Method and system for protecting the inner wall of a pipeline from erosion

Abstract

This invention relates to the field of pipeline anti-corrosion technology, and particularly to a method and system for anti-corrosion of pipeline inner walls. The method first determines the erosion influence range based on actual operating parameters, and obtains initial erosion state data of the pipeline through preprocessing detection and analysis. Second, it models the influence range and state data to construct an erosion evolution relationship model. Then, it identifies the critical erosion thickness for pipeline failure to obtain a critical threshold, and calculates the cumulative erosion amount on the pipeline inner wall. Subsequently, it maps and encodes the critical threshold and cumulative erosion amount using parameters to generate an anti-corrosion evaluation vector. Finally, it inputs this vector into a pre-set anti-corrosion process decision model, intelligently outputting a targeted anti-corrosion construction plan. This invention, by quantifying the nonlinear evolution process of erosion, achieves a closed loop from accurate state assessment to intelligent decision-making on maintenance plans, effectively solving problems such as inaccurate prediction and delayed maintenance in traditional methods, significantly reducing operation and maintenance costs and extending pipeline service life.

Description

Technical Field

[0001] This invention relates to the field of pipeline anti-corrosion technology, specifically to a method and system for anti-corrosion of the inner wall of a pipeline. Background Technology

[0002] In flue gas desulfurization (FGD) systems of thermal power plants, slurry circulation pipelines widely employ small-diameter, long-distance layouts to effectively transport desulfurizing agents (such as limestone slurry). Due to the high solids content, strong corrosiveness, and high flow velocity of the transported medium, the inner wall of the pipeline is subjected to a complex environment of coupled multiphase flow erosion and chemical corrosion, making it highly susceptible to severe erosion damage. This is particularly pronounced in areas with intense local flow field disturbances, such as elbows, reducers, and tees, where the erosion rate is significantly accelerated. The small-diameter pipelines, with their small cross-section and high flow velocity, further exacerbate the inner wall wear problem. Long-distance laying leads to difficult maintenance and high replacement costs. Once a perforation leak occurs, it not only affects desulfurization efficiency but may also trigger unplanned unit shutdowns, seriously threatening the safe and stable operation of the power plant.

[0003] Currently, conventional pipeline protection measures mainly rely on the selection of wear-resistant materials (such as duplex steel, plastic-lined pipes, or ceramic composite pipes) or regular manual thickness measurement and maintenance. However, these methods suffer from high costs, poor adaptability, and a lack of accurate predictive capabilities. Existing corrosion assessment methods are mostly based on empirical formulas or simple linear models, which are insufficient to accurately reflect the nonlinear evolution of corrosion under complex operating conditions, and even more so to quantify and classify the corrosion risk of different pipe sections. Furthermore, the lack of a dynamic assessment mechanism for corrosion development trends and remaining service life leads to outdated maintenance strategies, making it difficult to achieve preventative maintenance and precise corrosion prevention. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a method and system for anti-corrosion of the inner wall of a pipeline.

[0005] In a first aspect, embodiments of the present invention provide a method for resisting corrosion of the inner wall of a pipe, the method comprising: Based on the actual operating parameters of the desulfurization pipeline, the erosion influence range of the pipeline inner wall is determined. The erosion pretreatment detection and operating parameter analysis are carried out on the target small-diameter long-distance pipeline to obtain the initial erosion state data of the target pipeline. A relationship model is performed on the erosion influence range and the initial erosion state data of the target pipeline to obtain an erosion evolution relationship data model. The critical erosion thickness for pipeline failure is identified by performing the erosion evolution relationship data model to obtain the critical erosion threshold for the pipeline. The cumulative erosion amount is calculated by performing a cumulative erosion calculation on the erosion evolution relationship data model to obtain the cumulative erosion amount of the inner wall of the pipe; The critical erosion threshold of the pipeline and the cumulative erosion amount of the inner wall of the pipeline are parameter-mapped and encoded to generate a target pipeline erosion resistance evaluation vector. The anti-corrosion evaluation vector of the target pipeline is input into a pre-set anti-corrosion process decision model to output an anti-corrosion process scheme, thereby obtaining the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline.

[0006] Optionally, the process involves determining the erosion influence range of the pipeline inner wall based on the actual operating parameters of the desulfurization pipeline, performing erosion pretreatment detection and operating parameter analysis on the target small-diameter long-distance pipeline, and obtaining the initial erosion state data of the target pipeline, including: Based on the actual operating parameters of the desulfurization system of the thermal power plant, the erosion influence range of the inner wall of the pipeline is determined, and the erosion influence range is divided into multiple erosion influence gradient sub-ranges according to the medium flow rate and gypsum solid content. Based on the multiple erosion influence gradient sub-intervals, segmented erosion detection is performed on the target small-diameter long-distance pipeline, and multiple wall thickness parameters, medium parameters and material parameters of different pipe sections of the target pipeline are collected. The degree of abrasion is calculated for each of the multiple wall thickness parameters, medium parameters, and material parameters to obtain the initial abrasion state data for each parameter. Calculate the erosion influence weight of each parameter on pipeline erosion, and based on the erosion influence weight, perform weighted calculation on the initial erosion state data to obtain the weighted erosion state data of each parameter; The weighted erosion state data is fused to obtain the initial erosion state data of the target pipeline.

[0007] Optionally, the step of modeling the relationship between the erosion influence range and the initial erosion state data of the target pipeline to obtain an erosion evolution relationship data model includes: Based on the multiple erosion influence gradient sub-intervals, the initial erosion state data of the target pipeline is segmented to obtain the erosion state interval data corresponding to each erosion influence gradient sub-interval. The wear state interval data is discretized to obtain discretized wear state data points, and the wear influence intensity of each wear influence gradient sub-interval is discretized to obtain discretized wear influence intensity data points. By using a pre-set Markov model of erosion evolution, a quantitative analysis of the relationship between the discretized erosion state data points and the discretized erosion influence intensity data points is performed to obtain a data model of erosion evolution relationship.

[0008] Optionally, the step of identifying the critical erosion thickness for pipeline failure based on the erosion evolution relationship data model to obtain the critical erosion threshold for the pipeline includes: Curve fitting is performed on the erosion evolution relationship data model to generate the corresponding erosion thickness evolution relationship curve; The erosion thickness evolution curve is divided into a safe domain, an early warning domain, and a failure domain to obtain the corresponding safe domain, early warning domain, and failure domain. Based on the safety domain, the early warning domain, and the failure domain, the critical erosion threshold critical point is identified on the erosion thickness evolution curve to obtain the critical erosion critical point. The critical wear threshold is mapped to the wall thickness value to generate the corresponding critical wear threshold of the pipeline.

[0009] Optionally, the step of calculating the cumulative erosion amount on the erosion evolution relationship data model to obtain the cumulative erosion amount of the pipe inner wall includes: The corrosion trend of the corrosion evolution relationship data model is detected to obtain the pipeline corrosion trend index; Based on the pipeline erosion trend index, create an erosion amount calculation function corresponding to the erosion evolution relationship data model; Based on the erosion calculation function, the cumulative erosion of the inner wall of the pipeline is calculated using the erosion evolution relationship data model, and the cumulative erosion of the inner wall of the pipeline is obtained.

[0010] Optionally, the step of parameter mapping and encoding the critical erosion threshold of the pipeline and the cumulative erosion amount of the pipeline inner wall to generate a target pipeline erosion resistance evaluation vector includes: The K-Bins algorithm was used to calculate the cumulative erosion amount of the inner wall of the pipe using discrete intervals to obtain the target K value. Obtain the maximum and minimum values ​​of the cumulative abrasion amount on the inner wall of the pipe, and determine the corresponding discretization interval range based on the maximum and minimum values; K target interval ranges are determined based on the target K value and the discretized interval range, and the cumulative erosion amount of the inner wall of the pipe is mapped based on the K target interval ranges to obtain the cumulative erosion amount of the pipe corresponding to each target interval range. Multiple pipe erosion features are generated based on the cumulative pipe erosion amount corresponding to each target interval range. The multiple pipeline erosion characteristics and the pipeline critical erosion threshold are normalized and vectorized to obtain the target pipeline erosion resistance evaluation vector.

[0011] Optionally, the step of inputting the target pipeline anti-corrosion evaluation vector into a pre-set anti-corrosion process decision model to output an anti-corrosion process scheme, thereby obtaining the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline, includes: The anti-corrosion evaluation vector of the target pipeline is input into a preset anti-corrosion process decision model, wherein the anti-corrosion process decision model includes a first bidirectional gated loop unit, a second gated loop unit, and a double-layer fully connected output layer. Obtain the anti-corrosion process standard parameters for small-diameter long-distance desulfurization pipelines in thermal power plants, and construct the corresponding standard anti-corrosion requirement vector based on the anti-corrosion process standard parameters. The target pipeline anti-corrosion evaluation vector is input into the first bidirectional gated loop unit for feature extraction to obtain the first corrosion feature vector, and the standard anti-corrosion requirement vector is input into the second gated loop unit for feature extraction to obtain the second process requirement feature vector. The first wear-resistant feature vector and the second process requirement feature vector are fused to obtain a fused anti-wear-resistant feature vector. The anti-corrosion construction process parameters are calculated and output through the dual-layer fully connected output layer, and the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline is determined based on the anti-corrosion construction process parameters.

[0012] Secondly, embodiments of the present invention also provide a pipe inner wall anti-corrosion system, the system comprising: The initial working condition inspection module is used to determine the erosion influence range of the inner wall of the desulfurization pipeline based on the actual working condition parameters of the pipeline, and to perform erosion pretreatment detection and working condition parameter analysis on the target small-diameter long-distance pipeline to obtain the initial erosion state data of the target pipeline. The evolution modeling module is used to model the relationship between the erosion influence range and the initial erosion state data of the target pipeline, and obtain an erosion evolution relationship data model. The threshold identification module is used to identify the critical erosion thickness of the pipeline failure from the erosion evolution relationship data model, and obtain the critical erosion threshold of the pipeline. The cumulative erosion calculation module is used to calculate the cumulative erosion amount on the erosion evolution relationship data model to obtain the cumulative erosion amount of the inner wall of the pipeline. The vector encoding module is used to perform parameter mapping encoding on the critical erosion threshold of the pipeline and the cumulative erosion amount of the inner wall of the pipeline to generate a target pipeline erosion resistance evaluation vector. The scheme decision module is used to input the anti-corrosion evaluation vector of the target pipeline into a preset anti-corrosion process decision model to output the anti-corrosion process scheme, thereby obtaining the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline.

[0013] Thirdly, embodiments of the present invention also provide a pipe inner wall anti-corrosion device, the device comprising: a memory, a processor, and a pipe inner wall anti-corrosion program stored in the memory and executable on the processor, the pipe inner wall anti-corrosion program being configured to implement the steps of the pipe inner wall anti-corrosion method as described above.

[0014] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a pipe inner wall anti-corrosion program, wherein when the pipe inner wall anti-corrosion program is executed by a processor, it implements the steps of the pipe inner wall anti-corrosion method as described above.

[0015] This invention provides a method for combating erosion on the inner wall of pipelines. This method determines the erosion influence range based on actual operating parameters (such as flow velocity and solids content) and performs gradient division. It then establishes a data model of erosion evolution by combining pre-processed detection data, overcoming the limitations of traditional empirical formulas or simple linear models. This method accurately reflects the nonlinear evolution of erosion in complex multiphase flow environments, enabling refined calculation of the initial state and cumulative erosion amount of the pipeline inner wall, significantly improving the accuracy of the evaluation results. By performing curve fitting on the model and dividing it into reliable, early warning, and failure domains, this invention can intelligently identify the critical erosion threshold for pipeline failure. This data-driven threshold identification mechanism is more adaptable to the actual operating conditions of different pipe sections compared to fixed standard values, effectively avoiding over-maintenance or under-maintenance, and providing a scientific basis for preventing sudden perforation leaks. By mapping the critical erosion threshold and cumulative erosion amount into an evaluation vector and inputting it into a pre-set anti-erosion process decision model (including deep learning structures such as bidirectional gated loop units), an automated closed loop from data analysis to solution generation is achieved. This method can automatically match the optimal construction process parameters, such as material selection, coating thickness, and repair method, based on the specific erosion characteristics of the pipeline, and output targeted anti-erosion construction solutions, solving the problems of high subjectivity and low efficiency in traditional manual decision-making. Addressing the difficulties in maintenance and high replacement costs of small-diameter, long-distance pipelines, this method transforms maintenance work from reactive repair to proactive prevention through accurate remaining life prediction and preventative maintenance strategy formulation. This not only significantly reduces the risk of unplanned outages but also optimizes the allocation of maintenance resources, significantly reduces the overall life-cycle operation and maintenance costs, and extends the overall service life of desulfurization system pipelines. The method fully considers the characteristics of high solids content, strong corrosiveness, and large flow velocity variations in the media of thermal power plant desulfurization systems. Through multi-dimensional parameter weighted fusion and discretization processing, it can effectively cope with operating condition fluctuations at different operating stages and different pipeline locations, demonstrating strong engineering applicability and promotional value. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of an embodiment of the pipe inner wall anti-erosion method of the present invention; Figure 2 This is a structural block diagram of an embodiment of the anti-abrasion system for the inner wall of a pipe according to the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] Unless otherwise specifically stated, the technical or scientific terms used in the embodiments of this invention should be understood in their ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms "comprising" or "including," as used in the embodiments of this invention, do not limit the shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof mentioned, nor do they exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof, or the inclusion of these.

[0020] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale, and techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail; however, where appropriate, the illustrated techniques, methods, and apparatus should be considered part of the specification. In all the examples shown and discussed herein, any other specific example may have different values. It should be noted that similar symbols and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0021] In the description of the embodiments of the present invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of the present invention, as well as the features of different embodiments or examples.

[0022] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0023] Reference Figure 1 , Figure 1 This is a schematic flowchart of an embodiment of the pipe inner wall anti-corrosion method of the present invention, which presents an embodiment of the pipe inner wall anti-corrosion method of the present invention.

[0024] In one embodiment, the method for preventing erosion of the inner wall of the pipe includes: Step S100: Based on the actual operating parameters of the desulfurization pipeline, determine the erosion influence range of the pipeline inner wall, conduct erosion pretreatment detection and operating parameter analysis on the target small-diameter long-distance pipeline, and obtain the initial erosion state data of the target pipeline.

[0025] The actual operating parameters of the desulfurization pipeline can be a set of key physical and chemical variables that affect pipeline erosion during the operation of the desulfurization system. These parameters can characterize the multiphase flow environment of different pipe sections and support the dynamic division of the erosion influence zone. In this embodiment, the actual operating parameters of the desulfurization pipeline can include, but are not limited to, one or more of the following: flow velocity parameters, solids content parameters, slurry pH value, and temperature parameters. The target small-diameter long-distance pipeline can be a small-diameter, long-path pipeline used to transport high-solids-content limestone slurry in the flue gas desulfurization system of a thermal power plant. It can be used as the specific object for implementing this method. Its structural characteristics determine the high risk of wear and the difficulty of maintenance. The nominal diameter corresponding to the small diameter can be 60~325mm, and the long path refers to a single continuous laying length ≥100 meters. For example, the target small-diameter long-distance pipeline can include elbow sections, straight sections, and transition sections. The erosion pretreatment detection can be an initial state data acquisition process of the inner wall of the target pipeline. This can be used to provide a quantitative basis for the initial erosion state of the pipeline, providing real benchmark data for subsequent modeling. Furthermore, erosion pretreatment detection can acquire wall thickness distribution and surface morphology information through non-destructive testing techniques such as ultrasonic thickness measurement, endoscopic imaging, or electromagnetic detection. Operating condition parameter analysis involves normalizing, filtering, and extracting features from the collected actual operating parameters. This can be used to generate standardized operating condition feature vectors suitable for modeling, reflecting the intensity of local flow field disturbances. In an exemplary embodiment, operating condition parameter analysis can employ time-series sliding window statistics, outlier removal, and multi-dimensional parameter fusion algorithms for data cleaning and characterization. The initial erosion state data of the target pipeline can be a comprehensive dataset of the pipeline's initial wear condition obtained after erosion pretreatment detection and operating condition parameter analysis. This dataset can be used as the initial input condition for the erosion evolution relationship data model, ensuring that the model's starting point matches reality.

[0026] Determining the erosion-affected area of ​​the desulfurization pipeline based on its actual operating parameters can be achieved by segmenting the pipeline according to the spatial distribution characteristics of parameters such as flow velocity and solid content. Furthermore, this operation can be implemented by using the K-means clustering algorithm for unsupervised segmentation of the pipeline's operating parameters, or by regularizing interval segmentation based on preset flow field disturbance thresholds (such as Reynolds number mutation points). This allows for refined identification of high-risk areas, supporting differentiated modeling and evaluation. For erosion pretreatment detection and operating parameter analysis of the target small-diameter, long-distance pipeline, erosion pretreatment detection and operating parameter analysis can be performed simultaneously with internal wall condition detection and operational parameter acquisition, followed by data fusion processing. In a specific embodiment, this operation can be achieved by deploying an online ultrasonic thickness gauge array linked to the DCS system data interface for real-time detection and parameter synchronization, or by using an intelligent crawling robot equipped with multimodal sensors for offline comprehensive detection during periodic shutdowns. This allows for the acquisition of high-fidelity, spatiotemporally aligned initial state data, improving the reliability of subsequent models.

[0027] Step S200: Model the relationship between the erosion influence range and the initial erosion state data of the target pipeline to obtain the erosion evolution relationship data model.

[0028] The erosion influence zone can be a regional unit with similar erosion driving mechanisms, divided axially or spatially along the pipeline based on actual operating parameters. This allows for precise location of local high-risk areas, avoiding biases caused by global uniform assessments. In this embodiment, the erosion influence zone can include, but is not limited to, one or more of the following: high scouring disturbance zone, steady-state flow zone, and corrosion-dominant zone. The erosion evolution relationship data model can be a data-driven model describing the nonlinear mapping relationship of pipeline inner wall erosion amount with time and operating conditions. This model can accurately characterize the nonlinear evolution law of erosion under the coupled action of multiphase flow scouring and chemical corrosion. Furthermore, the erosion evolution relationship data model can be constructed using machine learning regression or deep time series modeling, with initial erosion state data and historical operating condition sequences as input. In an exemplary embodiment, the erosion evolution relationship data model can include a time series erosion model based on LSTM, a nonlinear regression model based on random forest, and a spatial correlation erosion model based on graph neural networks. In addition, the erosion evolution relationship data model can receive the erosion influence zone and initial erosion state data as input and output basic functions for critical threshold identification and cumulative erosion amount calculation.

[0029] Modeling the relationship between the erosion-affected regions and the initial erosion state data of the target pipeline can be achieved by using the region identifier and the initial state as input to construct a nonlinear function of the erosion amount evolving over time. For example, this operation can be implemented by independently training an LSTM sub-model within each erosion-affected region to form a partitioned modeling architecture, or by constructing a graph neural network to connect pipeline nodes by region and embed the initial state as node features. This allows for the establishment of a erosion evolution model that reflects local operating condition differences, breaking through the traditional global linear assumption.

[0030] Step S300: Identify the critical erosion thickness for pipeline failure using the erosion evolution relationship data model to obtain the critical erosion threshold for the pipeline.

[0031] The critical erosion thickness for pipeline failure can be the erosion depth value corresponding to the minimum remaining wall thickness that leads to the loss of pipeline structural integrity. It can be used as a safety boundary criterion to define whether the pipeline has entered a failure risk state. The critical erosion threshold for pipeline failure can be a dynamically determined critical erosion thickness value after curve fitting and domain partitioning from an erosion evolution relationship data model. It can be used to provide a dynamic failure criterion adapted to the differences in operating conditions of different pipe sections, replacing fixed standard values. In a specific embodiment, the critical erosion threshold for pipeline failure can be obtained by dividing the model output into three segments: a reliability domain, a warning domain, and a failure domain, and setting a threshold point in conjunction with a safety margin. Furthermore, the critical erosion threshold for pipeline failure can be generated by an erosion evolution relationship data model and participate in parameter mapping encoding together with the cumulative erosion amount.

[0032] Identifying the critical erosion thickness for pipeline failure using a data model of erosion evolution can be achieved by dividing the erosion-time curve output by the model into domains to determine the critical failure point. Furthermore, this operation can be accomplished by fitting the curve using cubic spline interpolation and combining it with a safety factor to set the starting point of the failure domain, or by introducing Monte Carlo simulation to generate multiple evolution paths and using a 95% confidence lower limit as the critical threshold. This allows for the generation of dynamic critical values ​​adapted to the specific degradation characteristics of the pipe section, avoiding misjudgments caused by fixed thresholds.

[0033] Step S400: Calculate the cumulative erosion amount on the erosion evolution relationship data model to obtain the cumulative erosion amount of the inner wall of the pipeline.

[0034] The cumulative erosion of the pipeline inner wall can be the total material thickness lost due to scouring and corrosion since the pipeline was put into operation. It can be used to reflect the current actual degradation state of the pipeline and is compared with the critical threshold to assess the remaining safety margin. In addition, the cumulative erosion of the pipeline inner wall can be calculated from the cumulative erosion amount and together with the critical erosion threshold constitute the evaluation vector input.

[0035] The cumulative erosion amount can be calculated by integrating or summing the instantaneous erosion rate output by the model along the time dimension. In an exemplary embodiment, this operation can be achieved by using numerical integration (such as the trapezoidal rule) to calculate the area of ​​the erosion rate curve under a continuous operating condition sequence, or by summing the erosion increments output by the model at each discrete time step to obtain the total amount. This allows for accurate quantification of the historical erosion process and supports the assessment of the current state.

[0036] Step S500: Perform parameter mapping encoding on the critical erosion threshold of the pipeline and the cumulative erosion amount of the inner wall of the pipeline to generate the target pipeline erosion resistance evaluation vector.

[0037] The parameter mapping encoding process can be a standardized process that converts the critical erosion threshold and cumulative erosion amount into a structured numerical vector. This can be used to form an input format recognizable by the anti-erosion process decision model, integrating risk status and safety boundary information. Furthermore, parameter mapping encoding can generate feature vectors through normalization, interpolation (e.g., safety margin = threshold - cumulative amount), and dimensional expansion. The target pipeline anti-erosion evaluation vector can be a numerical vector generated by parameter mapping encoding, characterizing the current erosion risk level and remaining lifespan of the pipeline. This vector can be used as direct input to the anti-erosion process decision model, driving the generation of automated solutions. Additionally, the target pipeline anti-erosion evaluation vector can be generated by parameter mapping encoding and input into the anti-erosion process decision model.

[0038] Parameter mapping encoding of the critical erosion threshold and the cumulative erosion amount of the pipeline inner wall can be achieved by converting the two scalar values ​​into a vector representation that includes relative relationships and absolute states. For example, this operation can be implemented by constructing a three-dimensional vector: [cumulative erosion amount, critical threshold, (critical threshold - cumulative amount) / critical threshold], or by using one-hot encoding to distinguish safety level intervals and appending the original numerical values ​​as continuous features. This can generate structured and information-rich input features, which facilitates deep models' understanding of risk states.

[0039] Step S600: Input the target pipeline anti-corrosion evaluation vector into the preset anti-corrosion process decision model to output the anti-corrosion process scheme, and obtain the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline.

[0040] The pre-set anti-wear process decision model can be a pre-trained deep learning model used to output optimal anti-wear construction parameters based on evaluation vectors. This model can be used to achieve automatic mapping from risk assessment to process decision-making, eliminating subjective human judgment. In one specific embodiment, the pre-set anti-wear process decision model can be trained using supervised learning based on historical maintenance records, material performance databases, and expert rules. Furthermore, the pre-set anti-wear process decision model can include a sequence decision model based on bidirectional gated recurrent units, a multi-task process recommendation model based on attention mechanisms, and a dynamic optimization decision model based on reinforcement learning. The bidirectional gated recurrent unit can be a recurrent neural network structure capable of simultaneously capturing forward and backward temporal dependencies, used to model the historical trend and future prediction correlation of wear evolution in the anti-wear process decision model. For example, the bidirectional gated recurrent unit can include a standard BiGRU structure, a BiGRU with attention weights, or a multi-layer stacked BiGRU. A targeted anti-corrosion construction plan can be an automatically generated repair or protection implementation plan, including materials, processes, and parameters, tailored to the specific corrosion condition of a pipe section. It can be used to guide precise anti-corrosion operations on-site, improving maintenance efficiency and targeting. In one exemplary embodiment, the targeted anti-corrosion construction plan may include an internal ceramic coating lining construction plan, a partial weld overlay repair plan, and a priority recommendation plan for overall replacement.

[0041] The anti-corrosion evaluation vector of the target pipeline is input into a pre-set anti-corrosion process decision model to output an anti-corrosion process scheme. This can be achieved by calling a trained deep learning model to generate construction parameter suggestions via forward propagation. Furthermore, this operation can be implemented by outputting multi-task results (such as material type (classification), coating thickness (regression), construction period (classification)) or by using an integrated model structure to predict the probability of different process options and select the scheme with the highest confidence. This enables end-to-end automation from data evaluation to engineering decision-making, improving the scientific nature of the scheme and execution efficiency.

[0042] Taking the preventive maintenance of the bend section of the desulfurization system as an example, the anti-erosion method for the inner wall of the pipeline in this embodiment can be as follows: at the 90° bend of the slurry circulation pipeline at the inlet of the desulfurization tower of a 600MW coal-fired unit, the system collects operating parameters such as flow velocity (8.2–10.5 m / s) and solid content (25–32%) for the past three months. Combined with ultrasonic thickness measurement data, the initial wall thickness reduction rate reaches 12%. After dividing the erosion influence area, this bend is identified as a high scouring disturbance area. The erosion evolution relationship data model predicts that its annual average erosion rate is 0.8 mm / year, and the critical erosion threshold is dynamically set to 3.5 mm (corresponding to a remaining wall thickness of 4.0 mm). The current cumulative erosion amount is 2.1 mm, and an evaluation vector [2.1, 3.5, 0.4] is generated through parameter mapping encoding. Based on this, the anti-corrosion process decision model outputs a target anti-corrosion construction plan of "lining with silicon carbide ceramic coating, thickness ≥ 2.5mm, and arranging the construction window during the next C-level maintenance", avoiding blindly replacing the entire pipeline and saving approximately 380,000 yuan in costs.

[0043] In one embodiment, the erosion influence range of the pipeline inner wall is determined based on the actual operating parameters of the desulfurization pipeline. Erosion pretreatment detection and operating parameter analysis are performed on the target small-diameter long-distance pipeline to obtain the initial erosion state data of the target pipeline, including: Based on the actual operating parameters of the desulfurization system of the thermal power plant, the erosion influence range of the inner wall of the pipeline is determined, and the erosion influence range is divided into gradients according to the medium flow rate and gypsum solid content, generating multiple erosion influence gradient sub-ranges. The medium flow velocity can be the linear velocity of the desulfurization slurry flowing within the pipeline, a key kinetic parameter affecting the intensity of erosion. In an exemplary embodiment, the medium flow velocity can be used as one of the main bases for dividing the erosion influence gradient sub-regions, reflecting the local erosion energy level. Exemplarily, the medium flow velocity can include, but is not limited to, one or more of the following: axial mainstream velocity, secondary flow velocity component, and turbulent pulsation velocity. The gypsum solids content can be the mass concentration or volume fraction of solid particles (mainly gypsum crystals) in the desulfurization slurry. Furthermore, the gypsum solids content can be used to characterize the degree of influence of abrasive particle concentration on the erosion rate, and is an important chemical-physical parameter for dividing the erosion gradient. In a specific embodiment, the gypsum solids content can include, but is not limited to, instantaneous solids content, average solids content, and particle size distribution-weighted solids content. The erosion influence gradient sub-region can be a sub-region with gradient erosion driving intensity, further subdivided based on the combined characteristics of the medium flow velocity and gypsum solids content, on the basis of the original erosion influence region. In this embodiment, the abrasion effect gradient sub-interval can be divided into two-dimensional hierarchical regions by using a bivariate threshold grid or clustering method. Furthermore, the abrasion effect gradient sub-interval can be used to achieve a refined spatial characterization of the coupling effect between local flow field disturbances and particle concentration, improving the modeling granularity. For example, the abrasion effect gradient sub-interval can include, but is not limited to, high flow velocity-high solids content intervals, low flow velocity-high solids content intervals, and high flow velocity-low solids content intervals.

[0044] The abrasion-affected region is divided into gradients based on the medium flow rate and gypsum solids content, generating multiple abrasion-affected gradient sub-regions. This can be achieved by using the medium flow rate and gypsum solids content as two orthogonal dimensions to perform two-dimensional meshing or clustering of the original abrasion-affected region. Furthermore, this operation can be achieved by setting multi-level thresholds for flow rate and solids content (e.g., high / medium / low) to form 3×3=9 combined sub-regions, or by using the DBSCAN algorithm to perform density clustering on historical operating condition data points to automatically identify naturally formed gradient clusters. This allows for precise spatial decoupling of the abrasion-driving mechanism under complex coupled operating conditions.

[0045] Based on multiple erosion influence gradient sub-intervals, segmented erosion detection is carried out on the target small-diameter long-distance pipeline, and multiple wall thickness parameters, medium parameters and material parameters of different pipe sections of the target pipeline are collected. Segmented erosion detection can be a process of dividing the pipeline into corresponding segments according to the boundaries of the erosion influence gradient sub-intervals, and performing detection operations independently in each segment. In this embodiment, segmented erosion detection can combine the pipeline geometric coordinates with the sub-interval mapping relationship to guide the detection equipment (such as a crawling robot or a fixed sensor array) to collect data segment by segment. Furthermore, segmented erosion detection can be used to ensure that the detection density matches the risk level, improving the spatial representativeness and operational condition alignment of the data. The wall thickness parameter can be the measured value or distribution data of the remaining thickness of the pipeline inner wall at a specific location. In an exemplary embodiment, the wall thickness parameter can be used to directly reflect the degree of material loss and is a core characterization indicator of the erosion state. Exemplarily, the wall thickness parameter can include, but is not limited to, one or more of the following: point measured wall thickness, circumferential average wall thickness, axial wall thickness gradient, etc. The medium parameter can be a set of operating parameters describing the physicochemical properties of the slurry, including but not limited to pH, temperature, particle hardness, etc. Furthermore, the medium parameter can be used to correct the corrosion and erosion coupling factor in the calculation of the degree of erosion. In a specific embodiment, the medium parameter can include, but is not limited to, slurry pH value, slurry temperature, particle Mohs hardness, etc. Material parameters can be the mechanical and corrosion resistance properties of the pipe material itself. In this embodiment, material parameters can be used as a benchmark for calculating the degree of abrasion, and to normalize the wear response of different pipe sections. For example, material parameters may include, but are not limited to, material yield strength, surface roughness, and erosion corrosion resistance coefficient.

[0046] Based on multiple abrasion-affected gradient sub-intervals, segmented abrasion detection of target small-diameter, long-distance pipelines can be performed. This can be achieved by planning the detection path according to the boundary coordinates of the sub-intervals and independently deploying detection resources within each sub-interval. Furthermore, this operation can be implemented by controlling a pipeline endoscope to adaptively adjust the sampling frequency according to the sub-interval length, or by assigning a dedicated sensor group to each sub-interval in a fixed measurement point system and synchronously triggering data acquisition. This allows the detection strategy to dynamically match the local risk level, improving data acquisition efficiency and representativeness. Multiple wall thickness parameters, medium parameters, and material parameters of different pipe sections of the target pipeline can be collected. This can involve simultaneously acquiring three types of heterogeneous parameters—structure, environment, and material—to form a multi-dimensional observation set. Further, this operation can be achieved by acquiring wall thickness using an ultrasonic array, reading medium parameters in real-time from the DCS system, and retrieving material parameters from equipment records, or by using a multi-functional composite probe to simultaneously collect wall thickness and local slurry samples and analyze the medium and material characteristics offline. This allows for the construction of an initial observation foundation covering all elements of abrasion.

[0047] The degree of abrasion was calculated for multiple wall thickness parameters, medium parameters, and material parameters respectively, and the initial abrasion state data corresponding to each parameter was obtained. The initial abrasion state data can be preliminary state characterization values ​​obtained by calculating the abrasion degree of three types of parameters: wall thickness, medium, and material. In this embodiment, the initial abrasion state data can be used as input for weighted calculation, preserving the independent information of each dimension's contribution to abrasion. Furthermore, the initial abrasion state data can be generated by calculating the abrasion degree of various parameters and participate in the construction of weighted abrasion state data. Abrasion degree calculations are performed on multiple wall thickness parameters, medium parameters, and material parameters to obtain the initial abrasion state data corresponding to each parameter. This can be achieved by designing a dedicated transformation function for each type of parameter, mapping it to a standardized abrasion degree index. Further, this operation can be implemented by converting the wall thickness parameter into a thinning rate relative to the design wall thickness, substituting the medium parameter into an empirical abrasion rate formula to deduce the equivalent abrasion amount, converting the material parameter into an equivalent reduction coefficient of the standard material, or using a pre-trained small neural network to perform abrasion sensitivity correction on the three types of parameters and output normalized state values, thereby achieving the mapping of heterogeneous parameters to a unified abrasion semantic space.

[0048] Calculate the erosion influence weight of each parameter on pipeline erosion, and based on the erosion influence weight, perform weighted calculation on the initial erosion state data to obtain the weighted erosion state data of each parameter; The erosion influence weight can be a coefficient that quantifies the relative contribution of parameters such as wall thickness, medium, and material to the overall erosion process. In this embodiment, the erosion influence weight can be determined based on statistical regression of historical failure cases, expert knowledge graphs, or sensitivity analysis methods. Furthermore, the erosion influence weight can be used to achieve differentiated fusion of multi-source heterogeneous parameters in erosion assessment, avoiding information distortion caused by equal-weighted averaging. For example, the erosion influence weight can include, but is not limited to, fair weights based on Shapley values, variance contribution weights based on principal component analysis, and causal weights based on Bayesian networks. Weighted erosion state data can be an intermediate state representation obtained by weighting various initial erosion state data according to their erosion influence weights. In an exemplary embodiment, the weighted erosion state data can generate corrected state values ​​for each parameter dimension through weighted summation or weighted averaging operations. Furthermore, the weighted erosion state data can be used to reflect the actual influence intensity of different factors under specific working conditions, improving the physical consistency of state characterization. In this embodiment, the weighted wear state data can be generated by the initial wear state data and the wear influence weights, and used as the input for data fusion processing.

[0049] Calculating the erosion influence weight of each parameter on pipeline erosion can be achieved by assessing the sensitivity or contribution of each parameter to the erosion outcome based on historical data or mechanistic models. Furthermore, this operation can be implemented by extracting feature importance from historical erosion models using interpretable AI methods such as LIME or SHAP, or by analyzing erosion rate changes under different parameter combinations and fitting weight coefficients through orthogonal experimental design, thereby establishing differentiated evaluation criteria for specific operating conditions. Based on the erosion influence weight, the initial erosion state data is weighted to obtain the weighted erosion state data for each parameter. This can be achieved by multiplying each initial state value by its corresponding weight to generate a weighted correction value. Further, this operation can be implemented using linear weighting (weighted value = weight × initial state value) or nonlinear weighting (weighted value = initial state value × weight, to enhance the distinguishability of high-weight parameters), thereby highlighting the influence of dominant factors and suppressing secondary noise interference.

[0050] The weighted erosion state data is fused to obtain the initial erosion state data of the target pipeline.

[0051] Data fusion processing can be an operation that integrates multiple weighted abrasion state data into a unified structured vector or tensor. In this embodiment, data fusion processing can achieve heterogeneous data alignment through features concatenating, principal component dimensionality reduction, or multimodal embedding. Furthermore, data fusion processing can be used to generate a standard input format that can be used for subsequent modeling, preserving multidimensional information while eliminating redundancy. To obtain the initial abrasion state data of the target pipeline, data fusion processing of the weighted abrasion state data can be performed by integrating multiple weighted state values ​​into a single structured data entity. Further, this operation can be achieved by concatenating all weighted values ​​to form a feature vector as the final initial state data, or by compressing and denoising the weighted state data using an autoencoder and outputting a low-dimensional fused representation, thereby generating a high-quality, multidimensional fused initial state input to support subsequent high-precision modeling.

[0052] Taking the joint evaluation of the variable diameter section and the straight pipe section as an example, the anti-erosion method of the inner wall of the pipe in this embodiment can be as follows: In the slurry circulation pipe at the outlet of a power plant absorption tower, the system identifies a region containing a DN200→DN150 diameter change as a high-disturbance erosion influence zone. Further, it is divided into four erosion influence gradient sub-zones according to the medium flow velocity (7.5–11.2 m / s) and gypsum solid content (28%–35%). In the high flow velocity-high solid content zone (within 5m downstream of the diameter change), the segmented erosion detection collected a wall thickness reduction rate of 18%, slurry pH = 4.8, and the pipe material was 316L stainless steel; while in the low flow velocity-medium solid content zone (straight pipe section), the wall thickness reduction rate was only 6%. After calculating the degree of erosion, the initial state values ​​of the three types of parameters are output as [0.18, 0.32, 0.10]. Based on historical data regression, the erosion impact weights were determined to be [0.6, 0.3, 0.1], which, after weighting, became [0.108, 0.096, 0.010]. Data fusion processing generated the initial erosion state data vector for the target pipeline [0.108, 0.096, 0.010], significantly higher than that of the straight pipe section [0.036, 0.045, 0.010]. Therefore, it was assigned a higher risk level in subsequent modeling, triggering a priority protection strategy.

[0053] In one embodiment, a relationship model is performed on the erosion influence range and the initial erosion state data of the target pipeline to obtain an erosion evolution relationship data model, including: Based on multiple erosion influence gradient sub-intervals, the initial erosion state data of the target pipeline is segmented to obtain the erosion state interval data corresponding to each erosion influence gradient sub-interval. The abrasion state interval data is discretized to obtain discretized abrasion state data points, and the abrasion influence intensity is discretized for each abrasion influence gradient sub-interval to obtain discretized abrasion influence intensity data points. By using a pre-set Markov model of erosion evolution, a quantitative analysis of the relationship between discrete erosion state data points and discrete erosion influence intensity data points is performed to obtain a data model of erosion evolution relationship.

[0054] The abrasion state interval data can be a subset of local initial abrasion state data corresponding to each sub-interval after dividing the data according to the abrasion influence gradient sub-intervals. This can be used to ensure that subsequent modeling is performed within physically consistent local regions, improving the alignment between the model and the actual flow field disturbances. In this embodiment, the abrasion state interval data can be extracted by slicing or indexing the initial abrasion state data of the overall target pipeline based on the spatial boundaries of the sub-intervals. Further, this operation can be performed by segmenting the initial abrasion state data of the target pipeline according to multiple abrasion influence gradient sub-intervals, obtaining the abrasion state interval data corresponding to each abrasion influence gradient sub-interval. This involves splitting the overall initial state data into segments based on the spatial or logical identifiers of the sub-intervals. For example, if the initial state data is indexed by pipeline coordinates, it is sliced ​​according to the start and end coordinates of the sub-intervals; if the data is organized by sensor ID, it is grouped and aggregated according to the sub-interval to which the sensor belongs. This allows for a one-to-one correspondence between modeling units and physical wear mechanism regions, improving local modeling accuracy.

[0055] Discretization of the degree of abrasion can be a process of mapping continuous abrasion state values ​​(such as wall thickness reduction rate) to a finite number of predefined discrete levels. This can be used to reduce data noise sensitivity and construct a finite state space that can be processed by a Markov model. In an exemplary embodiment, the discretization of the degree of abrasion can be quantized using equal-width binning, equal-frequency binning, or threshold partitioning methods based on wear mechanisms. Further, this operation can involve discretizing the degree of abrasion indices within a range of abrasion states to obtain discretized abrasion state data points, which map continuous abrasion indices to a predefined finite set of states. For example, three levels of discretization can be used: [0–1 mm) → “mild”, [1–2.5 mm) → “moderate”, ≥2.5 mm → “severe”; or K-means can be used to cluster historical abrasion amounts, with the cluster centers as the discrete level benchmark, thereby constructing a symbolic state space suitable for time-series state modeling and enhancing model robustness.

[0056] Discretized erosion state data points can be symbolic state values ​​representing the erosion level of the pipeline within a specific sub-interval, obtained after discretization of erosion degree. These values ​​can be used as inputs to the state nodes of a Markov model, supporting state transition probability calculations. In one specific embodiment, discretized erosion state data points can be generated from erosion state interval data after erosion degree discretization; and together with discretized erosion influence intensity data points, they constitute the joint input of the Markov model. Furthermore, the erosion degree discretization process can include, but is not limited to, one or more of the following: discretization based on wear stage (early / mid / late stage), discretization based on safety margin (safe / early warning / dangerous), and discretization based on statistical quantiles.

[0057] Discretization of abrasion influence intensity can be an operation that transforms continuous operating condition driving variables (such as flow velocity and solids content combinations) into a finite number of abrasion driving intensity levels. This can be used to simplify complex multiphase flow environments into enumerable external excitation states, facilitating joint modeling with abrasion states. In this embodiment, the discretization of abrasion influence intensity can be mapped based on the clustering results of multidimensional operating condition parameters or a preset intensity level table. Further, this operation can involve discretizing the abrasion influence intensity for each abrasion influence gradient sub-interval to obtain discretized abrasion influence intensity data points, which map the comprehensive operating condition parameters (such as flow velocity × solids content) of the sub-interval to intensity levels. For example, an intensity index can be defined as α·(flow velocity / reference flow velocity) + β·(solids content / reference solids content), then discretized into three levels; or intensity levels can be labeled for typical sub-intervals based on CFD simulation results, establishing a mapping rule base. This simplifies complex multidimensional operating conditions into enumerable external conditions, supporting conditional state transition modeling.

[0058] Discretized abrasion influence intensity data points can be discrete symbols representing the external abrasion driving intensity level of a specific abrasion influence gradient sub-interval at a certain time period. These data points can be used as external excitations or contextual conditions for Markov models to modulate state transition probabilities. In an exemplary embodiment, discretized abrasion influence intensity data points can be generated by discretizing the operating parameters of the abrasion influence gradient sub-interval and paired with discretized abrasion state data points for input into the model. Furthermore, the discretization of abrasion influence intensity can include, but is not limited to, high erosion-high concentration intensity levels, low erosion-medium concentration intensity levels, and steady-state low intensity levels.

[0059] The pre-set abrasion evolution Markov model can be a pre-trained time-series state transition probability model with discrete abrasion states as nodes and abrasion intensity as a condition. It can be used to characterize the evolution of abrasion states under different working conditions and intensity levels, supporting future state probability prediction. In this embodiment, the pre-set abrasion evolution Markov model can learn the state transition matrix based on historical detection sequence data through maximum likelihood estimation or Bayesian inference, with different intensity levels corresponding to different transition sub-matrices. Furthermore, this model can include, but is not limited to, Hidden Markov Models (HMMs), Conditional Markov Chains (CMCs), and Semi-Markov Processes (SMPs). Relationship quantification analysis can be the process of jointly modeling discrete state and intensity data using the Markov model to output the state transition probability distribution. This can be used to establish a mathematical representation of the abrasion evolution process, forming a computable and predictable abrasion evolution relationship data model. In a specific embodiment, relationship quantification analysis can input paired (state, intensity) sequences into the model to statistically analyze or infer the probability of transitioning to the next state under each state. Furthermore, this operation can be performed using a pre-set erosion evolution Markov model to quantify the relationship between discretized erosion state data points and discretized erosion influence intensity data points, resulting in an erosion evolution relationship data model. This model involves inputting paired (state, intensity) sequences into the Markov model to learn or invoke its state transition rules. For example, during the training phase, the EM algorithm can be used to learn the transition matrix for each intensity level from historical sequences; during the inference phase, a pre-set matrix can be directly invoked for forward prediction; alternatively, a reinforcement learning framework can be used to update the transition probabilities online to adapt to operating condition drift, thereby constructing a nonlinear erosion evolution model with time-series prediction capabilities, overcoming the limitations of traditional static or linear methods.

[0060] Taking the prediction of erosion trend in the high-disturbance zone of an elbow as an example, the pipeline inner wall anti-erosion method in this embodiment can be as follows: A 90° elbow in a desulfurization system is divided into a high-flow-high-solids-content erosion influence gradient sub-interval. Its erosion state interval data shows that the current wall thickness reduction is 1.8 mm, which is discretized and mapped to a "moderate" state (denoted as S2). The strength index calculated under the same operating conditions is 2.3, discretized to a "high strength" level (denoted as I3). The system calls a preset erosion evolution Markov model. Under I3 conditions, the annual transition probability from state S2 to S3 (severe) is 0.65, and the reverse probability to S1 (mild) is 0. Based on this, the erosion evolution relationship data model outputs that the probability of the elbow entering severe erosion within the next 12 months reaches 65%, triggering a warning domain judgment and automatically recommending ceramic lining repair within 6 months to avoid the risk of perforation.

[0061] In one embodiment, the critical erosion thickness for pipeline failure is identified using the erosion evolution relationship data model to obtain the critical erosion threshold for the pipeline, including: Curve fitting is performed on the erosion evolution relationship data model to generate the corresponding erosion thickness evolution relationship curve; The erosion thickness evolution curve can be a function curve representing the continuous change of the erosion thickness of the pipeline inner wall over time or operating cycle, generated by curve fitting from the erosion evolution data model. It can be used to explicitly present the nonlinear evolution trend of the erosion process, providing visualization and mathematical basis for domain partitioning and critical point identification. In this embodiment, the erosion thickness evolution curve can be formed as a differentiable and analyzable continuous function expression by interpolating or smoothing the erosion evolution data model in the time dimension. Furthermore, curve fitting the erosion evolution data model to generate the corresponding erosion thickness evolution curve can be achieved by using mathematical fitting methods to transform the discrete or black-box erosion evolution data model into a continuous smooth curve. For example, this operation can be achieved by using cubic spline interpolation to smoothly connect the model's output at multiple time steps, or by using Gaussian process regression to model the uncertainty of the model's prediction results and generate a mean evolution curve, thereby obtaining an analytical and differentiable erosion evolution expression, supporting subsequent refined domain partitioning and critical point location.

[0062] The erosion thickness evolution curve is divided into a safe domain, an early warning domain, and a failure domain to obtain the corresponding safe domain, early warning domain, and failure domain. The safety domain can be an operating state interval on the erosion thickness evolution curve corresponding to sufficient pipeline structural integrity and requiring no intervention. It can be used to define the normal operating range and support maintenance-free decision-making. In an exemplary embodiment, the safety domain can include, but is not limited to, one or more of the following: initial stable erosion section, low-rate corrosion section, and design margin retention section. The warning domain can be a transitional state interval on the erosion thickness evolution curve corresponding to significant pipeline degradation but not yet endangering safety. It can be used to trigger the formulation of preventive maintenance plans and resource pre-positioning. In a specific embodiment, the warning domain can include, but is not limited to, one or more of the following: accelerated erosion initiation section, corrosion-erosion coupling enhancement section, and mid-life section. The failure domain can be a dangerous state interval on the erosion thickness evolution curve corresponding to insufficient pipeline structural strength and the risk of perforation and leakage. It can be used to identify high-risk states that require immediate intervention to prevent sudden failure. Furthermore, the failure domain can include, but is not limited to, one or more of the following: critical strength loss section, leakage probability steep increase section, and structural instability imminent section. Dividing the erosion thickness evolution curve into a safe domain, a warning domain, and a failure domain can be done based on engineering safety criteria or historical failure data. Two boundary points can be set on the curve, dividing the entire curve into three continuous intervals. For example, this operation can be achieved by calculating the pressure-bearing capacity at different erosion depths based on residual strength theory and then back-calculating the boundaries of the three domains using the design pressure; or by using the erosion depth distribution of actual leaks in historical maintenance records to set the starting point of the failure domain and proportionally dividing it into warning domains. This discretizes the continuous erosion process into state intervals with clear operational and maintenance implications, enabling risk-level management.

[0063] Based on the safety domain, early warning domain, and failure domain, the critical erosion threshold critical point is identified from the erosion thickness evolution curve to obtain the critical erosion critical point. The critical erosion threshold can be the boundary point between the warning domain and the failure domain on the erosion thickness evolution curve. It represents the turning point from controllable degradation to uncontrollable failure and can be used as a key location basis for dynamic threshold generation, reflecting the actual failure initiation characteristics of a specific pipe section. In this embodiment, the critical erosion threshold can be determined by detecting abrupt changes in the curve slope, extreme values ​​of the second derivative, or based on a risk probability threshold. Identifying the critical erosion threshold on the erosion thickness evolution curve based on the safety domain, warning domain, and failure domain can identify the boundary between the warning domain and the failure domain, thus determining the critical erosion threshold. Furthermore, this operation can be achieved by searching for the erosion depth point on the curve that satisfies a preset failure probability (e.g., Pf = 0.1) as the critical point, or by detecting the maximum curvature of the curve or the abrupt change in erosion rate acceleration to locate the failure initiation turning point. This allows for accurate capture of the key turning point from controllable degradation to high-risk failure under specific operating conditions, avoiding subjective setting deviations.

[0064] The critical wear threshold is generated by numerically mapping the wall thickness to the critical wear point.

[0065] The wall thickness numerical mapping can be a mathematical mapping process that converts the position of the critical wear point on the time or operating cycle axis into the corresponding actual wall thickness loss value. This can be used to generate executable threshold indicators with engineering units (such as millimeters) to support on-site decision-making. In a specific embodiment, wall thickness numerical mapping can be achieved by using the functional expression of the wear thickness evolution curve, substituting the horizontal coordinate of the critical point into the corresponding vertical coordinate (i.e., cumulative wear depth). Performing wall thickness numerical mapping on the critical wear point to generate the corresponding pipeline critical wear threshold can be achieved by converting the coordinates of the critical point on the evolution curve into a specific wear depth or remaining wall thickness value. For example, this operation can be implemented by directly reading the vertical coordinate value of the critical point as the threshold if the vertical axis of the curve represents the cumulative wear amount, or by subtracting the value from the original wall thickness if the vertical axis of the curve represents the remaining wall thickness, or by directly using the remaining wall thickness as the threshold standard. This allows for the output of a quantitative threshold with engineering feasibility, which can be compared with the measured cumulative wear amount to determine the current state.

[0066] Taking the dynamic threshold generation of a variable-diameter pipe section as an example, the anti-erosion method for the inner wall of the pipe in this embodiment can be as follows: A slurry pipe in a desulfurization system is classified as a high-disturbance erosion zone due to a sudden increase in flow velocity at the DN150 to DN100 diameter change. Based on six months of operating data and initial thickness measurement results for this section, the erosion evolution relationship data model predicts that the erosion exhibits an exponentially accelerating trend. After obtaining the erosion thickness evolution relationship curve through curve fitting, based on relevant pipe strength assessment standards and historical leakage data, the remaining wall thickness ≥6.0mm is classified as the safety zone, 4.2–6.0mm as the warning zone, and <4.2mm as the failure zone. The critical erosion threshold is located at a remaining wall thickness of 4.2mm, corresponding to a cumulative erosion amount of 3.8mm. Through wall thickness numerical mapping, the system outputs a critical erosion threshold of 3.8mm. The current measured cumulative erosion amount is 3.1mm, which is within the warning zone, triggering the coating repair plan and avoiding the risk of delayed intervention caused by using a traditional fixed threshold (such as a uniform 3.0mm).

[0067] In one embodiment, the cumulative erosion amount is calculated from the erosion evolution relationship data model to obtain the cumulative erosion amount of the pipe inner wall, including: The corrosion trend was detected by analyzing the corrosion evolution relationship data model to obtain pipeline corrosion trend indicators. Based on pipeline erosion trend indicators, create a erosion calculation function corresponding to the erosion evolution relationship data model. Based on the erosion calculation function, the cumulative erosion of the inner wall of the pipeline is calculated from the erosion evolution relationship data model, and the cumulative erosion of the inner wall of the pipeline is obtained.

[0068] The erosion trend detection process involves extracting dynamic features of erosion rate changes from an erosion evolution relationship data model. This can be used to generate trend indicators reflecting the dynamic evolution characteristics of erosion, providing a basis for customized cumulative calculations. In this embodiment, erosion trend detection can perform first / second derivative calculations, sliding window slope analysis, or time series pattern recognition on the erosion-time series output by the model to determine whether erosion is accelerating, stable, or slowing down. Furthermore, erosion trend detection can form an input-output relationship with pipeline erosion trend indicators. For example, erosion trend detection can calculate the local slope within the sliding window of the erosion depth sequence output by the model and cluster it into trend categories, or use wavelet transform to extract multi-scale variation features of the erosion sequence to construct a trend intensity index, thereby transforming the static model into an evaluation tool with dynamic perception capabilities and revealing the non-stationary characteristics of the erosion process.

[0069] Pipeline erosion trend indicators can be quantitative characteristic values ​​that characterize the direction and intensity of pipeline inner wall erosion rate changes over time. They can guide the selection of the structure and parameter configuration of the erosion calculation function, ensuring that the cumulative calculation adapts to the actual nonlinear evolution process. In an exemplary embodiment, the pipeline erosion trend indicator is output by the erosion trend detection operation and can be a classification label (such as "accelerating" / "stable") or a continuous value (such as the acceleration coefficient). Furthermore, the pipeline erosion trend indicator can form a condition-response relationship with the erosion calculation function. In a specific embodiment, the pipeline erosion trend indicator can include, but is not limited to, one or more of the following: erosion acceleration indicator, erosion fluctuation frequency indicator, and erosion stage transition indicator.

[0070] Erosion trend detection is performed on the erosion evolution relationship data model to obtain pipeline erosion trend indicators. This can be achieved by analyzing the time-series erosion data output by the erosion evolution relationship data model and identifying its changing trend characteristics. Furthermore, this operation can be achieved by calculating the local slope within a sliding window of the erosion depth sequence output by the model and clustering it into trend categories, or by using wavelet transform to extract multi-scale change features of the erosion sequence and constructing a trend intensity index. This transforms the static model into an assessment tool with dynamic perception capabilities, revealing the non-stationary characteristics of the erosion process. Based on the pipeline erosion trend indicators, a erosion amount calculation function corresponding to the erosion evolution relationship data model is created. This can be achieved by configuring or generating suitable cumulative calculation rules based on the type and value of the trend indicator. Further, this operation can be achieved by using a quadratic or exponential integral kernel if the trend indicator is "accelerating" and a linear accumulation if it is "stationary," or by constructing a neural network submodule with the trend indicator as a conditional input and dynamically outputting an integral weight vector. This couples the cumulative erosion amount calculation with actual operating condition fluctuations and the nonlinear characteristics of material degradation, improving calculation accuracy.

[0071] The erosion calculation function can be a nonlinear mathematical mapping relationship constructed based on pipeline erosion trend indicators to calculate cumulative erosion. It can be used to achieve high-fidelity backtracking of historical erosion processes, overcoming the limitations of fixed-rate or linear accumulation assumptions. In this embodiment, the erosion calculation function can dynamically select the integral kernel function, piecewise weight coefficients, or recursive update rules based on the trend indicators to form a calculation logic adapted to the current evolution state. For example, the erosion calculation function can include, but is not limited to, variable-weight integral functions, piecewise exponential decay accumulation functions, and recursive summation functions with memory effects. Based on the erosion calculation function, the cumulative erosion of the pipeline inner wall is calculated using the erosion evolution relationship data model. This can be achieved by substituting the historical output of the erosion evolution relationship data model into the customized erosion calculation function and performing numerical solutions. Furthermore, this operation can be achieved by weighting and accumulating the instantaneous erosion rate under discrete time steps according to the weights adjusted by trend indicators, or by using an adaptive step-size numerical integration method combined with a variable kernel function to solve the continuous model output, thereby realizing high-precision cumulative quantification of the nonlinear erosion process under complex working conditions and providing a reliable benchmark for risk assessment.

[0072] Taking the cumulative erosion assessment under the condition of periodic fluctuations in solid content as an example, the pipeline inner wall erosion resistance method in this embodiment can be as follows: Due to unstable limestone supply in the desulfurization system of a power plant, the solid content of the slurry fluctuates periodically between 20% and 35%. The output of the erosion evolution relationship data model shows that the erosion depth increases in a stepwise manner. Through erosion trend detection, an alternating pattern of "acceleration-stable" occurs once every 72 hours, generating an erosion trend index of "periodic acceleration". Based on this, a piecewise weighted integral function is created: 1.8 times the weight is assigned to the acceleration segment, and 1.0 times to the stable segment. The final cumulative erosion amount is calculated to be 2.7 mm, which is closer to the measured value (2.65 mm) than the traditional linear accumulation (2.1 mm), significantly improving the assessment accuracy and avoiding delays in maintenance due to underestimating the risk.

[0073] In one embodiment, the critical erosion threshold of the pipeline and the cumulative erosion amount of the pipeline inner wall are parameter-mapped and encoded to generate a target pipeline erosion resistance evaluation vector, including: The K-Bins algorithm is used to calculate the cumulative erosion amount of the inner wall of the pipe using discrete intervals to obtain the target K value; Obtain the maximum and minimum values ​​of the cumulative erosion on the inner wall of the pipe, and determine the corresponding discretization interval range based on the maximum and minimum values; K target intervals are determined based on the target K value and the discretization interval range. The cumulative erosion of the inner wall of the pipe is mapped based on the K target interval ranges to obtain the cumulative erosion of the pipe corresponding to each target interval range. Multiple pipe erosion features are generated based on the cumulative pipe erosion amount corresponding to each target interval range. Multiple pipeline erosion characteristics and critical erosion thresholds are normalized and vectorized to obtain the target pipeline erosion resistance evaluation vector.

[0074] The K-Bins algorithm is an unsupervised data discretization method that divides continuous numerical variables into K discrete intervals. It can be used to intelligently divide the cumulative erosion amount on the inner wall of a pipe into intervals, improving the ability of features to distinguish erosion stages. In this embodiment, the K-Bins algorithm can automatically select the optimal number of discrete intervals based on the distribution characteristics of the cumulative erosion amount data, through minimizing information entropy, maximizing the profile coefficient, or the elbow rule. The target K value can be the optimal number of discrete intervals determined by the K-Bins algorithm, which can be used to control the discretization granularity, preserving the differences in key erosion stages while avoiding overfitting. Furthermore, the target K value can be achieved by automatically selecting the optimal number of discrete intervals K through analysis of the cumulative erosion amount data distribution. In an exemplary embodiment, the target K value can be evaluated using the elbow rule to assess the change in variance within the interval under different K values, selecting the inflection point as the target K value; or based on the information gain or the rate of change of Gini impurity, selecting the K value that maximizes the class discrimination, thereby achieving adaptive discretization granularity control and matching the quantitative characteristics of the actual erosion evolution stages.

[0075] The maximum value of the cumulative erosion amount on the pipe inner wall can be the maximum cumulative erosion depth recorded at all detection points within the current evaluation period. This can be used as the upper bound of the discretization interval range to ensure coverage of actual extreme erosion conditions. The minimum value of the cumulative erosion amount on the pipe inner wall can be the minimum cumulative erosion depth recorded at all detection points within the current evaluation period. This can be used to reflect the lowest wear level of the system. In a specific embodiment, obtaining the maximum and minimum values ​​of the cumulative erosion amount on the pipe inner wall and determining the corresponding discretization interval range based on these values ​​can be achieved by extracting extreme values ​​from historical or current detection data to construct effective numerical boundaries. Furthermore, this operation can be implemented by removing outliers other than 3σ and taking the extreme values ​​as the interval boundaries, or by using a sliding window to dynamically update the extreme values ​​to adapt to operating condition drift, thereby ensuring that the discretization range closely follows the actual data distribution and avoiding the waste of expressive power by invalid intervals.

[0076] The discretized interval range can be a numerical span consisting of the maximum and minimum cumulative abrasion amounts. This range can be used to define the effective scope of the K-Bins partition and exclude outlier interference. The K target interval ranges can be K mutually exclusive and continuous sub-intervals generated within the discretized interval range according to the K-Bins rule. These can be used to provide a structured segmentation basis for subsequent mapping processing, achieving semantic representation of the abrasion state. Furthermore, by determining the K target interval ranges based on the target K value and the discretized interval range, K sub-intervals can be partitioned within a given numerical span using the K-Bins strategy. For example, this operation can employ equal-width partitioning, where the interval width equals the difference between the maximum and minimum values ​​divided by K; or equal-frequency partitioning, ensuring each interval contains approximately the same number of sample points, thereby generating a clearly structured and fully covered abrasion state partition to support subsequent feature mapping.

[0077] The cumulative pipe erosion value corresponding to each target interval range can be a set of original cumulative erosion value observations falling within each target interval, which can be used as the basic data unit for generating pipe erosion features. Further, the cumulative pipe erosion value of the inner wall is mapped according to the K target interval ranges to obtain the cumulative pipe erosion value corresponding to each target interval range. This can be achieved by assigning the original cumulative erosion value to its respective interval and aggregating it by interval. In a specific embodiment, this operation can record a list of specific erosion value values ​​contained in each interval, or only retain the interval affiliation identifier while discarding the original continuous values, thereby completing the initial transformation from continuous measurement to discrete state and providing a foundation for feature generation.

[0078] Multiple pipeline erosion features can be representative variables generated based on the cumulative erosion amount statistics or encoding within each interval. These features can be used to transform continuous erosion amounts into discrete features with stage semantics, enhancing model interpretability. In this embodiment, multiple pipeline erosion features can be constructed as structured features based on interval affiliation and internal statistics. Furthermore, these features can be generated using interval numbering, interval mean, existence indicators, etc. For example, multiple pipeline erosion features can include, but are not limited to, one or more of interval index features, interval mean erosion features, and interval activation binary features. In an exemplary embodiment, multiple pipeline erosion features are generated based on the cumulative pipeline erosion amount corresponding to each target interval range. This can be achieved by generating a binary feature (whether it is activated) for each interval, or by calculating the mean or variance of the erosion amount within each interval as a continuous feature, thereby realizing a semantic expression of the erosion state and highlighting the differences in key degradation stages.

[0079] Normalization is the process of scaling features of different dimensions or scales to a uniform numerical range. It can be used to eliminate the dimensional differences between the critical corrosion threshold and corrosion features of a pipeline, thereby improving the stability of model training. Vector transformation is the operation of integrating normalized multidimensional features into a fixed-length numerical vector. It can be used to generate a standard input format that the anti-corrosion process decision model can directly process. Furthermore, normalization and vector transformation are performed on multiple pipeline corrosion features and the critical corrosion threshold to obtain the target pipeline anti-corrosion evaluation vector. This can be done by unifying the feature scale and integrating it into a fixed-dimensional vector. In a specific embodiment, this operation can be performed by performing Min-Max normalization to [0, 1] on all features and then concatenating them into a vector; or by normalizing the corrosion features and the critical threshold separately and then fusing them into a low-dimensional embedding vector through linear projection, thereby generating a high-quality, highly compatible model input that takes into account both state description and risk boundary information.

[0080] Taking the refined coding of straight pipe section erosion status as an example, the pipe inner wall anti-erosion method in this embodiment can be as follows: The historical data range of the cumulative erosion of a horizontal straight pipe section in a desulfurization system is 0.8mm to 2.9mm. After the system analyzes the data distribution using the K-Bins algorithm, it automatically determines the target K value to be 4, dividing it into four intervals: [0.8–1.3), [1.3–1.8), [1.8–2.3), and [2.3–2.9]. The current detection value of 2.1mm falls into the third interval, generating an erosion feature vector [0, 0, 1, 0] (interval activation code). At the same time, the critical erosion threshold of this pipe section is 3.2mm. After Min-Max normalization, the threshold is converted to 0.96, which is concatenated with the erosion feature to form a 5-dimensional evaluation vector [0, 0, 1, 0, 0.96]. After inputting into the anti-erosion process decision model, the output scheme of "local spraying of a polymer wear-resistant coating with a thickness of 1.8mm" is given, avoiding over-protection of low-risk pipe sections.

[0081] In one embodiment, the target pipeline anti-corrosion evaluation vector is input into a pre-set anti-corrosion process decision model to output an anti-corrosion process scheme, resulting in a target anti-corrosion construction scheme for the target small-diameter long-distance pipeline, including: The anti-corrosion evaluation vector of the target pipeline is input into the preset anti-corrosion process decision model, which includes a first bidirectional gated loop unit, a second gated loop unit, and a double-layer fully connected output layer. Obtain the anti-corrosion process standard parameters for small-diameter long-distance desulfurization pipelines in thermal power plants, and construct the corresponding standard anti-corrosion requirement vector based on the anti-corrosion process standard parameters; The target pipeline anti-corrosion evaluation vector is input into the first bidirectional gated loop unit for feature extraction to obtain the first anti-corrosion feature vector, and the standard anti-corrosion requirement vector is input into the second gated loop unit for feature extraction to obtain the second process requirement feature vector. The first wear-resistant feature vector and the second process requirement feature vector are fused to obtain a fused wear-resistant feature vector. The anti-corrosion construction process parameters are calculated and matched by a double-layer fully connected output layer. Based on the anti-corrosion construction process parameters, the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline is determined.

[0082] The anti-erosion process decision model can be a deep neural network architecture comprising a first bidirectional gated recurrent unit, a second gated recurrent unit, and a two-layer fully connected output layer, used to fuse the actual erosion state of the pipeline with process specification constraints to generate a construction plan. In this embodiment, the anti-erosion process decision model can optimize weights by jointly training two feature extraction branches and a fusion output head, using historical maintenance data and standard parameter alignment. Furthermore, the anti-erosion process decision model can receive the target pipeline anti-erosion evaluation vector and the standard anti-erosion requirement vector as input, and output anti-erosion construction process parameters. For example, the anti-erosion process decision model can include, but is not limited to, one or more of the following: a bi-branch BiGRU-GRU fusion model, a specification alignment decision model with attention mechanism, and a multi-task constraint-aware process recommendation model.

[0083] Obtain the anti-corrosion process standard parameters for small-diameter, long-distance desulfurization pipelines in thermal power plants, and construct corresponding standard anti-corrosion requirement vectors based on these parameters. This can be achieved by extracting structured process constraints from power plant technical regulations, industry standards, or equipment manuals and encoding them as numerical vectors. In one exemplary embodiment, this operation can be performed by parsing PDF-formatted corrosion protection design guidelines for thermal power plant desulfurization systems, extracting key parameters through natural language processing entity recognition, and quantifying them; or by calling an enterprise knowledge graph API to query standard process entries matching pipeline material and media characteristics, generating standardized requirement vectors. This transforms unstructured or semi-structured engineering specifications into numerical inputs that the model can process, ensuring the compliance of the solution.

[0084] The target pipeline anti-corrosion evaluation vector is input into a first bidirectional gated recurrent unit for feature extraction to obtain a first corrosion feature vector. This can be achieved by using the bidirectional gated recurrent unit to perform context-aware encoding on the temporal or logical correlation dimensions in the evaluation vector. In one specific embodiment, this operation can treat the evaluation vector as a single time step sequence, and generate features by concatenating the forward and backward hidden states of the bidirectional gated recurrent unit. If the evaluation vector contains historical sliding window data, the bidirectional gated recurrent unit is expanded step-by-step to capture the evolution trend, thereby enhancing the ability to express the implicit dynamic patterns in the corrosion state and improving feature discriminability. The first corrosion feature vector can be a high-dimensional feature representation containing the dynamic characteristics of corrosion evolution extracted from the target pipeline anti-corrosion evaluation vector by the first bidirectional gated recurrent unit, used to capture the temporal correlation and nonlinear degradation pattern between the cumulative corrosion amount and the critical threshold. Furthermore, the first corrosion feature vector can be generated by the first bidirectional gated recurrent unit and participate in vector fusion to generate a fused anti-corrosion feature vector.

[0085] The standard anti-wear requirement vector is input into the second gated loop unit for feature extraction to obtain the second process requirement feature vector. This can be achieved by using a unidirectional gated loop unit to perform semantic compression and constraint relationship modeling on the standard requirement vector. For example, this operation can sort the standard parameters by category and input them into the gated loop unit to learn the dependency logic between parameters (such as the influence of temperature resistance on material selection); or a gated loop unit with a discard mechanism can be used to prevent overfitting to sparse standard vectors, improving generalization ability. This allows for the extraction of potential process rules from static specifications, supporting intelligent alignment with actual conditions. The second process requirement feature vector can be a feature representation reflecting the semantic structure of the process specification extracted from the standard anti-wear requirement vector by the second gated loop unit, used to parse the implicit constraint logic in the standard parameters, such as material-thickness-environment matching rules. Furthermore, the second process requirement feature vector can be generated by the second gated loop unit and participate in vector fusion to generate a fused anti-wear feature vector.

[0086] Vector fusion is performed on the first abrasion feature vector and the second process requirement feature vector to obtain a fused anti-abrasion feature vector. This can be achieved by combining the two types of features in a vector space to form a unified decision representation. In a specific embodiment, this operation can be performed by splicing followed by linear transformation and normalization; or a gating mechanism can be introduced to dynamically adjust the weights of standard constraints according to the severity of abrasion, thereby achieving a coordinated representation of the actual risk state and engineering specifications, and preventing the solution from deviating from the site or violating standards. The fused anti-abrasion feature vector can be a comprehensive feature representation formed by fusing the first abrasion feature vector and the second process requirement feature vector through splicing, weighting, or attention mechanisms. In this embodiment, the fused anti-abrasion feature vector can integrate state information and specification constraints through feature-level fusion strategies (such as splicing and batch normalization) to construct a decision space that simultaneously satisfies the actual wear condition and engineering feasibility, supporting the generation of accurate process parameters. Furthermore, the fused anti-abrasion feature vector can be generated by vector fusion operations and used as input to a two-layer fully connected output layer.

[0087] The anti-abrasion construction process parameters are calculated and output through a two-layer fully connected output layer. This can be achieved by mapping the fused features to a multi-dimensional process parameter space through two layers of nonlinear transformation. For example, this operation can extract key decision factors through dimensionality reduction in the first fully connected layer, and output classification (e.g., material type) and regression (e.g., thickness) results through bifurcation in the second layer; or residual connections can be introduced into the output layer to retain some information from the original fused features to enhance stability. This can generate a combination of construction parameters that is structurally clear, highly interpretable, and conforms to multi-objective optimization. The anti-abrasion construction process parameters can be a set of specific executable technical indicators output by the model, used to guide the implementation of anti-abrasion operations on site. Furthermore, the anti-abrasion construction process parameters can directly determine the technical details of the construction plan, such as the type of material selected, the coating thickness, and the repair process used. For example, the anti-abrasion construction process parameters can include, but are not limited to, one or more of the following: wear-resistant coating material type, lining structure form, and local repair priority.

[0088] Determining the target anti-corrosion construction scheme for a small-diameter, long-distance pipeline based on anti-corrosion construction process parameters can involve mapping the parameters output by the model into executable engineering instructions or work order content. In an exemplary embodiment, this operation can automatically fill in a standard construction template to generate a PDF scheme document containing a material list, process steps, and acceptance criteria; or write the parameters into the work order field of a computerized maintenance management system to trigger subsequent approval and execution processes, thereby completing the final link from algorithm output to project implementation and achieving closed-loop automation.

[0089] Taking the optimization of coating selection for straight pipe sections as an example, the anti-corrosion method for the inner wall of the pipe in this embodiment can be as follows: The cumulative corrosion of a DN150 straight pipe in the desulfurization system of a power plant is 1.3mm, the critical threshold is 3.0mm, and the evaluation vector is [1.3, 3.0, 0.57]. The system simultaneously obtains the plant's "Standard for Corrosion Protection of Slurry Pipelines," which stipulates that limestone slurry pipeline coatings must meet the requirements of pH>4 tolerance, minimum thickness of 2.0mm, and service life ≥5 years, and constructs a standard anti-corrosion requirement vector. The first bidirectional gated loop unit extracts the characteristic that the pipe section is in a state of "moderate wear but stable trend," and the second gated loop unit parses the specification tendency of "priority of ceramic materials." The fused features are output through a double-layer fully connected process: silicon carbide modified epoxy coating, thickness 2.2mm, local spraying process.

[0090] In addition, refer to Figure 2 To achieve the above objectives, the present invention also provides a pipe inner wall anti-corrosion system, the system comprising: The initial inspection module 10 is used to determine the erosion influence range of the inner wall of the pipeline based on the actual operating parameters of the desulfurization pipeline, perform erosion pretreatment detection and operating parameter analysis on the target small-diameter long-distance pipeline, and obtain the initial erosion state data of the target pipeline. Evolution modeling module 20 is used to model the relationship between the erosion influence range and the initial erosion state data of the target pipeline to obtain an erosion evolution relationship data model; Threshold identification module 30 is used to identify the critical erosion thickness of the pipeline failure in the erosion evolution relationship data model and obtain the critical erosion threshold of the pipeline. The cumulative erosion calculation module 40 is used to calculate the cumulative erosion amount on the erosion evolution relationship data model to obtain the cumulative erosion amount of the inner wall of the pipeline. The vector encoding module 50 is used to perform parameter mapping encoding on the critical erosion threshold of the pipeline and the cumulative erosion amount of the inner wall of the pipeline to generate a target pipeline erosion resistance evaluation vector. The scheme decision module 60 is used to input the anti-corrosion evaluation vector of the target pipeline into the preset anti-corrosion process decision model to output the anti-corrosion process scheme, thereby obtaining the target anti-corrosion construction scheme of the target small-diameter long-distance pipeline.

[0091] Other embodiments or specific implementations of the pipe inner wall anti-corrosion system of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0092] In addition, to achieve the above objectives, the present invention also provides a pipe inner wall anti-corrosion device, the device comprising: a memory, a processor, and a pipe inner wall anti-corrosion program stored in the memory and executable on the processor, the pipe inner wall anti-corrosion program being configured to implement the steps of the pipe inner wall anti-corrosion method as described above.

[0093] In addition, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a pipe inner wall anti-corrosion program, which, when executed by a processor, implements the steps of the pipe inner wall anti-corrosion method as described above.

[0094] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for preventing abrasion on the inner wall of a pipe, characterized in that, The method includes: Based on the actual operating parameters of the desulfurization pipeline, the erosion influence range of the pipeline inner wall is determined. The erosion pretreatment detection and operating parameter analysis are carried out on the target small-diameter long-distance pipeline to obtain the initial erosion state data of the target pipeline. A relationship model is performed on the erosion influence range and the initial erosion state data of the target pipeline to obtain an erosion evolution relationship data model. The critical erosion thickness for pipeline failure is identified by performing the erosion evolution relationship data model to obtain the critical erosion threshold for the pipeline. The cumulative erosion amount is calculated by performing a cumulative erosion calculation on the erosion evolution relationship data model to obtain the cumulative erosion amount of the inner wall of the pipe; The critical erosion threshold of the pipeline and the cumulative erosion amount of the inner wall of the pipeline are parameter-mapped and encoded to generate a target pipeline erosion resistance evaluation vector. The anti-corrosion evaluation vector of the target pipeline is input into a pre-set anti-corrosion process decision model to output an anti-corrosion process scheme, thereby obtaining the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline.

2. The method for preventing corrosion of the inner wall of a pipe according to claim 1, characterized in that, The process involves determining the erosion influence range of the pipeline inner wall based on actual operating parameters of the desulfurization pipeline, conducting erosion pretreatment detection and operating parameter analysis on the target small-diameter long-distance pipeline, and obtaining initial erosion state data of the target pipeline, including: Based on the actual operating parameters of the desulfurization system of the thermal power plant, the erosion influence range of the inner wall of the pipeline is determined, and the erosion influence range is divided into multiple erosion influence gradient sub-ranges according to the medium flow rate and gypsum solid content. Based on the multiple erosion influence gradient sub-intervals, segmented erosion detection is performed on the target small-diameter long-distance pipeline, and multiple wall thickness parameters, medium parameters and material parameters of different pipe sections of the target pipeline are collected. The degree of abrasion is calculated for each of the multiple wall thickness parameters, medium parameters, and material parameters to obtain the initial abrasion state data for each parameter. Calculate the erosion influence weight of each parameter on pipeline erosion, and based on the erosion influence weight, perform weighted calculation on the initial erosion state data to obtain the weighted erosion state data of each parameter; The weighted erosion state data is fused to obtain the initial erosion state data of the target pipeline.

3. The method for preventing corrosion of the inner wall of a pipe according to claim 2, characterized in that, The process of modeling the relationship between the erosion influence range and the initial erosion state data of the target pipeline to obtain an erosion evolution relationship data model includes: Based on the multiple erosion influence gradient sub-intervals, the initial erosion state data of the target pipeline is segmented to obtain the erosion state interval data corresponding to each erosion influence gradient sub-interval. The wear state interval data is discretized to obtain discretized wear state data points, and the wear influence intensity of each wear influence gradient sub-interval is discretized to obtain discretized wear influence intensity data points. By using a pre-set Markov model of erosion evolution, a quantitative analysis of the relationship between the discretized erosion state data points and the discretized erosion influence intensity data points is performed to obtain a data model of erosion evolution relationship.

4. The method for preventing corrosion of the inner wall of a pipe according to claim 1, characterized in that, The step of identifying the critical erosion thickness for pipeline failure using the erosion evolution relationship data model to obtain the critical erosion threshold for the pipeline includes: Curve fitting is performed on the erosion evolution relationship data model to generate the corresponding erosion thickness evolution relationship curve; The erosion thickness evolution curve is divided into a safe domain, an early warning domain, and a failure domain to obtain the corresponding safe domain, early warning domain, and failure domain. Based on the safety domain, the early warning domain, and the failure domain, the critical erosion threshold critical point is identified on the erosion thickness evolution curve to obtain the critical erosion critical point. The critical wear threshold is mapped to the wall thickness value to generate the corresponding critical wear threshold of the pipeline.

5. The method for preventing erosion of the inner wall of a pipe according to claim 1, characterized in that, The calculation of cumulative erosion amount on the erosion evolution relationship data model to obtain the cumulative erosion amount of the pipe inner wall includes: The corrosion trend of the corrosion evolution relationship data model is detected to obtain the pipeline corrosion trend index; Based on the pipeline erosion trend index, create an erosion amount calculation function corresponding to the erosion evolution relationship data model; Based on the erosion calculation function, the cumulative erosion of the inner wall of the pipeline is calculated using the erosion evolution relationship data model, and the cumulative erosion of the inner wall of the pipeline is obtained.

6. The method for preventing erosion of the inner wall of a pipe according to claim 1, characterized in that, The step of parameter mapping and encoding the critical erosion threshold of the pipeline and the cumulative erosion amount of the pipeline inner wall to generate a target pipeline erosion resistance evaluation vector includes: The K-Bins algorithm was used to calculate the cumulative erosion amount of the inner wall of the pipe using discrete intervals to obtain the target K value. Obtain the maximum and minimum values ​​of the cumulative abrasion amount on the inner wall of the pipe, and determine the corresponding discretization interval range based on the maximum and minimum values; K target interval ranges are determined based on the target K value and the discretized interval range, and the cumulative erosion amount of the inner wall of the pipe is mapped based on the K target interval ranges to obtain the cumulative erosion amount of the pipe corresponding to each target interval range. Multiple pipe erosion features are generated based on the cumulative pipe erosion amount corresponding to each target interval range. The multiple pipeline erosion characteristics and the pipeline critical erosion threshold are normalized and vectorized to obtain the target pipeline erosion resistance evaluation vector.

7. The method for preventing corrosion of the inner wall of a pipe according to claim 1, characterized in that, The step of inputting the target pipeline anti-corrosion evaluation vector into a pre-set anti-corrosion process decision model to output an anti-corrosion process scheme, thereby obtaining the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline, includes: The anti-corrosion evaluation vector of the target pipeline is input into a preset anti-corrosion process decision model, wherein the anti-corrosion process decision model includes a first bidirectional gated loop unit, a second gated loop unit, and a double-layer fully connected output layer. Obtain the anti-corrosion process standard parameters for small-diameter long-distance desulfurization pipelines in thermal power plants, and construct the corresponding standard anti-corrosion requirement vector based on the anti-corrosion process standard parameters. The target pipeline anti-corrosion evaluation vector is input into the first bidirectional gated loop unit for feature extraction to obtain the first corrosion feature vector, and the standard anti-corrosion requirement vector is input into the second gated loop unit for feature extraction to obtain the second process requirement feature vector. The first wear-resistant feature vector and the second process requirement feature vector are fused to obtain a fused anti-wear-resistant feature vector. The anti-corrosion construction process parameters are calculated and output through the dual-layer fully connected output layer, and the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline is determined based on the anti-corrosion construction process parameters.

8. A pipe inner wall anti-corrosion system, characterized in that, The system includes: The initial working condition inspection module is used to determine the erosion influence range of the inner wall of the desulfurization pipeline based on the actual working condition parameters of the pipeline, and to perform erosion pretreatment detection and working condition parameter analysis on the target small-diameter long-distance pipeline to obtain the initial erosion state data of the target pipeline. The evolution modeling module is used to model the relationship between the erosion influence range and the initial erosion state data of the target pipeline, and obtain an erosion evolution relationship data model. The threshold identification module is used to identify the critical erosion thickness of the pipeline failure from the erosion evolution relationship data model, and obtain the critical erosion threshold of the pipeline. The cumulative erosion calculation module is used to calculate the cumulative erosion amount on the erosion evolution relationship data model to obtain the cumulative erosion amount of the inner wall of the pipeline. The vector encoding module is used to perform parameter mapping encoding on the critical erosion threshold of the pipeline and the cumulative erosion amount of the inner wall of the pipeline to generate a target pipeline erosion resistance evaluation vector. The scheme decision module is used to input the anti-corrosion evaluation vector of the target pipeline into a preset anti-corrosion process decision model to output the anti-corrosion process scheme, thereby obtaining the target anti-corrosion construction scheme for the target small-diameter long-distance pipeline.

9. A device for preventing wear and corrosion on the inner wall of a pipeline, characterized in that, The device includes: a memory, a processor, and a pipe inner wall anti-corrosion program stored in the memory and executable on the processor, the pipe inner wall anti-corrosion program being configured to implement the steps of the pipe inner wall anti-corrosion method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a pipe inner wall anti-corrosion program, which, when executed by a processor, implements the steps of the pipe inner wall anti-corrosion method as described in any one of claims 1 to 7.

Images