.

[0021] S1.4: Constructing a coordinate mapping relationship between the physical space and the virtual space according to the geometric model, the material property model and the operation parameter model; The coordinate mapping relationship between physical space and virtual space is the key to realizing digital twins. First, a unified coordinate system needs to be established to establish a correspondence between the location information of the pipeline system in the physical space and the model coordinates in the virtual space. Secondly, a coordinate transformation matrix is ​​constructed to achieve accurate mapping between the physical space and the virtual space. In addition, considering the possible deformation of the pipeline system during operation, a dynamic coordinate mapping mechanism needs to be established to adapt to the geometric changes of the system.

[0022] S1.5: Based on the coordinate mapping relationship, a state parameter mapping of each component of the pipeline system is established to realize a real-time data synchronization mechanism between the physical system and the virtual model; The state parameter mapping in step S1.5 realizes the real-time synchronization of the physical system and the virtual model. First, establish the state parameter mapping table of each component of the pipeline system, and clarify the parameter type and data format that need to be synchronized for each component. Secondly, design the data transmission and update mechanism to ensure that the state changes in the physical system can be reflected in the virtual model in real time. Thirdly, realize the automated processing flow of data synchronization, including data collection, transmission, conversion and update. It should be noted that in order to ensure the real-time and reliability of synchronization, the system should also have data caching and exception handling mechanisms to deal with abnormal situations such as network delays or data transmission interruptions.

[0023] S2: Collect pipeline detection data through the deployed multi-source sensor network; like Figure 3As shown, the multi-source sensor network deployment and data acquisition process in step S2 includes: arranging various types of sensors at key nodes of the pipeline system, including ultrasonic probes for detecting pipe wall thickness and internal defects, electromagnetic sensor arrays for monitoring surface corrosion and cracks, radiographic detection equipment for detecting weld quality, and vibration sensors and temperature sensors for collecting operating status parameters. It should be noted that the installation location and sampling frequency of the sensors need to be optimized based on the digital twin model to ensure the comprehensiveness and accuracy of data acquisition. For the collected raw data, data cleaning is first performed, including eliminating noise, removing outliers, and filling missing values; secondly, corresponding filtering algorithms are used for signal optimization for different types of sensor data, such as using wavelet transform to denoise ultrasonic data and applying Kalman filtering to vibration data; finally, the processed data is stored in a standardized manner, and timestamps and spatial location tags are added to ensure data traceability.

[0024] S2.1: deploy a multi-source sensor network at key nodes of the pipeline system, wherein the sensor network includes ultrasonic probes, electromagnetic sensor arrays, radiation detection equipment, vibration sensors, and temperature sensors; The multi-source sensor network deployment in step S2.1 is the basis of the entire detection system. First of all, it is necessary to scientifically plan the deployment location of the sensors based on the structural characteristics and operation requirements of the pipeline system. Specifically, ultrasonic probes are mainly arranged at key nodes of the pipeline system to monitor the changes in pipe wall thickness and the development of internal defects in real time. The electromagnetic sensor array needs to be evenly distributed along the surface of the pipeline to achieve comprehensive monitoring of pipeline surface corrosion and cracks. X-ray detection equipment is usually arranged at the location of pipeline welds to collect weld quality data regularly. The installation location of vibration sensors and temperature sensors needs to take into account the dynamic characteristics and thermal characteristics of the pipeline system to ensure that the operating status parameters of the system can be accurately captured. In addition, the installation method and fixed structure of various sensors need to ensure the stability and reliability of the measurement, while considering the convenience of maintenance.

[0025] S2.2: Collecting detection data of the pipeline system using the multi-source sensor network; In step S2.2, the data collection process of the multi-source sensor network needs to coordinate the working modes of various sensors. First, a reasonable sampling strategy is formulated according to the characteristics and monitoring requirements of different types of sensors. For example, the ultrasonic probe can use a timed scanning method to measure the wall thickness at specific time intervals; the electromagnetic sensor array needs to be continuously monitored to collect surface status data in real time; the X-ray detection equipment can carry out regular inspections according to the predetermined inspection cycle. Secondly, a clock synchronization mechanism for the sensor network is established to ensure that the data collected by different sensors have a unified time reference. In addition, according to the data transmission characteristics of different sensors, corresponding data caching and transmission strategies are designed to avoid data loss or delay.

[0026] S2.3: performing data cleaning on the collected detection data; Data cleaning in step S2.3 is a key link in improving the quality of detection data. First, the collected raw data is preprocessed to identify and mark abnormal data points. Specifically, it is necessary to use statistical analysis methods to identify data that deviates significantly from the normal range, and combine professional knowledge to determine the causes of these abnormal values. Secondly, take corresponding processing measures for different types of abnormal data. For example, random noise can be smoothed by a filtering algorithm, abnormal values ​​caused by sudden interference can be corrected by interpolation methods, and data missing caused by equipment failures need to be reasonably supplemented based on adjacent data. It should be noted that the original data records should be retained during the data cleaning process to facilitate subsequent tracing and verification.

[0027] S2.4: performing signal optimization on the cleaned detection data; Regarding the signal optimization of step S2.4, it is necessary to select appropriate processing methods according to the characteristics of different types of sensor data. First, for ultrasonic detection data, the wavelet transform method is used for noise reduction processing, which can effectively remove high-frequency interference signals while retaining useful defect information. Secondly, for electromagnetic detection signals, an adaptive filtering algorithm is used to eliminate the influence of environmental electromagnetic interference. Thirdly, for vibration signals, the Kalman filtering method is applied to achieve real-time optimization of the signal and improve the recognition accuracy of vibration characteristics. In addition, the processing of temperature signals needs to take into account the influence of ambient temperature and improve the measurement accuracy through a temperature compensation algorithm.

[0028] S2.5: adding a timestamp and a spatial position tag to the optimized data to generate a standardized detection data set; In step S2.5, standardization is the basis for realizing multi-source data fusion. First, a unified data standardization framework is established to convert detection data of different sources and types into the same metric space. Specifically, for numerical data, the data is mapped to the interval [0,1] through normalization; for categorical data, the encoding conversion method is used to achieve standardized representation of the data. Secondly, complete timestamp information and spatial location tags are added to each set of data to establish a spatiotemporal index system for the data. Thirdly, a data quality assessment indicator system is constructed to conduct quality assessment on the standardized data to ensure the availability and reliability of the data. It should be noted that standardization also needs to consider the scalability of the data and reserve interfaces to adapt to newly added data types and formats.

[0029] S3: Extract and analyze pipeline status features; Among them, S3 includes the following steps: S3.1: Based on the standardized inspection data set, a multidimensional feature extraction framework including geometric features, physical features and dynamic features is constructed in the pipeline inspection digital twin model; The construction of a multidimensional feature extraction framework is the core of pipeline state analysis. First, based on the standardized detection data in the digital twin model, a multidimensional feature extraction framework containing geometric features, physical features, and dynamic features is constructed. Specifically, geometric features include spatial features such as pipeline deformation and wall thickness changes, which directly reflect the geometric state changes of the pipeline. Physical features include state features such as material performance parameters and stress distribution, which reveal the mechanical performance state of the pipeline. Dynamic features mainly include time series features such as vibration characteristics and temperature changes, which reflect the dynamic characteristics of the operation of the pipeline system. In addition, the feature extraction framework also needs to establish a correlation analysis mechanism between features, identify the dependencies between features, and ensure the integrity and representativeness of the extracted features.

[0030] S3.2: using the multidimensional feature extraction framework, extracting the geometric features of the pipeline system through a morphological analysis method, extracting the physical features of the pipeline system through a stress-strain analysis model, and extracting the dynamic features of the pipeline system through a time-frequency analysis technique; For the feature extraction process of step S3.2, it is necessary to adopt corresponding analysis methods for different types of features. First, for the extraction of geometric features, the morphological analysis method is adopted. In the specific implementation, the geometric contour changes of the pipeline are analyzed by image processing technology, the deformation and displacement of key parts are calculated, and the accurate description of the geometric state of the pipeline is achieved. Secondly, in terms of physical feature extraction, the stress-strain analysis model is applied. The stress distribution of the pipeline system is calculated by the finite element analysis method, the fatigue state of the material is evaluated, and the possible failure location is predicted. Thirdly, for the extraction of dynamic features, time-frequency analysis technology is used. Through methods such as Fourier transform and wavelet analysis, spectral features are extracted from the vibration signal, abnormal vibration modes are identified, and the dynamic performance of the system is evaluated. It should be noted that in the feature extraction process, the physical meaning of the features should be fully considered to ensure that the extracted features can accurately reflect the actual state of the pipeline system.

[0031] S3.3: constructing a feature hierarchy model based on the extracted geometric features, physical features and dynamic features, and using a deep learning feature fusion algorithm to fuse the geometric features, physical features and dynamic features to generate fused features; The feature fusion in step S3.3 is a key link to improve the accuracy of state assessment. First, the multi-dimensional features extracted in step S3.2 are hierarchically classified to establish a feature hierarchy model. The model needs to clearly express the correlation between features at different levels. Secondly, a feature fusion algorithm based on deep learning is designed, and a multi-layer neural network structure is used to achieve adaptive fusion of features. Specifically, local features are extracted through the convolution layer, the feature dimension is reduced through the pooling layer, and the deep fusion of features is achieved through the fully connected layer. Thirdly, a feature importance evaluation module is constructed, and a feature weight calculation method based on the attention mechanism is used to highlight the influence of key features. For example, for the key feature of pipeline wall thickness, a higher attention weight can be given to enhance its contribution in the fusion process. In addition, the fusion algorithm also needs to have adaptive learning capabilities and be able to continuously optimize the fusion strategy based on historical data.

[0032] S3.4: Generate a pipeline system state feature vector based on the fusion feature, and track the dynamic change law of the state feature vector using a time series analysis method; Regarding the generation of the state feature vector in step S3.4, it is necessary to establish a system feature organization mechanism. First, based on the fusion features of step S3.3, construct a feature vector that reflects the overall state of the pipeline system. This feature vector needs to contain comprehensive information such as the health status of the pipeline system and performance change trends. Secondly, use the time series analysis method to track the feature vector in time series. By establishing a time series prediction model, analyze the changing patterns of the features and predict future development trends. Thirdly, design a feature anomaly detection mechanism to identify abnormal feature changes by setting dynamic thresholds. In specific implementation, a method based on statistical control charts can be used to monitor the changing trends of features, and a warning signal can be issued in time when the feature value exceeds the control limit.

[0033] S3.5: construct a feature evolution report including feature change trends and abnormal warning information according to the dynamic change law; In the process of generating the feature evolution report in step S3.5, the above analysis results need to be systematically integrated. First, based on the state feature vector obtained in step S3.4, the temporal evolution law of the feature is analyzed. Through statistical analysis methods, the characteristics of the periodicity, trend and mutation of feature changes are identified. Secondly, a feature trend prediction model is established, and the time series analysis method is used to predict the changing trend of the feature. This prediction model needs to consider the role of external conditions such as seasonal factors and environmental influences. Thirdly, a feature anomaly warning mechanism is constructed to generate targeted warning information by combining the prediction results and threshold criteria. Finally, a complete feature evolution report is formed. The report needs to present the feature change law, trend prediction results and abnormal warning information in a clear logical structure to provide comprehensive data support for subsequent heterogeneous tree network evaluation.

[0034] S4: construct heterogeneous tree networks and perform hierarchical evaluation; like Figure 4 As shown, the heterogeneous tree network construction process in step S4 is as follows: First, based on the feature change rules in the feature evolution report, a network topology structure that reflects the physical connection relationship and logical dependency relationship of the pipeline system is constructed. This structure adopts a tree hierarchy, which can effectively express the hierarchical relationship between the components of the pipeline system. Secondly, an edge weight calculation model is designed, which comprehensively considers the reliability, importance and timeliness of the detection data, and quantifies the uncertainty of the data through information entropy theory. Specifically, for the edge e(A, B) connecting the main pipeline and the branch pipeline, the weight calculation formula is:

[0035] Where R represents the detection reliability component, E represents the environmental impact component, T represents the time evolution component, and α, β, and γ are the corresponding weight coefficients. In this embodiment, the value range of α is [0.3, 0.5], the value range of β is [0.2, 0.4], and the value range of γ is [0.2, 0.3]. The calculation of the detection reliability component R takes into account the uncertainty of multiple detection sources:

[0036] Among them, wi is the weight of each detection source, ri is the normalized detection result, and δr is the comprehensive uncertainty.

[0037] Among them, S4 is specifically: S4.1: Based on the feature change rules in the feature evolution report, a network topology structure reflecting the physical connection relationship and logical dependency relationship of the pipeline system is constructed; The construction of the heterogeneous tree network topology structure in step S4.1 is the basis for realizing hierarchical evaluation. First, it is necessary to deeply analyze the feature change rules in the feature evolution report and clarify the physical connection relationship and logical dependency relationship between the components of the pipeline system. Specifically, the physical connection relationship reflects the actual connection method of the pipeline system, including the connection between the main pipeline and the branch pipeline, the connection between the pipe sections, and the connection between the pipeline and the control equipment; the logical dependency relationship reflects the functional association between the system components, such as the flow relationship and pressure transmission relationship between the upstream and downstream pipe sections. Secondly, a hierarchical network structure is constructed based on these relationships to divide the pipeline system into different levels. For example, the main pipeline can be set as the first-level node, the branch pipeline directly connected to it can be set as the second-level node, and the terminal pipe section can be set as the third-level node, thereby forming a clear hierarchical structure. In addition, the network structure also needs to consider the dynamic characteristics of the nodes and reserve interfaces to adapt to changes in the system configuration.

[0038] S4.2: Design an edge weight calculation model based on the reliability, importance and timeliness of the detection data; In the design of the edge weight calculation model in step S4.2, multiple influencing factors need to be considered comprehensively. First, from the perspective of the reliability of the detection data, it is necessary to evaluate the data quality obtained by different detection methods. For example, for the connection edge from the main pipeline A to the branch pipeline B1, its detection data may come from multiple sources: ultrasonic detection shows that the wall thickness loss rate is 5%±1%, electromagnetic detection shows that the surface corrosion degree is level 3±0.5, and radiographic detection shows that the weld quality grade is level II±0.3. These detection data have different uncertainties and need to be quantified through information entropy theory. Secondly, from the perspective of data importance, it is necessary to consider the contribution of different data to the system status assessment. In specific implementation, a scoring system for data importance can be established based on expert experience and historical data analysis. Thirdly, from the perspective of timeliness, it is necessary to evaluate the time decay characteristics of the data. The time decay function can be used to describe the change in the credibility of the data over time to ensure that the evaluation results can reflect the latest status of the system in a timely manner.

[0039] S4.3: using the edge weight calculation model, using an adaptive weighting method to calculate the weight of each edge; For the edge weight calculation process in step S4.3, a scientific adaptive weighting mechanism needs to be established. First, for the weight calculation of each edge, a multi-factor comprehensive evaluation method is adopted. Taking the edge e(A, B1) connecting the main pipeline A and the branch pipeline B1 as an example, its weight calculation formula is: , where R represents the detection reliability component, E represents the environmental impact component, T represents the time evolution component, and α, β, and γ are the corresponding weight coefficients. In the specific implementation, the value range of α is [0.3, 0.5], which mainly considers the quality characteristics of the detection data; the value range of β is [0.2, 0.4], which reflects the degree of influence of environmental factors; the value range of γ is [0.2, 0.3], which reflects the importance of time evolution. Secondly, the calculation of the detection reliability component R needs to consider the uncertainty of multiple detection sources: , where wi is the weight of each detection source, ri is the normalized detection result, and δr is the comprehensive uncertainty. Thirdly, the environmental impact component E needs to consider the comprehensive effects of environmental parameters such as temperature, humidity, and soil corrosivity, and the time evolution component T needs to reflect the impact of pipeline use time on the connection relationship.

[0040] S4.4: Generate a dynamic adjustment factor for the edge weight, and update the edge weight in real time based on the dynamic adjustment factor; The dynamic update of edge weights in step S4.4 is the key to ensuring the accuracy of the evaluation. First, it is necessary to build a dynamic adjustment factor generation mechanism for edge weights. This mechanism needs to monitor the various factors that affect the weights in real time, including the update of detection data, changes in environmental conditions, and the evolution of system operation status. Secondly, based on these monitoring information, an adaptive algorithm is used to calculate the weight adjustment factor. For example, when the detection shows that the reliability of a certain edge has changed significantly, the system will automatically adjust the weight coefficient of the edge; when the environmental conditions are abnormal, the weights of the relevant edges will also be adjusted accordingly. Thirdly, a real-time update mechanism for weights is implemented. By establishing an efficient data processing process, it is ensured that the weight adjustment can respond to changes in the system status in a timely manner. In addition, it is also necessary to establish a validity verification mechanism for weight updates to ensure the rationality of weight adjustments through historical data analysis and expert evaluation.

[0041] In the petrochemical pipeline system, assume that there is a network structure consisting of a main pipeline, branch pipelines, and terminal pipe segments. The main pipeline A is connected to three branch pipelines B1, B2, and B3, and each branch pipeline is connected to several terminal pipe segments C1, C2, C3, etc. It is necessary to build a heterogeneous tree network that can reflect this hierarchical structure.

[0042] In this heterogeneous tree network, the uncertainty of edge weights is mainly reflected in the following aspects: First, the reliability of the test data is different. Taking the connection from main pipe A to branch pipe B1 as an example, the weight of this edge depends on the data from multiple test sources: ultrasonic testing shows that the wall thickness loss rate is 5%±1%, electromagnetic testing shows that the surface corrosion level is level 3±0.5, and radiographic testing shows that the weld quality level is level II±0.3. These test data all have different degrees of uncertainty, and their uncertainty needs to be quantified through information entropy theory.

[0043] Second, the dynamic influence of environmental factors. For example, the edge weight of the connection between branch pipe B2 and terminal pipe section C4 will be affected by factors such as ambient temperature, humidity, and soil corrosivity. Fluctuations in these environmental parameters can cause dynamic changes in weights. Through the Bayesian reasoning method, the embodiment of the present application can establish a probabilistic relationship model between environmental factors and weight changes.

[0044] Third, the uncertainty of time evolution. Taking branch pipe B3 as an example, as the service time increases, the edge weights between it and the connected pipe sections will change. This change has a cumulative effect and randomness, and it is necessary to establish a weight evolution model through time series analysis methods.

[0045] The specific edge weight calculation process is as follows: First, for the edge e(A,B1) connecting the main pipeline A and the branch pipeline B1, its weight calculation formula is:

[0046] Among them, R represents the detection reliability component, E represents the environmental impact component, T represents the time evolution component, and α, β, and γ are the corresponding weight coefficients.

[0047] The calculation of R takes into account the uncertainty of multiple detection sources:

[0048] Among them, wi is the weight of each detection source, ri is the normalized detection result, and δr is the comprehensive uncertainty.

[0049] This uncertain edge weight design enables the heterogeneous tree network to more accurately reflect the actual status of the pipeline system and provide more reliable data support for subsequent status assessment.

[0050] Through this example, the uncertainty of edge weights is reflected in multiple dimensions such as multi-source data fusion, environmental factors, and time evolution characteristics. This design is more in line with the needs of actual engineering applications. At the same time, this uncertainty quantification and processing method also provides a more scientific theoretical basis for the status assessment of pipeline systems.

[0051] S5: Optimize maintenance plan based on hierarchical assessment results; like Figure 5 As shown, the k-swap neighborhood smoothing analysis method in step S5 includes: first, establishing an initial maintenance plan based on the hierarchical evaluation results, which includes the priority, execution time and resource requirements of the maintenance tasks. Exemplarily, the initial maintenance task sequence is set to ACBDE, where tasks A and C are high priority tasks, tasks B and D are medium priority tasks, and task E is a low priority task. Secondly, the k-swap neighborhood generator performs task exchange operations in the time dimension and the space dimension to generate candidate maintenance plans. In this embodiment, k=2 is set, that is, the execution order of two maintenance tasks can be exchanged at the same time each time. After the k-swap operation, multiple candidate plans can be generated, such as CABDE, ABCDE, etc. Thirdly, a smooth optimization algorithm is constructed to evaluate the candidate plans, and the evaluation indicators include resource utilization balance, time span smoothness and priority satisfaction. Finally, a smooth maintenance plan is obtained through iterative optimization.

[0052] S5.1: Based on the hierarchical assessment results, establish an initial maintenance plan including maintenance task priorities, execution time and resource requirements; Based on the hierarchical assessment results, the maintenance demand priority of each component of the pipeline system is clarified. Specifically, for components that the assessment results show to be at higher risk, such as pipe sections with wall thickness loss close to the critical value or areas with obvious corrosion, a higher maintenance priority should be given. Secondly, the execution time requirements of the maintenance tasks should be comprehensively considered. For example, the weld inspection of the main pipeline section A is expected to take 4 days, and considering its importance, it should be given priority; while the replacement of the anti-corrosion layer of branch pipeline 1 is expected to take 3 days, and the execution sequence can be adjusted appropriately according to its risk level. Thirdly, the manpower, equipment and other resources required for the maintenance tasks should be planned in a coordinated manner. It should be noted that resource planning should take into account the professional skills matching of maintenance personnel, the availability of equipment and the continuity requirements of maintenance work.

[0053] S5.2: Generate candidate maintenance plans by performing task swap operations in the time dimension and the space dimension using the k-swap neighborhood generator; For the k-swap neighborhood generation process in step S5.2, a scientific task exchange mechanism needs to be established. First, in the time dimension, candidate solutions are generated by adjusting the execution order of maintenance tasks. Taking the initial sequence ACBDE as an example, when k=2, multiple candidate sequences can be generated by exchanging adjacent tasks, such as CABDE (exchanging A and C), ABCDE (exchanging B and C), etc. Secondly, in the spatial dimension, the spatial distribution characteristics of maintenance tasks are considered to optimize the execution location of tasks. For example, for maintenance tasks with similar spatial locations, the movement costs of maintenance personnel and equipment can be reduced by adjusting their execution order. Thirdly, when generating candidate solutions, it is necessary to ensure that various constraints are met. For example, constraints such as the limit on the number of maintenance personnel (up to 5 people per day), the equipment downtime of no more than 7 days, and the priority of high-priority tasks must be arranged first need to be strictly observed in the solution generation process.

[0054] S5.3: Construct a smoothing optimization algorithm based on k-swap neighborhood; The construction of the smoothing optimization algorithm in step S5.3 is the core of achieving maintenance plan optimization. First, an iterative search strategy based on the k-swap neighborhood is designed. This strategy searches for a better maintenance solution by continuously exploring the neighborhood solution space. Secondly, a smoothing factor evaluation mechanism is introduced to evaluate the smoothness of candidate solutions from multiple dimensions such as resource utilization balance, time distribution rationality, and priority satisfaction. Thirdly, a simulated annealing mechanism is used to prevent the algorithm from falling into a local optimal solution, and a taboo search strategy is introduced to increase the exploration capability of the solution space. In specific implementation, the convergence efficiency of the algorithm can be improved by dynamically adjusting search parameters such as the temperature attenuation coefficient and the taboo length.

[0055] S5.4: using the smoothing optimization algorithm, performing iterative optimization on the candidate maintenance solutions; In step S5.4, a systematic evaluation system needs to be established for the iterative optimization process of candidate maintenance solutions. First, a comprehensive evaluation is conducted on each candidate solution, including resource utilization balance evaluation, time span smoothness evaluation, and priority satisfaction evaluation. For example, the evaluation of solution CABDE shows that: the staff demand is high in the early stage (7 people), low in the later stage (3 people), and the balance needs to be improved; the time distribution is relatively concentrated and the smoothness is good; high-priority tasks C and A are ranked at the front, and the priority satisfaction is high. Secondly, based on the evaluation results, the dynamic weight adjustment method is used to continuously optimize the search direction. Thirdly, through iterative calculation, the quality of the maintenance solution is gradually improved until the predetermined optimization goal or number of iterations is reached.

[0056] S5.5: Based on maintenance cost, equipment availability and risk control, a multi-objective evaluation system is constructed to evaluate the optimization results; For the construction of the multi-objective evaluation system in step S5.5, it is necessary to comprehensively consider multiple important indicators of the maintenance plan. First, from the perspective of maintenance costs, evaluate various expenses such as human resource investment, equipment use costs, and material consumption. Secondly, from the perspective of equipment availability, evaluate the impact of the maintenance plan on production operations, including factors such as downtime and capacity loss. Thirdly, from the perspective of risk control, evaluate the degree to which the maintenance plan guarantees system safety, including whether high-risk areas can be handled in a timely manner and whether emergency response needs are considered. In addition, it is necessary to establish a weight distribution mechanism among the evaluation indicators to ensure that the evaluation results can accurately reflect the overall pros and cons of the maintenance plan.

[0057] S5.6: Generate an optimized maintenance plan based on the evaluation results; The generation of the optimized maintenance plan in step S5.6 is the final output of the entire optimization process. First, based on the multi-objective evaluation results, the optimal maintenance plan is selected from the candidate solutions. Specifically, the optimized maintenance plan should achieve a relative balance in resource utilization, such as maintaining the demand for maintenance personnel at an average of 4-5 people per day. Secondly, formulate a detailed implementation plan, including the specific execution time, required resource allocation and technical requirements for each maintenance task. Thirdly, establish a plan adjustment mechanism to verify the robustness of the plan by simulating different interference situations (such as weather changes and staff absences), and formulate corresponding emergency plans. It should be noted that the final maintenance plan should also include a monitoring plan and quality assessment standards for the execution process to ensure that the plan can be effectively implemented.

[0058] Assume that a petrochemical plant has a set of critical pipelines that need maintenance. Based on the assessment results, the following initial maintenance tasks are required: Task A: Main pipeline section weld inspection (estimated time: 4 days, priority: high); Task B: Replacement of anti-corrosion layer of branch pipeline 1 (estimated time: 3 days, priority: medium); Task C: Valve assembly replacement (estimated time: 2 days, priority: high); Task D: Pipe wall reinforcement (estimated time: 3 days, priority: medium); Task E: Corrosion point repair (estimated time: 2 days, priority: low); In the k-swap neighborhood optimization process, the embodiment of the present application may be operated as follows: Step 1: Determine the k value and constraints Set k=2, that is, the execution order of two maintenance tasks can be swapped at the same time each time Constraints include: limited number of maintenance personnel (up to 5 per day), equipment downtime no more than 7 days, and high-priority tasks must be scheduled first Step 2: Generate neighborhood solutions The initial sequence is: ACBDE By exchanging k=2, the following neighborhood solutions can be generated: Solution 1: CABDE (swap A and C); Solution 2: ABCDE (swap B and C); Solution 3: ACDBE (swap B and D); Solution 4: ACBED (swap D and E); Step 3: Smoothing Analysis Each neighborhood solution is evaluated, considering the following smoothness metrics: 1. Balance of resource utilization: assess whether the use of maintenance personnel and equipment is balanced; 2. Time span smoothness: avoid over-concentration or over-dispersion of tasks; 3. Priority satisfaction: whether high-priority tasks are processed in a timely manner; For example, the smoothing analysis of solution 2 (ABCDE): Balance of resource utilization: high staff demand in the early stage (7 people), low in the later stage (3 people), poor balance; Time span: Tasks are distributed relatively evenly and have good smoothness; Priority satisfaction: High priority task A is ranked first, but C is delayed, which is not ideal.

[0059] S6: Predict the remaining life of the pipeline; like Figure 6As shown, the pipeline remaining life prediction process in step S6 is as follows: First, based on the historical data of equipment performance, environmental factors and maintenance records, a multidimensional state tensor of the pipeline system is constructed. The state tensor contains performance indicator dimensions (such as wall thickness loss rate, surface roughness, stress state, crack length), environmental factor dimensions (such as operating temperature, medium corrosiveness, soil resistivity, vibration intensity) and maintenance record dimensions (such as maintenance frequency, maintenance quality score, replacement component type, maintenance response time). Secondly, nonlinear features are extracted through high-order statistical analysis, and the third-order cross moment is calculated:

[0060] For example, the third-order moment of wall thickness loss rate, temperature and maintenance frequency is 0.83, the third-order moment of surface roughness, corrosion and maintenance quality is 0.67, and the third-order moment of stress state, vibration intensity and response time is 0.92. Again, the kernel method is used to reduce the dimension of the features, and a feature extraction model based on tensor decomposition is constructed to analyze the potential evolution law of the pipeline system. Finally, the attention mechanism and Bayesian reasoning method are combined to generate a prediction report containing the estimated value of remaining life, credible interval and key influencing factors.

[0061] S6.1: Construct a multidimensional state tensor of the pipeline system using historical data on equipment performance, environmental factors, and maintenance records; S6.2: extracting nonlinear features from the multidimensional state tensor; S6.3: performing dimensionality reduction processing on the extracted nonlinear features using a kernel method; S6.4: Based on the features after dimensionality reduction, construct a feature extraction model based on tensor decomposition; S6.5: Analyze and mine the potential evolution law of the pipeline system using the feature extraction model to predict the remaining life; S6.6: Calculate the weights of the impact of key time points in the pipeline based on the attention mechanism; S6.7: Use Bayesian inference methods to assess the confidence intervals of predictions; S6.8: Generate a forecast report containing remaining life estimates, credibility analysis, and key influencing factors.

[0062] Specifically, In step S6.1, the construction of the multidimensional state tensor requires the systematic integration of various historical data. First, data is collected and organized from the equipment performance dimension, including key performance indicators such as wall thickness loss rate (unit: mm / year), surface roughness (unit: μm), stress state (unit: MPa) and crack length (unit: mm). Second, data is sorted from the environmental factor dimension, including environmental parameters such as operating temperature (unit: ℃), medium corrosivity (pH value), soil resistivity (unit: Ω·m) and vibration intensity (unit: g). Third, data is summarized from the maintenance record dimension, including maintenance information such as maintenance frequency (times / year), maintenance quality score (1-10 points), replacement component type and maintenance response time (hours). By organizing these data into a third-order tensor T(i,j,k), where i represents performance indicators (4 features), j represents environmental factors (4 features), and k represents maintenance records (4 features), a multidimensional description of the pipeline system state is achieved.

[0063] The nonlinear feature extraction process in step S6.2 requires the use of high-order statistical analysis methods. First, a high-order moment analysis is performed on the constructed state tensor to calculate the third-order cross moment: , where Xi, Yj, and Zk represent characteristic variables of different dimensions, and μi, μj, and μk are the corresponding means. For example, by calculation, the third-order moment of the wall thickness loss rate, temperature, and maintenance frequency is 0.83, the third-order moment of the surface roughness, corrosiveness, and maintenance quality is 0.67, and the third-order moment of the stress state, vibration intensity, and response time is 0.92. Secondly, based on the results of high-order moment analysis, the nonlinear correlation between features is explored to identify key influencing factors. It should be noted that this high-order statistical analysis can effectively capture complex dependencies that are difficult to discover using traditional second-order statistical methods.

[0064] Regarding the feature dimensionality reduction processing in step S6.3, it is necessary to use the kernel method to perform dimensionality reduction mapping of high-dimensional features. First, select an appropriate kernel function to map the data in the original high-dimensional feature space to the low-dimensional feature space. In specific implementation, the radial basis kernel function (RBF) or polynomial kernel function can be used to select the most suitable kernel function type according to the characteristics of the data. Secondly, the nonlinear dimensionality reduction is achieved through the kernel principal component analysis (KPCA) method to retain the essential characteristics of the data while reducing the computational complexity. Thirdly, the validity of the dimensionality reduction results is verified to ensure that the features after dimensionality reduction can maintain the main information of the original data.

[0065] The process of constructing the feature extraction model based on tensor decomposition in step S6.4 requires systematic consideration. First, the tensor decomposition method is used to further analyze the features after dimensionality reduction. Methods such as CP decomposition (CANDECOMP / PARAFAC decomposition) or Tucker decomposition can be selected. Secondly, the potential factors affecting the system state are identified through the factor matrix obtained by decomposition. Thirdly, a feature reconstruction mechanism is established to ensure that the model can accurately reflect the evolution characteristics of the system state. In addition, the computational efficiency of the model needs to be considered, and the processing speed can be improved by parallel computing and other methods.

[0066] In step S6.5, the analysis and mining of the potential evolution laws of the pipeline system is the key to achieving accurate life prediction. First, based on the output of the feature extraction model, a state evolution trajectory analysis method is established. By tracking the changing trends of key features, the laws of system performance degradation are identified. Secondly, combined with historical data, a degradation rate prediction model is constructed. This model needs to consider the coupling effect of multiple influencing factors, such as the impact of environmental factors on material properties, the improvement of system status by maintenance activities, etc. Thirdly, based on the results of the evolution law analysis, a remaining life prediction model is established.

[0067] The application of the attention mechanism in step S6.6 needs to focus on the impact of key time points. First, a multi-head attention mechanism is designed to focus on state changes at different time scales. Second, through self-attention calculation, the importance of state changes at different time points is evaluated, such as the impact of key events such as major repairs and replacement of key components in the pipeline system. Third, based on the attention weight, the contribution of data in different time periods in the prediction model is dynamically adjusted.

[0068] For the Bayesian reasoning process in step S6.7, a scientific uncertainty quantification framework needs to be established. First, a parameter prior distribution model is constructed, which needs to fully consider historical data and expert experience. Secondly, through the Bayesian update process, the model parameters are continuously corrected to improve the accuracy of the prediction. Thirdly, the credible interval of the prediction result is calculated to provide a reliable confidence assessment for decision-making. Exemplarily, the prediction result can be expressed as "the predicted remaining life is 14±1.5 years (95% confidence interval)".

[0069] Finally, in the process of generating the prediction report in step S6.8, it is necessary to systematically integrate the results of various analyses. First, summarize the prediction results of the remaining life, including point estimates and interval estimates. Second, analyze the credibility of the prediction results, including quantitative evaluation of various sources of uncertainty, such as the impact of temperature fluctuations (±0.8 years), maintenance quality impact (±0.5 years), environmental corrosion impact (±0.2 years), etc. Third, identify and summarize the key factors affecting life prediction to provide a basis for subsequent maintenance decisions. In addition, the prediction report should also include model performance evaluation results, update records, and optimization suggestions to support the full life cycle management of the pipeline system.

[0070] Assume that the embodiment of the present application focuses on a section of a main transport pipeline, which has the following characteristic data dimensions: Performance indicator dimensions: Wall thickness loss rate (unit: mm / year); Surface roughness (unit: μm); Stress state (unit: MPa); Crack length (unit: mm); Environmental factors: Working temperature (unit: ℃); Corrosivity of the medium (pH value); Soil resistivity (unit: Ω·m); Vibration intensity (unit: g); Maintain record dimensions: Maintenance frequency (times / year); Repair quality rating (1-10 points); Replacement part type; Maintenance response time (hours); The process of building a tensor estimation model for this data is as follows: Step 1: Build the state tensor The data of the above three dimensions are organized into a third-order tensor T(i,j,k), where: i represents the performance index (4 features); j represents environmental factors (4 characteristics); k represents maintenance records (4 features); For example, the tensor element T(1,2,3) at a certain point in time may represent the state association of the replacement part type (valve assembly) under the condition of corrosive media (pH=6.5) with the wall thickness loss rate (0.15mm / year).

[0071] Step 2: Calculate implicit high-order moments Perform high-order statistical analysis on the tensor to extract nonlinear features: For example, to compute the third-order cross moment:

[0072] Specific numerical examples: The third moment of wall thickness loss rate with temperature and maintenance frequency: 0.83; Third moment of surface roughness, corrosion and repair quality: 0.67; The third moment of stress state, vibration intensity and response time: 0.92; Step 3: Implementation of life prediction model Build a time series forecasting model based on historical data: At t=0: the wall thickness loss rate is 0.15mm / year, and the remaining life is predicted to be 15 years; At t=1 year: the measured loss rate is 0.17mm / year, and the revised forecast is 14 years; At t=2 years: the measured loss rate is 0.16mm / year, and it is predicted to last for 14 years; Introducing uncertainty quantification: Prediction value interval: 14 years ± 1.5 years (95% confidence interval); Main influencing factors: Temperature fluctuation influence: ±0.8 years; Maintenance quality impact: ±0.5 years; Environmental corrosion influence: ±0.2 years; Step 4: Dynamically update the example Assuming a major overhaul in year 3: Predicted remaining life before update: 13.5 years; The wall thickness loss rate was detected to be reduced to 0.14mm / year; Surface roughness improved by 20%; Updated estimated remaining lifespan: 15.2 years; Corrected confidence interval: ±1.2 years; Through this specific example, we can see how the implicit high-order moment tensor estimation method captures the complex nonlinear relationships in the pipeline system and achieves more accurate life prediction through these relationships. This method not only considers the influence of a single factor, but also captures the interaction between multiple factors through high-order moment analysis, thereby providing more reliable prediction results. At the same time, the dynamic update mechanism ensures that the prediction model can adapt to changes in the system state in a timely manner and maintain the accuracy of the prediction.

[0073] In addition to the above methods, it also includes: Build predictive models; Obtain deviation results by comparing predicted values ​​with actual observed values; According to the deviation result, the parameters of the prediction model are dynamically adjusted using an adaptive correction algorithm; The prediction model is used to predict the remaining life of the pipeline and generate a prediction model evaluation report including model performance indicators, update records and optimization suggestions.

[0074] The construction of a prediction model requires systematic consideration of multiple key links. First, based on the previous feature extraction and data analysis results, a prediction model framework containing multiple sub-modules is constructed. The framework needs to include core components such as data preprocessing module, feature selection module, prediction algorithm module and result output module. Secondly, in the selection of prediction algorithms, an integrated learning method is used to fuse the prediction results of multiple basic models to improve the stability and reliability of the prediction. In specific implementation, algorithms such as random forest and gradient boosting tree can be selected as basic models, and the final prediction results can be obtained through weighted combination. In addition, the prediction model also needs to have online learning capabilities and be able to continuously optimize and update model parameters based on new data.

[0075] The analysis process of forecast deviation requires the establishment of a scientific evaluation mechanism. First, collect the forecast values ​​and actual observation values ​​within a certain period of time to construct a forecast effect evaluation data set. Secondly, use multi-dimensional evaluation indicators to quantitatively analyze the forecast deviation, including statistical indicators such as mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE). Thirdly, perform a time series analysis on the forecast deviation to identify the changing trend and periodic characteristics of the deviation. It should be noted that the deviation analysis should also consider the forecast performance under different operating conditions to comprehensively evaluate the adaptability of the model.

[0076] The design of the adaptive correction algorithm is directly related to the dynamic optimization effect of the prediction model. First, based on the prediction deviation analysis results, a parameter sensitivity analysis model is constructed to identify the key parameters that have a greater impact on the prediction results. Secondly, a parameter optimization algorithm based on gradient descent is designed to continuously adjust the model parameters by minimizing the prediction error. Specifically, when it is detected that the prediction deviation exceeds the preset threshold, the system automatically triggers the parameter adjustment process. Thirdly, an adaptive learning rate mechanism is introduced to dynamically adjust the parameter update step size according to the convergence situation during the optimization process, which not only ensures the stability of the optimization, but also improves the convergence speed.

[0077] Finally, in the application of the prediction model and the generation of the evaluation report, it is necessary to systematically integrate the various analysis results. First, use the optimized prediction model to predict the remaining life of the pipeline and give the confidence interval of the prediction result. Secondly, generate a comprehensive prediction model evaluation report, which should include the following key contents: model performance indicators (such as prediction accuracy, stability evaluation, etc.), parameter update records (including update time, trigger conditions, adjustment range, etc.), model optimization suggestions (such as feature selection optimization, algorithm improvement suggestions, etc.). In addition, the evaluation report should also include an analysis of the limitations of the model application and suggestions for improvement, providing a reliable theoretical basis for maintenance decisions of the pipeline system.

[0078] The present application also provides a petrochemical pipeline full life cycle detection and evaluation device, comprising a processor and a memory, wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor can perform each step of the above petrochemical pipeline full life cycle detection and evaluation method.

[0079] The embodiment of the present disclosure also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the petrochemical pipeline full life cycle detection and evaluation described in the above method embodiment are executed. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0080] In addition, an embodiment of the present disclosure also provides a computer program product, which stores a computer program. When the computer program is executed by a processor, the steps of the petrochemical pipeline full life cycle detection and evaluation method provided in any of the above embodiments of the present disclosure are executed. For details, please refer to the above method embodiments, which will not be repeated here.

[0081] The computer program product may be implemented in hardware, software or a combination thereof. In an optional embodiment, the computer program product is embodied as a computer storage medium, which may be a volatile or non-volatile computer-readable storage medium. In another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK) and the like.

[0082] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the above-described devices and devices can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed devices, equipment and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0083] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0084] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0085] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0086] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the aforementioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be based on the protection scope of the claims.

[0087] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A petrochemical pipeline full life cycle detection and evaluation method, characterized in that: include: Build a pipeline inspection digital twin model including geometric models and material properties; Collect pipeline inspection data through the deployed multi-source sensor network; Inputting the detection data into the pipeline detection digital twin model, and performing feature extraction and analysis of pipeline status in the pipeline detection digital twin model, so as to generate a pipeline status feature evolution report; According to the pipeline state characteristic evolution report, a heterogeneous tree network is constructed, and a hierarchical evaluation is performed on the pipeline system based on the hierarchical structure of the heterogeneous tree network to generate a hierarchical evaluation result; optimizing a maintenance plan based on the tiered evaluation results; Based on the optimized maintenance plan, the remaining life of the pipeline is predicted.

2. The method according to claim 1, characterized in that The pipeline inspection digital twin model including geometric model and material properties is constructed, including: Construct the geometric model of the physical piping system based on the engineering drawings and field measurement data of the piping system; According to the physical characteristic parameters of the pipeline system, a material property model including elastic modulus, Poisson's ratio and thermal expansion coefficient is established; Based on the geometric model and the material property model, a pipeline system operation parameter model including pressure, temperature and flow rate is established; Constructing a coordinate mapping relationship between the physical space and the virtual space according to the geometric model, the material property model and the operation parameter model; Based on the coordinate mapping relationship, the state parameter mapping of each component of the pipeline system is established to realize the real-time data synchronization mechanism between the physical system and the virtual model.

3. The method according to claim 1, characterized in that The method of collecting pipeline detection data through the deployed multi-source sensor network includes: Deploy a multi-source sensor network at key nodes of the pipeline system, the sensor network including ultrasonic probes, electromagnetic sensor arrays, ray detection equipment, vibration sensors and temperature sensors; Using the multi-source sensor network to collect detection data of the pipeline system; Performing data cleaning on the collected detection data; Performing signal optimization on the cleaned detection data; Timestamps and spatial location tags are added to the optimized data to generate a standardized detection data set.

4. The method according to claim 3, characterized in that The pipeline state feature extraction and analysis includes: Based on the standardized inspection data set, a multidimensional feature extraction framework including geometric features, physical features and dynamic features is constructed in the pipeline inspection digital twin model; Using the multidimensional feature extraction framework, the geometric features of the pipeline system are extracted by a morphological analysis method, the physical features of the pipeline system are extracted by a stress-strain analysis model, and the dynamic features of the pipeline system are extracted by a time-frequency analysis technique; According to the extracted geometric features, physical features and dynamic features, a feature hierarchy model is constructed, and a deep learning feature fusion algorithm is used to fuse the geometric features, physical features and dynamic features to generate fused features; Generate a pipeline system state feature vector based on the fusion feature, and track the dynamic change law of the state feature vector using a time series analysis method; According to the dynamic change rules, a feature evolution report including feature change trends and abnormal warning information is constructed.

5. The method according to claim 1, characterized in that The heterogeneous tree network is a heterogeneous tree network with uncertainty in edge weights, and the construction of the heterogeneous tree network includes: Based on the characteristic change rules in the characteristic evolution report, a network topology structure reflecting the physical connection relationship and logical dependency relationship of the pipeline system is constructed; Design edge weight calculation model based on the reliability, importance and timeliness of detection data; Using the edge weight calculation model, an adaptive weighting method is used to calculate the weight of each edge; A dynamic adjustment factor of the edge weight is generated, and based on the dynamic adjustment factor, the edge weight is updated in real time.

6. The method according to claim 1, characterized in that The optimizing the maintenance plan based on the hierarchical evaluation result comprises: Based on the hierarchical evaluation results, establish an initial maintenance plan including maintenance task priorities, execution time and resource requirements; Using the k-swap neighborhood generator, candidate maintenance plans are generated by performing task swap operations in the time and space dimensions. Construct a smooth optimization algorithm based on k-swap neighborhood; Using the smoothing optimization algorithm, performing iterative optimization on candidate maintenance solutions; Based on maintenance cost, equipment availability and risk control, a multi-objective evaluation system is constructed to evaluate the optimization results; An optimized maintenance plan is generated according to the evaluation results.

7. The method according to claim 1, characterized in that The predicted remaining life of the pipeline includes: Utilize historical data on equipment performance, environmental factors, and maintenance records to construct a multi-dimensional state tensor of the pipeline system; Extracting nonlinear features in the multidimensional state tensor; Performing dimensionality reduction processing on the extracted nonlinear features using a kernel method; According to the features after dimensionality reduction, a feature extraction model based on tensor decomposition is constructed; The feature extraction model is used to analyze and mine the potential evolution law of the pipeline system to predict the remaining life.

8. The method according to claim 1, characterized in that The predicted remaining life of the pipeline includes: Analyze dependencies in pipeline time series data using long short-term memory networks; The first weight calculation is performed based on the influence of the attention mechanism on the key time points of the pipeline; Using a Bayesian inference method, based on the dependency relationship and the first weight, a credible interval of the prediction result is evaluated; Based on the assessment results, a forecast report is generated containing remaining life estimates, credibility analysis and key influencing factors.

9. The method according to claim 1, characterized in that: The predicted remaining life of the pipeline includes: Build predictive models; Obtain deviation results by comparing predicted values ​​with actual observed values; According to the deviation result, the parameters of the prediction model are dynamically adjusted using an adaptive correction algorithm; The prediction model is used to predict the remaining life of the pipeline and generate a prediction model evaluation report including model performance indicators, update records and optimization suggestions.

10. A petrochemical pipeline full life cycle detection and evaluation equipment, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the petrochemical pipeline full life cycle detection and evaluation method as described in any one of claims 1 to 9.

Images