A storage tank corrosion state analysis method based on a digital twin model

Abstract

This invention discloses a method for analyzing the corrosion status of storage tanks based on a digital twin model, belonging to the field of industrial twin technology. The method includes: establishing a base plate region division, recording the initial thickness and minimum allowable thickness, and calculating the allowable corrosion margin; real-time acquisition and preprocessing of acoustic emission, electrochemical, and environmental data, extracting monitoring indices to form a data package; determining the validity of the three modes to obtain evidence weights; mapping the thickness loss observations to logarithmic compression by region, and calculating the residuals with the prior thickness loss; generating prior uncertainty based on the observed dispersion superimposed with the natural growth rate, and using the residuals and uncertainty to calculate the calibration gain and recursively update the thickness, thickness loss, corrosion rate, and uncertainty; integrating thickness loss, corrosion rate, uncertainty, and weights to calculate a risk confidence index and classifying it for appropriate handling; converting the index into sampling intervals and wake-up thresholds to trigger encrypted sampling; calibrating the twin thickness status when verification conditions are met, achieving self-calibration and interpretable assessment.

Description

Technical Field

[0001] This invention relates to the field of industrial twin technology, and in particular to a method for analyzing the corrosion status of storage tanks based on digital twin models. Background Technology

[0002] In petrochemical and storage safety management, the corrosion status assessment of the bottom plate and contact area of ​​large vertical storage tanks is an important part of integrity management. In engineering, a combination of regular manual thickness measurement, visual inspection and local non-destructive testing is often used, supplemented by electrochemical monitoring, online acoustic emission monitoring and environmental parameter acquisition, to obtain characterization data on thickness changes, corrosion activity and service environment, thereby supporting the formulation of maintenance plans and risk classification management.

[0003] However, existing methods still have two limitations: First, the lack and drift of multimodal data fluctuate with operating conditions, and conventional fusion methods cannot quantify data availability into update intensity constraints. They also lack a mechanism to map multi-source indices into thickness loss observations, which affects the consistency of regional characterization. Second, state updates and risk assessments often lack recursive calibration of calibration gains and risk confidence index expressions for sliding window evidence accumulation, making it difficult to continuously output interpretable risk conclusions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for analyzing the corrosion status of storage tanks based on a digital twin model, which solves the problems in existing technologies where evidence weights are difficult to use to constrain the multimodal fusion strength and risk confidence indices are difficult to generate recursively based on calibration gains.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for analyzing the corrosion status of storage tanks based on a digital twin model. The method includes: establishing a tank bottom plate region, obtaining allowable corrosion margins; real-time acquisition and preprocessing of acoustic emission, electrochemical, and environmental data within the tank bottom plate region to extract three types of monitoring indices and form a monitoring evidence package; calculating evidence weights through validity determination; performing logarithmic compression mapping on the tank bottom plate region based on the monitoring evidence package to calculate the thickness loss observation and thickness loss residual; generating prior uncertainty by superimposing natural growth on the dispersion of the thickness loss observation; performing gain recursive calibration on the thickness status based on the thickness loss residual and prior uncertainty, and simultaneously recursively calculating the thickness uncertainty to output updated thickness, thickness loss, and corrosion rate; calculating a risk confidence index based on thickness loss, corrosion rate, thickness uncertainty, and evidence weights, and performing graded treatment determination; converting the risk confidence index into the sampling interval and wake-up threshold for the next cycle to trigger encrypted sampling; and calibrating the twin thickness status when the graded treatment determination meets the verification conditions.

[0008] As a preferred embodiment of the tank corrosion state analysis method based on digital twin model described in this invention, the step of obtaining the allowable corrosion margin includes: establishing a coordinate system for the tank bottom plate and dividing the region set according to structural characteristics, assigning region identifiers, recording the initial thickness and the minimum allowable thickness, and calculating the allowable corrosion margin for each tank bottom plate region.

[0009] As a preferred embodiment of the tank corrosion state analysis method based on a digital twin model as described in this invention, the following steps are taken: real-time acquisition and preprocessing of acoustic emission, electrochemical, and environmental data are performed in the tank bottom plate area to extract three types of monitoring indices and form a monitoring evidence package. Specifically, the acoustic emission waveform, electrochemical sequence, and environmental sequence are sampled in real time and a unified time stamp and sequence number are generated. The area identifier, time stamp, and sequence number are used as the primary key for preprocessing. Acoustic emission monitoring indices are constructed using the energy ratio of adjacent time windows plus nonlinear compression. Electrochemical monitoring indices and environmental monitoring indices are calculated using a fixed difference window and a sliding window to obtain the monitoring evidence package. The preprocessing includes bandpass filtering, amplitude limiting, and pulse interference removal for the acoustic emission waveform; moving median filtering for the electrochemical sequence; and missing value imputation and moving average smoothing for the environmental sequence.

[0010] As a preferred embodiment of the tank corrosion state analysis method based on digital twin model described in this invention, the following steps are taken: calculating the evidence weight by validity determination, and calculating the thickness loss observation by performing logarithmic compression mapping on the monitoring evidence package by region. Specifically, the validity of acoustic emission mode, electrochemical mode and environmental mode is determined and the number of valid modes is counted. The evidence weight is calculated based on the number of valid modes. The thickness loss observation is calculated by region based on the monitoring evidence package through ratio coupling and logarithmic compression mapping.

[0011] As a preferred embodiment of the tank corrosion state analysis method based on digital twin model described in this invention, the thickness loss residual includes: taking the difference between the initial thickness and the prior thickness as the prior thickness loss amount; taking the product of the evidence weight and the thickness loss observation amount as the intermediate loss value; and taking the difference between the intermediate loss value and the prior thickness loss amount as the thickness loss residual.

[0012] As a preferred embodiment of the tank corrosion state analysis method based on the digital twin model described in this invention, the steps of generating prior uncertainty by superimposing the natural growth amount on the dispersion of the thickness loss observation are as follows: converting the dispersion of the thickness loss observation into initial uncertainty; reading the updated thickness and the updated thickness uncertainty of the previous time step immediately adjacent to the previous unified time mark to generate the current prior thickness; and superimposing the natural growth amount on the updated thickness uncertainty of the previous time step to obtain the prior uncertainty.

[0013] As a preferred embodiment of the tank corrosion state analysis method based on a digital twin model as described in this invention, the following steps are involved: performing gain recursive calibration on the thickness state based on the thickness loss residual and prior uncertainty, and simultaneously recursively ...

[0014] As a preferred embodiment of the tank corrosion state analysis method based on a digital twin model as described in this invention, the steps for calculating the risk confidence index based on thickness loss, corrosion rate, thickness uncertainty, and evidence weight are as follows: The ratio of updated thickness loss to allowable corrosion margin is monotonically compressed to obtain a thickness loss evidence term; the corrosion rate is converted to an annual timescale to obtain a predicted loss, and its ratio with the allowable corrosion margin is monotonically compressed to obtain a corrosion rate evidence term; the ratio of updated uncertainty to the square of the allowable corrosion margin is used to obtain an uncertainty suppression term; the product of evidence weight, thickness loss evidence term, corrosion rate evidence term, and uncertainty suppression term is used as the risk evidence quantity at a single moment; and the risk confidence index is calculated by accumulating the risk evidence quantity at a single moment.

[0015] As a preferred embodiment of the tank corrosion status analysis method based on digital twin model described in this invention, the specific steps for determining the graded treatment are as follows: when the risk confidence index is not less than the first-level risk threshold, it is determined to be a high-level treatment; when the risk confidence index is less than the first-level risk threshold but not less than the second-level risk threshold, it is determined to be a medium-level treatment; when the risk confidence index is less than the second-level risk threshold but not less than the third-level risk threshold, it is determined to be a low-level treatment; and when the risk confidence index is less than the third-level risk threshold, it is determined to be a routine treatment.

[0016] As a preferred embodiment of the tank corrosion status analysis method based on digital twin model described in this invention, the steps are as follows: The risk confidence index is converted into the sampling interval and wake-up threshold for the next period to trigger encrypted sampling. When the graded treatment determination meets the verification conditions, the twin thickness status is calibrated. Specifically, the risk confidence index is converted into the sampling interval and wake-up threshold for the next period using a monotonically decreasing mapping; the maximum change amplitude of the three monitoring indices in adjacent sampling periods is used as the wake-up criterion and compared with the wake-up threshold for the next period to trigger encrypted sampling; when an area is determined to be subject to high-level or medium-level treatment, a verification task is generated and verification thickness measurement data is received, replacing the twin thickness status at the same verification time marker in the same area with the actual thickness value.

[0017] The beneficial effects of this invention are as follows: by using the thickness loss residual, it is possible to stably and quantitatively characterize the thickness loss changes in each region even when multi-source data is missing or fluctuating; by calibrating the gain and recursively calibrating the thickness status, and by combining the risk confidence index, it is possible to achieve continuous self-calibration of the thickness status and more robust risk classification, suppress the impact of occasional fluctuations and support disposal decisions. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a method for analyzing the corrosion status of storage tanks based on a digital twin model.

[0020] Figure 2 A flowchart for generating a monitoring evidence package.

[0021] Figure 3 A flowchart for outputting updated thickness loss and corrosion rate.

[0022] Figure 4 This is a flowchart for active sampling and verification calibration. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a method for analyzing the corrosion status of storage tanks based on a digital twin model, comprising the following steps:

[0027] S1. Establish the division of the tank bottom plate area and record the initial thickness and minimum allowable thickness of each area to obtain the allowable corrosion margin. In the area, collect and preprocess acoustic emission, electrochemical and environmental data in real time, extract three types of monitoring indices, and form a monitoring evidence package.

[0028] Obtain the planar geometric information of the tank bottom plate and establish a bottom plate coordinate system. Divide the bottom plate into multiple regions according to structural features (e.g., densification of the area near the ring plate and lap weld), forming a region set, where each region corresponds to a unique region identifier. For each region, enter the initial thickness from the construction records, factory data, and baseline thickness measurement records, and enter the minimum allowable thickness from the design documents to obtain the allowable corrosion allowance for the current region. Write the region identifier and thickness parameters into the region index table.

[0029] The allowable corrosion margin is the difference between the initial thickness and the minimum allowable thickness.

[0030] Acoustic emission sensors, electrochemical sensors, and environmental sensors are deployed within the monitoring area of ​​the base plate. The installation location of each sensor is mapped to a unique area identifier in the area set. When multiple sensors of the same type exist in the same area, the sensor outputs are aligned by timestamp and the average value is used as the input for the current mode of the current area. When the coverage of different mode sensors is not completely consistent, the available mode data is packaged and marked with missing flags using the area identifier as the convergence key.

[0031] The acoustic emission waveform, electrochemical sequence (e.g., potential), and environmental sequence (e.g., temperature, humidity, soil resistivity) are sampled in real time. A unified time stamp and sequence number are generated after each sampling data is generated. The unified time stamp comes from the time source (e.g., a local high-precision clock), and the sequence number is generated by an incrementing counter. The region identifier, time stamp, and sequence number are used as the primary key of the data structure.

[0032] Preprocessing is performed for different modes. Specifically, the acoustic emission waveform is bandpass filtered, limited, and pulse interference is removed; the electrochemical sequence is subjected to moving median filtering to suppress spikes; and the environmental sequence is imputed (using the most recent valid value) and smoothed by moving average.

[0033] Acoustic emission monitoring index, electrochemical monitoring index, and environmental monitoring index were extracted separately and constrained to... The acoustic emission monitoring index is constructed using the energy ratio of adjacent time windows plus nonlinear compression to enhance sensitivity to increased corrosion activity and suppress absolute amplitude drift. The expression is:

[0034] ;

[0035] ;

[0036] in, Indicates the first A unified time stamp, Represents the time variable of integration. This indicates the energy ratio between adjacent time windows. This indicates the length of the acoustic emission energy integration window. This represents the preprocessed acoustic emission waveform amplitude sequence. This indicates the acoustic emission monitoring index.

[0037] The electrochemical monitoring index is calculated using a fixed differential window (e.g., 10 min). For each unified time mark, the absolute value of the difference between the potential and the corresponding value obtained by moving the fixed differential window forward is taken, and the electrochemical monitoring index is obtained by normalizing it according to the median of the first 24 hours of samples. The environmental monitoring index is calculated using a fixed sliding window (e.g., 30 min). The sliding average values ​​of temperature, relative humidity, and dielectric resistivity are calculated respectively within the time interval of "moving the fixed sliding window forward from the current time mark to the current time mark", and the arithmetic mean is obtained after normalization.

[0038] The acoustic emission monitoring index, electrochemical monitoring index, and environmental monitoring index are combined to obtain a monitoring evidence package.

[0039] S2. Calculate the evidence weight by validity determination, and perform logarithmic compression mapping of the monitoring evidence package by region to obtain the thickness loss observation. Obtain the prior thickness loss amount from the twin state at the previous time step, and calculate the thickness loss residual.

[0040] The validity of acoustic emission, electrochemical, and environmental modes is determined. A mode is considered valid if it simultaneously meets the following criteria at the time marker: continuous and unordered time markers, a missing proportion not exceeding the missing threshold, and no prolonged saturation or constant sampling values. Otherwise, it is considered invalid. The number of valid modes among the three types is counted, and this count is taken as the number of valid modes. For a specific base plate region within a unified time marker corresponding to a statistical time window, the theoretical number of sampling points should be calculated based on the start and end times of the time window and the predetermined sampling period of the mode. The actual number of data points arriving within the time window is counted, and the difference between the theoretical number of sampling points and the actual number of valid data points is taken as the number of missing data points. The ratio of the number of missing data points to the theoretical number of sampling points is taken as the missing proportion of the time window. The in-window mean of a preprocessed key statistic within the time window is calculated, as is the in-window mean of the same key statistic in the immediately preceding statistical time window. The relative change between the current in-window mean and the previous in-window mean is taken as an indicator of the time window's drift.

[0041] It should be noted that the missing threshold is calculated by taking the first 24 hours after commissioning, using a predetermined sampling period as a benchmark, calculating the missing proportion for each statistical time window to form a modal missing proportion sequence, and taking the corresponding quantile of the missing proportion sequence as the missing threshold. For example, taking the 95th percentile as the missing threshold. If the missing threshold is lower than the 95th percentile, normal fluctuations will be misjudged as abnormal, resulting in excessive rejection of valid data. If the missing threshold is higher than the 95th percentile, the judgment will be relaxed, making it difficult to identify abnormal missing data in a timely manner.

[0042] Evidence weights are generated based on effective modality counts and monitoring evidence packages to control the strength of the effect of time window observations on twin updates. The expression is as follows:

[0043] ;

[0044] in, Indicates the region In a unified time stamp Weight of evidence Indicates the base plate area markings. Indicates the region In a unified time stamp The number of effective modes, area In a unified time stamp The proportion of missing time windows, Indicates the region In a unified time stamp The degree of drift in the time window, This indicates an indicator value; it takes the value 1 if the condition within the parentheses is true, and 0 otherwise.

[0045] For each region, based on the monitoring evidence package, a thickness loss observation is constructed using a mapping method combining ratio coupling and logarithmic compression, expressed as:

[0046] ;

[0047] in, Indicates the region In a unified time stamp The thickness loss observation, Indicates the region The scale parameter, Indicates the electrochemical monitoring index, This indicates an environmental monitoring index.

[0048] It should be noted that, The method involves selecting at least one manually confirmed inspection time, extracting the acoustic emission monitoring index, electrochemical monitoring index, and environmental monitoring index corresponding to the region for a period of time before and after the inspection time, calculating the original thickness loss observation quantity at the inspection time, and using the constraint that the actual thickness loss value obtained from the regional thickness measurement and the original thickness loss observation quantity are consistent, and obtaining the proportional coefficient that makes the two consistent as the regional scale parameter.

[0049] Create a regional status record table based on the regional index table. Write an initial record for each base plate region. The thickness of the initial record is taken as the initial thickness of the current region, and the corresponding timestamp is taken as the first received unified timestamp. When each new unified timestamp arrives, retrieve the status record immediately preceding the current unified timestamp in the regional status record table according to the same regional identifier. Read the previous updated thickness and the previous corrosion rate from the status record. When there is no previous updated thickness, take the current region's initial thickness as the current prior thickness to make the prior thickness loss zero. If it exists, read the thickness record immediately preceding the current unified timestamp from the historical records, take the previous thickness as the starting point, and combine it with the previous corrosion rate and the time interval between the two timestamps to obtain the predicted thickness loss increment. Take the difference between the previous thickness and the predicted thickness loss increment as the current prior thickness, and take the difference between the initial thickness and the current prior thickness as the prior thickness loss.

[0050] Generate thickness loss residuals. Specifically, the product of the evidence weight and the thickness loss observation is used as the intermediate loss value, and the difference between the intermediate loss value and the prior thickness loss value is used as the thickness loss residual.

[0051] S3. Generate prior uncertainty based on the dispersion of thickness loss observations and superimpose the natural growth amount. Perform gain recursive calibration on thickness state based on thickness loss residual and prior uncertainty, and simultaneously recursively derive thickness uncertainty, outputting updated thickness, thickness loss and corrosion rate.

[0052] A time-series state database with regional identifiers and unified time stamp indexes is established. The initial thickness of each region is written as the initial thickness state. Starting from the initial time stamp, monitoring evidence packets of the region are continuously collected for subsequent fixed time periods (e.g., within 30 minutes) to generate thickness loss observations. The dispersion of the thickness loss observations is calculated within the fixed time period, and the dispersion is converted into thickness dimensions as the initial uncertainty of the region. For the current unified time stamp, the state record of the region at the point immediately adjacent to the previous unified time stamp is read to obtain the updated thickness and the updated thickness uncertainty at the previous time stamp. The current prior thickness is generated starting from the updated thickness at the previous time stamp. Prior recursion is performed on the uncertainty, and the natural growth amount is added to the updated thickness uncertainty at the previous time stamp to obtain the current prior uncertainty.

[0053] It should be noted that, within the first 24 hours after the equipment is put into operation, the absolute value sequence of thickness loss residuals is statistically analyzed for each base plate area. The corresponding percentile (e.g., the 60th percentile) of the absolute value sequence is taken as the upper limit threshold of the residual for the area. If the value is lower than the 60th percentile, normal fluctuations will be more easily judged as exceeding the limit, resulting in an overly strict threshold. If the upper limit threshold of the residual is higher than the 60th percentile, the judgment of exceeding the limit will be too lenient, thus weakening the constraint on significant deviations. Between two adjacent unified time marks where the absolute value of the thickness loss residual does not exceed the upper limit threshold of the residual, the increment of uncertainty is calculated and the arithmetic mean is obtained. The arithmetic mean is taken as the natural growth of the thickness uncertainty of the area.

[0054] The current prior thickness is recursively calibrated using the thickness loss residual, expressed as:

[0055] ; ;

[0056] ;

[0057] in, Indicates the region In a unified time stamp Calibration gain at the location, Indicates the prior uncertainty. Indicates the region The level of observation noise, Indicates the thickness loss residual. Represents the residual scaling parameter. Indicates the current prior thickness. This indicates the updated thickness after calibration. This represents the updated uncertainty after calibration.

[0058] It should be noted that, The observation noise level is determined by generating thickness loss observations for each base plate area at a unified time mark within a fixed time period after the equipment is put into operation (e.g., the first 24 hours), forming an observation sequence. The observation difference sequence between two adjacent unified time marks is calculated, and the median absolute deviation of the observation difference sequence is taken as the statistical measure of regional observation fluctuation. Finally, the statistical measure is converted into the thickness dimension as the observation noise level. The thickness loss residuals are generated by uniformly marking each base plate region with a time stamp, forming a sequence of absolute residual values ​​for the region. The corresponding quantile of the absolute residual value sequence (e.g., the 90th percentile) is taken as the residual scale parameter for the region. If the quantile is lower than the 90th percentile, it is easy to be pulled by common fluctuations, resulting in a smaller scale and making the residuals more likely to be judged as overscale. If the quantile is higher than the 90th percentile, it is easy to be raised by a small number of extreme deviations, resulting in a larger scale and making overscale residuals difficult to identify.

[0059] The initial thickness of the region is read from the region index table, and the difference between the initial thickness and the calibrated updated thickness is used as the updated thickness loss. The time series of the current region thickness loss is extracted from the time series state library at the most recent fixed time (e.g., 24 hours), the growth rate within the fixed time is calculated, and it is converted into an annualized rate as the corrosion rate. If the corrosion rate is negative due to noise, it is set to zero. The updated thickness, updated thickness loss, corrosion rate, update uncertainty, corresponding unified time stamp, and region identifier are written into the time series state library.

[0060] S4. Calculate the risk confidence index based on thickness loss, corrosion rate, thickness uncertainty, and evidence weight, and make a graded disposal judgment.

[0061] The ratio of updated thickness loss to allowable corrosion margin, after monotonically compression, is used as the thickness loss evidence term; after converting 24 hours to an annual time scale, the ratio of the predicted loss to allowable corrosion margin within the annual time scale, after monotonically compression, is used as the corrosion rate evidence term; the proportion of updated uncertainty to the square of allowable corrosion margin is used as the uncertainty suppression term; and the product of the evidence weight, thickness loss evidence term, corrosion rate evidence term, and uncertainty suppression term is used as the risk evidence quantity at a single moment.

[0062] It should be noted that the predicted loss refers to the corrosion rate of the area read at the current unified time mark, and the corrosion rate is converted into the thickness loss increment within 24 hours according to the annual time length corresponding to 24 hours as the predicted loss.

[0063] To reduce the impact of occasional fluctuations on the judgment, the amount of risk evidence at a single moment within the statistical time window is accumulated, and the risk confidence index is calculated, as expressed by:

[0064] ;

[0065] ;

[0066] in, Indicates the level of risk intensity. Indicates the number of valid time points. This represents a uniform time marker for each time interval falling within a 24-hour sliding window. Add them up one by one. Indicates the first in the window A unified time stamp is The discrete time points within the previous 24-hour range, Indicates the region In a unified time stamp The amount of risk evidence at a single moment in time. Indicates the area In a unified time stamp Risk confidence index at the location.

[0067] The risk confidence index is used to determine the level of handling. Specifically, when the risk confidence index is not less than the first-level risk threshold, it is determined to be a high-level handling; when the risk confidence index is less than the first-level risk threshold but not less than the second-level risk threshold, it is determined to be a medium-level handling; when the risk confidence index is less than the second-level risk threshold but not less than the third-level risk threshold, it is determined to be a low-level handling; and when the risk confidence index is less than the third-level risk threshold, it is determined to be a routine handling.

[0068] It should be noted that within the first predetermined period after the equipment is put into operation (e.g., 24 hours), the risk confidence index of the base plate area is continuously calculated according to a unified time mark to form a historical series. The corresponding quantiles of the historical series are used as the risk level 1 threshold, risk level 2 threshold, and risk level 3 threshold. For example, the 90th percentile of the historical series is used as the risk level 1 threshold, the 70th percentile as the risk level 2 threshold, and the 50th percentile as the risk level 3 threshold. If the risk level 1 threshold is lower than the 90th percentile, the high-risk judgment will be too sensitive and frequently triggered. If it is higher than the 90th percentile, the high-risk judgment will be too sluggish and difficult to enter level 1 in time. If the risk level 2 threshold is lower than the 70th percentile, the coverage of medium risk will be expanded, resulting in a decrease in the classification distinction. If it is higher than the 70th percentile, the medium-risk range will be compressed, causing a slight increase to be prematurely raised to high risk. If the risk level 3 threshold is lower than the 50th percentile, the low-risk range will be too narrow, weakening the representation of normal operation. If it is higher than the 50th percentile, the low-risk range will be too wide, making it difficult for general fluctuations to be raised to the level that requires attention.

[0069] S5. Convert the risk confidence index into the sampling interval and wake-up threshold for the next cycle, trigger encrypted sampling, and calibrate the twin thickness status when the graded disposal judgment meets the review conditions.

[0070] To ensure consistency, higher-risk areas require more frequent sampling and are more sensitive to wake-up, a monotonically decreasing mapping is used to convert the risk confidence index into sampling interval and wake-up threshold, expressed as:

[0071] ;

[0072] ;

[0073] in, Indicates the maximum allowed sampling interval. Indicates the minimum allowed sampling interval. Indicates the region In a unified time stamp The next cycle sampling interval is calculated. Indicates the minimum wake-up threshold. Indicates the maximum wake-up threshold. Indicates the region In a unified time stamp Calculate the wake-up threshold for the next cycle.

[0074] It should be noted that the minimum sampling interval is determined by adjusting the sampling interval in stages and running it for a period of time, comparing the peak cache usage, packet loss ratio, and reporting latency statistics between the two periods before and after the period. When no cache overflow occurs and the latter three indicators are not worse than the former, the current sampling interval is determined to be usable, and the smallest interval that meets the determination criteria is taken as the minimum sampling interval. The maximum sampling interval is determined by reading the update thickness, minimum allowable thickness, and corrosion rate of each region at the current time, using the difference between the update thickness and the minimum allowable thickness as the remaining margin, and converting the remaining margin and corrosion rate to obtain the maximum time length that will not cross the minimum allowable thickness before the next sampling, and taking the minimum of the maximum time lengths of each region as the maximum sampling interval. The minimum wake-up threshold and the maximum wake-up threshold are determined by obtaining the maximum value of the three types of monitoring indices in each sampling period to form a historical sequence of the region, and using the corresponding quantile of the historical sequence as the minimum wake-up threshold and the maximum wake-up threshold, such as the 70th percentile as the minimum wake-up threshold and the 95th percentile as the maximum wake-up threshold. If the minimum wake-up threshold and the maximum wake-up threshold are lower than the corresponding quantile, the wake-up will be too sensitive and frequently triggered, and if they are higher than the corresponding quantile, the wake-up will be too sluggish and difficult to trigger in a timely manner.

[0075] The next sampling interval is used as the timed sampling period for the region. The wake-up criteria for the region are continuously calculated in each sampling period. When the wake-up criteria reach the wake-up threshold of the next period, an encrypted sampling and instant reporting are immediately triggered.

[0076] It should be noted that the wake-up criterion is determined by reading the acoustic emission monitoring index, electrochemical monitoring index, and environmental monitoring index formed in the region during the current sampling period at the end of each sampling period, and simultaneously reading the three types of monitoring indices corresponding to the previous sampling period. The change range between the current sampling period value and the previous sampling period value is calculated for each type of monitoring index, and then the maximum value among the three types of change ranges is taken as the wake-up criterion for the current sampling period.

[0077] When a region is determined to be subject to high-level or medium-level handling at the current unified time marker, a review task is generated. The review task includes the region identifier, the suggested review location, and the suggested review time window.

[0078] Based on the verification task, verification thickness measurement data is obtained. The verification thickness measurement data includes region identifier, verification time marker and actual thickness value of the region. The actual thickness value of the region is bound to the twin thickness state (the updated thickness after calibration) at the same time marker. The twin thickness state of the region at the verification time marker is directly replaced with the actual thickness value. The uncertainty of the region at the verification time marker is reset to the thickness order of magnitude corresponding to the observation noise level. The calibrated thickness state and uncertainty are written into the time series state library.

[0079] The latest thickness state record immediately preceding the current unified time mark is read from the time-series state database as the updated thickness at the previous moment. If a verification forced calibration record exists, the thickness record at the same region and the same unified time mark is updated to the verification thickness value and written into the time-series state database, thereby ensuring that the thickness source used to generate the prior thickness loss in the next cycle is the twin state after verification and calibration.

[0080] In summary, this invention achieves stable and quantitative characterization of thickness loss changes in each region even when multi-source data is missing or fluctuating, through thickness loss residuals; by calibrating the gain and recursively calibrating the thickness state, and combining it with the risk confidence index, it achieves continuous self-calibration of the thickness state and more robust risk classification, suppressing the impact of occasional fluctuations and supporting disposal decisions.

[0081] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for analyzing the corrosion status of storage tanks based on a digital twin model, characterized in that: include, Establish the tank bottom plate area, obtain the allowable corrosion margin, collect and preprocess acoustic emission, electrochemical and environmental data in real time within the tank bottom plate area, extract three types of monitoring indices, and form a monitoring evidence package. The evidence weight is calculated by determining the validity of the evidence package, and the thickness loss observation and thickness loss residual are calculated by performing a logarithmic compression mapping on the bottom plate area of ​​the tank based on the monitoring evidence package. Based on the dispersion of the thickness loss observations and the natural growth, a priori uncertainty is generated. The thickness state is then calibrated by gain recursion based on the thickness loss residual and the prior uncertainty, and the thickness uncertainty is simultaneously recursively calculated. The updated thickness, thickness loss, and corrosion rate are then output. The risk confidence index is calculated based on thickness loss, corrosion rate, thickness uncertainty, and evidence weight, and a graded treatment decision is made accordingly. The risk confidence index is converted into the sampling interval and wake-up threshold for the next cycle, triggering encrypted sampling. When the graded treatment judgment meets the review conditions, the twin thickness status is calibrated.

2. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 1, characterized in that: The process of obtaining the allowable corrosion margin includes establishing a coordinate system for the tank bottom plate and dividing the area into sets according to structural characteristics, assigning area identifiers, entering the initial thickness and the minimum allowable thickness, and calculating the allowable corrosion margin for each tank bottom plate area.

3. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 2, characterized in that: The process involves real-time acquisition and preprocessing of acoustic emission, electrochemical, and environmental data within the tank bottom area to extract three types of monitoring indices and form a monitoring evidence package. The specific steps are as follows: Real-time sampling of acoustic emission waveforms, electrochemical sequences, and environmental sequences generates unified timestamps and sequence numbers. Preprocessing is performed using region identifiers, timestamps, and sequence numbers as primary keys. Acoustic emission monitoring indices were constructed by using the energy ratio of adjacent time windows plus nonlinear compression. Electrochemical monitoring indices and environmental monitoring indices were calculated using fixed differential windows and sliding windows to obtain monitoring evidence packages. The preprocessing includes bandpass filtering, amplitude limiting, and pulse interference removal of the acoustic emission waveform; moving median filtering of the electrochemical sequence; and missing value imputation and moving average smoothing of the environmental sequence.

4. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 3, characterized in that: The specific steps are as follows: Calculating evidence weights through validity determination, performing logarithmic compression mapping on the monitoring evidence package by region, and calculating the thickness loss observation. The validity of acoustic emission modes, electrochemical modes, and environmental modes is determined and the number of valid modes is counted. The evidence weight is calculated based on the number of valid modes. Based on the monitoring evidence package, thickness loss observations are calculated by region using ratio coupling and logarithmic compression mapping.

5. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 4, characterized in that: The thickness loss residual includes using the difference between the initial thickness and the prior thickness as the prior thickness loss amount; The product of the evidence weight and the thickness loss observation is used as the median loss value, and the difference between the median loss value and the prior thickness loss value is used as the thickness loss residual.

6. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 5, characterized in that: The specific steps for generating prior uncertainty by superimposing the natural growth amount on the dispersion of the thickness loss observations are as follows: The dispersion of thickness loss observations is converted into initial uncertainty. The current prior thickness is generated by reading the thickness updated at the previous time point immediately adjacent to the previous unified time mark and the uncertainty of the thickness updated at the previous time point. The prior uncertainty is obtained by superimposing the natural growth amount on the uncertainty of the thickness updated at the previous time point.

7. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 6, characterized in that: The method involves performing gain recursive calibration on the thickness state based on the thickness loss residual and prior uncertainty, and simultaneously recursively calculating the thickness uncertainty to output the updated thickness, thickness loss, and corrosion rate. The specific steps are as follows: The calibration gain is calculated based on the thickness loss residual and prior uncertainty. Recursive calibration is performed using the calibration gain to obtain the updated thickness, and the uncertainty is updated simultaneously. The difference between the initial thickness and the updated thickness is taken as the updated thickness loss, and the growth rate of the thickness loss is converted into the corrosion rate.

8. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 7, characterized in that: The specific steps for calculating the risk confidence index based on thickness loss, corrosion rate, thickness uncertainty, and evidence weight are as follows: The thickness loss evidence term is obtained by monotonically compressing the ratio of updated thickness loss to allowable corrosion margin. The corrosion rate is converted to an annual time scale to obtain the predicted loss, and the ratio of this to the allowable corrosion margin is monotonically compressed to obtain the corrosion rate evidence term. The uncertainty suppression term is obtained by taking the proportion of the update uncertainty relative to the square of the allowable corrosion margin. The product of the evidence weight, thickness loss evidence term, corrosion rate evidence term, and uncertainty suppression term is used as the risk evidence quantity at a single moment. The risk confidence index is calculated by accumulating the amount of risk evidence at a single moment.

9. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 8, characterized in that: The specific steps for determining the tiered treatment are as follows: When the risk confidence index is not less than the first-level risk threshold, it is judged as a high-level disposal. When the risk confidence index is less than the first-level risk threshold and not less than the second-level risk threshold, it is judged as a medium-level disposal. When the risk confidence index is less than the risk level 2 threshold and not less than the risk level 3 threshold, it is judged as a low-level disposal. When the risk confidence index is less than the risk level 3 threshold, it is determined to be a routine procedure.

10. The method for analyzing the corrosion status of storage tanks based on a digital twin model as described in claim 9, characterized in that: The process involves converting the risk confidence index into the sampling interval and wake-up threshold for the next cycle, triggering encrypted sampling, and calibrating the twin thickness state when the tiered handling determination meets the review conditions. The specific steps are as follows: A monotonically decreasing mapping is used to convert the risk confidence index into the next period sampling interval and the next period wake-up threshold; The maximum change in the three monitoring indices of adjacent sampling periods is used as the wake-up criterion and compared with the wake-up threshold of the next period to trigger encrypted sampling. When an area is determined to be subject to high-level or medium-level treatment, a review task is generated and review thickness measurement data is received. The actual thickness value replaces the twin thickness status at the same review time marker in the same area.

Images